0% found this document useful (0 votes)

24 views

Survey of Security Protocols and Vulnerabilities I

Uploaded by

Kratika Choudhary

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

24 views

Survey of Security Protocols and Vulnerabilities I

Uploaded by

Kratika Choudhary

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 27

This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3066778, IEEE Access

Digital Object Identifier 10.1109/ACCESS.2017.DOI

Survey of Security Protocols and

Vulnerabilities in Unmanned Aerial
Vehicles
ARSLAN SHAFIQUE1 , ABID MEHMOOD 2 , MOURAD ELHADEF 2
1
Department of Electrical Engineering, Riphah International University, Pakistan.
2
Department of Computer Science and Information Technology, Abu Dhabi University, UAE.
Corresponding author:Arslan Shafique (e-mail:[email protected])

ABSTRACT With the rapid growth in technology, the use of Unmanned Aerial Vehicles (UAVs) have
increased in civil and military applications including rescue operations, disaster recovery, and military
operations. Despite the utility and advantages of UAVs, they may lead to major security breaches in the
context of hardware, software, and communication channel, due their ease of use and availability. UAVs are
vulnerable to various types of attacks such as spoofing, false data injection, jamming, fuzzing, availability,
confidentiality, and integrity attacks. To overcome these security threats, researchers have been investigating
strong security protocols to keep UAVs safe from the attackers. Nevertheless, there are many flaws in the
developed protocols which can be exploited by hackers. Therefore, it is becomes crucial to study and analyze
the existing security protocols used in UAVs to discover and address their vulnerabilities and weaknesses.
The purpose of this study is to explore the vulnerabilities in the security protocols and propose guidelines
to improve the security and provide future research directions.

INDEX TERMS Unmanned Aerial Vehicles (UAVs), Security, Vulnerabilities, Attacks, Drones, Security
Threats.

I. INTRODUCTION include the ability to carry heavy payloads, detec-

HE Unmanned Aerial Vehicles (UAVs), also tion of mines, and scanning of unethical/unwanted
T known as drones, have the capability of flying
with and without a human pilot, and can be remotely
activities in certain areas. All of these features are
made possible thanks to the recent advancement
controlled by wireless connections such as WiFi or in software an hardware technologies. In particu-
radio. Other flying objects such as quadcopters and lar, Artificial Intelligence and Machine Learning
gliders can also be classified as UAVs. Recently, have been playing a vital role in enabling UAVs
military has intensified the use of UAVs for critical to perform complex and sophisticated tasks.This
operations in order to reduce the exposure of their has resulted in making UAVs highly susceptible to
valued human resources in high-risk environments. security threats.
Apart from military uses, there are several other UAVs are prone to different security threats and
applications in the private sectors as well. These can be attacked in various ways. The consequences
applications include search and rescue missions, of some attacks can be devastating. Even the expen-
surveillance, fire-fighting, courier services, ... [1]. sive professional ones are not secure [2], [3]. Some
UAV applications are exponentially expanding be- of the attacks focus on stealing information affecting
cause of its rapid movement, low maintenance cost, hence its integrity, confidentiality and availability
and its ability to float and monitor real-time environ- [4]. UAVs carry and transfer a lot of information
ments. using their communication channels. Such informa-
Figure 1 shows a sample from the various appli- tion exchange should be secured. The information
cations of UAVs. The advanced features of UAVs can be in different forms such as images, text, audio

VOLUME 4, 2016 1

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2021.3066778, IEEE Access

FIGURE 1: Applications of UAVs.

and videos. Several encryption schemes are avail- A. UAV SYSTEM

able in the literature by which one can encrypt the Knowing the components of UAV systems and how
sensitive information [5], [6]. the information flows between them is crucial in
The fundamental security concern about the com- analyzing their vulnerabilities. In addition, UAVs
munication protocols is how to secure the data that are highly exposed to technical system failures. The
is being sent over an insecure connection such as basic components of UAVs and the information
WiFi. UAVs normally send the data to the ground flows between them are described in Figure 2.
station over a wireless link which can be easily
targeted by the attackers. To prevent the data from
being intercepted by the attackers, the data should
be protected. One common mechanism used to pro-
tect the data is use of encryption. For instance, the
Advanced Encryption Standard (AES) is one of of
the secure mechanisms used nowadays. However, it
cannot be used efficiently in real-time applications
because of the communication overhead, especially
where the data transfer rate is very high [7].
The other major concern about the security of
UAVs is an attacker interfering with the UAVs in
various ways to either take control of the UAVs or
disabling the communication between the UAVs and
the ground control station (GCS). Different attacks,
which can be launched against UAVs to either take
control or disable the communication, include jam-
ming [8], spoofing [9], and false data injection [10]
attacks.
In this section, we will describe the different FIGURE 2: UAV components and information
components of a UAV system and how the infor- flows.
mation flows between these components. We will
also highlight the motivations behind this study and
The UAV base system is built for UAVs which is
discuss the contributions of this survey paper.
responsible for linking all the components together.
It is used for inter-component communications, and
2 VOLUME 4, 2016

for controlling the sensors and the communication/- Most of those reviews are not comprehensive and
navigation systems. The UAV base system is also addressed only few security concerns. In addition,
used for the integration of the optional components, the protocols discussed are not properly scrutinized
e.g., the weapon systems. The UAV sensor system for security and vulnerabilities. To the best of our
is composed of all the sensory equipment, with knowledge, there is no exhaustive survey in the
integrated pre-processing functionalities, such as literature that overviews all possible security threats
sensors with cameras, GPS, and radars. The avionic and the vulnerabilities that exist in the security
system is responsible for executing control com- protocols used in UAVs. Our survey explores two
mands received from the controller such as engine major aspects which are:
commands, spoilers, flaps, and stabilizers. UAVs • Security protocols used to secure the UAVs.
rely mainly on a wireless communication system • Vulnerabilities in the existing security proto-
which can either be a direct line of sight communi- cols.
cation or indirect communication through satellites. The first generations of drones were initially used
To attack a UAV, an attacker must influence the mainly by the military for relief operations such
UAV externally unless the attacker has a physical as delivering of goods and rescue operations as
access to its system. Due to the wireless nature shown in Figure 1. Nowadays, UAVs are widely
of UAVs’ communication system, they are highly used for civil applications such as capturing images
dependent on external inputs. This provides various and videos of different areas. Due to the ease of
input channels for an attacker to attack their sys- availability and the tremendous increase in their
tems. As shown in Figure 2, information flows be- applications, UAVs are becoming a very attractive
tween the UAV and its environment through various target for cyber criminals. As a matter of fact, it
channels. The bidirectional communication between is riskier to execute sensitive operations like res-
the communication system and the GCS is the most cue operations via UAVs without a strong security
exposed channel that can be used in an attack. The protocol. The protocols that secure the UAVs must
second most critical component is the information be strong enough to resist different types of cyber-
flow from the environment to the UAV sensors. attacks on availability, such as denial of service
These two links are very receptive to manipulations. attacks (DoS), or confidentiality and integrity at-
Furthermore, the reliability of the sensors cannot be tacks. The vulnerabilities that exist in the security
trusted. The key to control a UAV during a cyber- protocols can lead to different cyber-threats. To
attack is the host’s knowledge about the receptive- prevent the UAVs from cyber-threats, the security
ness of the components to the commands. protocols used in the UAVs should be free of vul-
There are various cyber-attacks which could be nerabilities. Several security protocols have been
used to exploit vulnerabilities of existing UAV sys- proposed recently and some of them offer a very
tems. GPS jamming and spoofing are one of them. good sense of security. However, it remains crucial
Jamming refers to the process of preventing the host to find out the vulnerabilities in the existing security
from receiving the normal signal. While, spoofing protocols. Exploring vulnerabilities and providing
encompasses the process of sending a malicious solutions to the corresponding vulnerabilities is the
signal to foul the host to consider it as a legitimate basic motivation behind the proposed survey.
one. Another form of intrusion attacks that exploits
the GPS functionalities and that can be used to C. CONTRIBUTIONS OF THIS SURVEY
attack a UAV is the “GPS spoofing attack”. I such Several survey papers were published recently high-
intrusion, the attacker can overlay the GPS signal by lighting issues related to the security of UAVs such
a spoofed GPS system with a stronger signal. This as, secure communication, intrusion detection, and
leads to the false estimation of the UAV’s current security of the routing protocols used by UAVs
position. Attacks on UAVs have been continuously [11], [16]–[19]. Most of these survey papers have
investigated by researchers in order to explore them discussed UAVs’ security in a very sophisticated
in depth and to develop countermeasures that will manner, However, to the best of our knowledge,
make the future generations of UAVs more secure only few of the existing survey papers have reflected
and less prone to attacks. upon the vulnerabilities of the UAVs’ security proto-
cols. In [16], security vulnerabilities of UAVs were
B. MOTIVATIONS discussed. The survey paper highlighted the vulner-
In the last decades, several review papers related abilities that can be used to attack UAVs such as
to the security of UAVs were published [11]–[15]. attacks on the communication and control systems,
VOLUME 4, 2016 3

spoofing, and jamming attacks. However, the paper to security, safety, and privacy, which must be ad-
did not discuss on how to secure UAVs from the dressed before fully relying on them [20], [21].
attacks. The drones used by cyber criminals or terrorists
Zhi et. al discussed security and privacy issues of can invade the privacy of the individuals as well
UAVs in [17]. The two major categories of attacks as the privacy and safety of the general public. A
were covered: i) spoofing attack, and ii) WiFi at- number of drone properties are utilized in attacks
tacks. However, the survey paper is not comprehen- that include high-level operations and unauthorized
sive as it lacks in-depth discussions on UAVs’ vul- inspections. Drone utilization involves unauthorized
nerabilities and attacks. In [18], secure communica- spying on individuals, resulting in safety and pri-
tion protocols used in UAVs and their vulnerabilities vacy issues [22]. Drones must not be used to capture
were discussed. Nevertheless, the review paper does images of individuals and record their videos with-
not discuss the techniques used to prevent cyber- out their prior consent. The use of drones must be
attacks. A comparison of the proposed survey and prohibited in residential areas and public properties
the most recent review papers related to UAVs’ that cause privacy issues, as the images captured by
security is given in Table 1. In our survey paper, we these drones may be used for illegal purposes that
have covered all the important aspects such as UAV include scamming. Most of the drones nowadays
security, vulnerabilities in the existing protocols, are WiFi enabled so that the captured video can be
and their countermeasures. broadcasted to smart devices. Some drones also use
The contributions of this survey are as follows: WiFi for remote control using smart mobile devices.
• We have done an in-depth literature review of As the WiFi connections are not strongly protected
the past research work related to the security of due to weak passwords, the attackers can easily
UAVs. access the WiFi and interfere with the communica-
• We have analyzed the vulnerabilities that exist tion, especially when there is no encryption protocol
in the security protocols and provided possible applied to WiFi passwords. [23].
solutions to overcome the issues of the existing Attackers can also use their unauthorized UAVs
security protocols. to destroy the authorized UAVs by performing phys-
• We have explained and analysed how different ical collisions. As the unauthorized and authorized
vulnerabilities such as WiFi insecurity, jam- UAVs come across often, it is crucial to avoid any
ming attacks, fuzzing attacks, ..., can be used collision between them. Several modes were investi-
to attack the UAVs. gated by researchers to prevent UAVs from colliding
• We have also highlighted the vulnerabilities in [24]. The purpose was to design a UAV-Sense-and-
the packet forwarding and routing protocols Avoid (SAA) system to sense and elude the obsta-
used in UAVs and how they can be a threat to cles placed by the attackers. Another mode for SAA
the security. was also introduced by Barfield in [25]. Barfield
• Last, but not the least, we have highlighted the proposed an autonomous collision avoidance system
number of possible future research directions that is fully capable to protect the UAVs from any
to enhance the security of the UAVs. unnecessary accidents. Practical trials showed no
In this paper, we will highlight the existing se- failure during the tested the flights. The collision
curity protocols for UAVs. We will also analyze avoidance algorithms were designed to accomplish
the vulnerabilities in the existing protocols. This some important challenges that include Individual
research study is organized as follows: Section II Collision Avoidance (ICA) and Group Collision
is devoted to the survey of the existing security Avoidance (GCA) [26]. Another method was pre-
protocols which are designed in the past few years sented by Yang et al. in [27] was based on a UAV 3D
to enhance the security of UAVs. In Section III, the path planning, which consists of locating a collision-
vulnerabilities in the UAVs’ security protocols are free path in a 3D cluttered environment considering
discussed. In Section V, we will propose potential geometric, physical, and temporal constraints. Dif-
research directions and Section VI will conclude the ferent obstacle-collision avoidance methods were
paper. also presented to overcome any obstacle facing the
UAVs. In [28], Ueno et al. presented a new al-
II. SECURITY IN UAVS gorithm that enables an UAV to accurately locate
The use of drones brought several advantages that objects in its vicinity. In [29], Brandt et al. stated
include commercial gains as well as personal ben- that quad-rotors are more suitable to operate indoors
efits. However, there are several drawbacks related due to their flexible and fine-controlled operations
4 VOLUME 4, 2016

TABLE 1: Comparison of the current survey with existing review papers.

Categories Sub-categories Krishna Zhi Khan Sharma McCoy Current

discussed et al. [16] et al. [17] et al. [18] et al. [11] et al. [19] Survey

Symmetric security protocols X X

Secure communication Asymmetric security protocols X X

Authentication protocols X X X X

Learning based X X
Intrusion detection systems

Rules based X X

Security of Routing protocols X X

Packet forwarding X

GPS spoofing X X X X X X
Spoofing attacks

False data injection X X X X

WiFi insecurity X X X X X

Jamming attacks X X X X X

Attacks on control systems X X X

Fuzzing attacks X X

Malicious UAVs detection X X X

in small and confined areas. Furthermore, an algo-

rithm was presented by Israelsen et al. to manually
control UAVs using automatic Obstacle Collision
Avoidance (OCA) [30].
In addition to protecting the UAVs from colli-
sions, it is very important to protect the commu-
nication between the UAVs and the GCS. For se-
cure communication, several security protocols are
proposed. The use of security protocols depends
on the nature of the application which we will
discuss later. The security protocols for UAVs can
be broadly classified into three categories: i) secure
communication, ii) physical layer security, and iii)
intrusion detection system. The schematic diagram
for the security protocols used in UAVs are shown
in Figure 3.
FIGURE 3: Security categories in UAVs.

A. SECURE COMMUNICATION IN UAVS

UAVs can be utilized for the surveillance of a large
area without any additional help from the network portant information. This exchange of information
infrastructure. During the flight, UAVs communi- creates new challenges due to the dynamic topol-
cate with the GCS and continuously exchange im- ogy. UAVs are frequently used for data transmission
VOLUME 4, 2016 5

from one node to the GCS. The transmitted data can created which can be secured by using an encryption
be attacked in various ways. In most of the mili- scheme.
tary applications, the sensitive information is sent OTP-based encryption schemes have some draw-
between two authentic users through the wireless backs. For instance, the key size must be equal to
communication channels. As the wireless channel is the length of the data. If we want to send a large
an insecure medium, it is quite possible to access size data, we must share the key with the receiver.
the information by initiating cyber-attacks such as Hence, key distribution becomes a problem as it
integrity, availability, and confidentiality attacks. To consumes a lot of bandwidth. Moreover, the key can
protect the information from the attackers, different be only used once, which means for each secure
types of security protocols are used to secure the transmission there is a need for a new key [31]. The
transmission and authenticate the users. For exam- scheme proposed in [31] can be improved in terms
ple, symmetric and asymmetric security protocols of security by applying some robust transformation
are used to secure the communication between the techniques such as discrete wavelet and discrete
UAV and the GCS. In symmetric security proto- cosine transforms. These techniques first convert
cols, only one private shared key can be used for the original message into different frequency coeffi-
the encryption and decryption process. While in cients which are completely different from the orig-
asymmetric security protocols, two different keys, inal message. Moreover, transformation performed
one private and one public, are used. Public key is using frequency coefficients is faster as compared
used for encryption whereas private key is used for to transformation performed directly on the original
decryption. These two types of security protocols message [32], [33].
are further discussed in sections II-A1 and II-A2. In [34], a chaotic Lorenz system is used that
Section II-A2 also highlights the authentication pro- encrypts and decrypts the original and transformed
tocols that are used to verify the identity the trans- messages, respectively. Chaotic Lorenze systems
mitter, i.e., guarantee that the received message is have long-term unpredictability and can generate
authentic and was not sent by an attacker. Section more randomness with minor changes in the seed
II-A3 presents lightweight authentication protocols values. The UAV collects the data from the sensors
which are used where less memory and low compu- and camera, and passes them to the Lorenz chaotic
tational complexity are required. based encoder. It does not directly encrypt the plain
message. All the information is converted into bits
1) Cryptographic Symmetric Security Protocols and then undergoes encryption. The bits are contin-
To ensure integrity, confidentiality, and availability, uously encrypted till the end of the original data.
cryptographic protocols are frequently used. In par- Following the encryption process, the UAV sends
ticular, symmetric protocols are used to protect the the encrypted information to the receiver which
sensitive data such as text, images, audio, and video. decrypts it by applying the reverse process of the
In symmetric security protocols, the same key is chaotic Lorenze system. The proposed encryption
used to encrypt and decrypt the information, i.e., scheme is symmetric in nature which means that the
the transmitter and receiver must have identical keys receiver uses the same key by which the original
to access the original information. One time pad data was encrypted by the transmitter. However,
(OTP) is an example of symmetric security proto- the proposed protocol [34] has some weaknesses
cols, which is often used to secure the transmission. as well. For instance, there is no data scrambling
To secure the data, OTP requires the same key size process in the proposed technique. In fact, the se-
as the size of the data. For example, in the case curity of any encryption scheme depends on both
of images, if an image contains M rows and N confusion (scrambling) and diffusion [35].
columns of pixels, the key must be equal to the
length of the original image, i.e., M ×N . In [31], the 2) Cryptographic Asymmetric Security
security of the wireless communication MAVlink Protocols
is enhanced using OTP encryption. To secure the In asymmetric security protocols, two different keys
transmitted data, an encryption-decryption function are used. One is the public key, while the other is the
is used. There are several commands to control the private key. The user at the transmitter and receiver
UAVs, such as start UAV, takeoff command, and ends encrypts and decrypts the information using
autopilot enable. All these commands are in the the public key and private key, respectively. The
form of bits which can be represented by 0 or 1. secrecy of the public key is not necessary because
By combining the different bits, a long text can be if anyone encrypts the information using the public
6 VOLUME 4, 2016

key, it cannot be decrypted with that same public the basis of their characteristics to make the ADS-
key. To retrieve the information, a secret (private) B more secure. After evaluating the cryptography in
key must be used instead of the public key. In ADS-B, Wesson et al. declared that the asymmetric-
[36], to check whether the data received by the key elliptic curve digital signature algorithm is vi-
UAV is sent from the authentic ground station or able. The use of asymmetric cryptography in ADS-
the eavesdropper, the authors have proposed a data B can be more costly and time-inefficient because in
authentication protocol using an asymmetric key asymmetric cryptography different keys are used to
algorithm technique. encrypt and decrypt the original message which can
Asymmetric security protocols are used for the take some time to process.
secure transmission of messages between the UAV The information authentication protocols are also
and the GCS. However, because of the communi- used to ensure the integrity of the transmitted data.
cation overhead, asymmetric protocols are mostly To exchange the key between the nodes, Diffie and
used for the symmetric key exchange between the Hellman proposed a key exchange protocol that
UAV and the GCS. Asymmetric security protocols has been frequently used in the past few decades
are also used to ensure the integrity of the transmit- [38]. When exchanging the keys between two par-
ted data between different sensors or devices. ties, they have no prior knowledge whether the
In [36], X.509 certificate with the elliptic curve keys which are sent over an insecure channel (e.g.,
cryptography (ECC) is used. By using X.509, the Internet) by an authentic person or not. The au-
generated signature is shorter in length which makes thors in [39] have developed a public-key exchange
the authentication process significantly faster. The protocol in which the sensor nodes exchange the
scheme proposed in [37] performs signature ver- keys and communicate with each other after authen-
ification after receiving the data. Once the UAV ticating their neighboring nodes. In the proposed
receives the data from the ground station or eaves- framework, two sensor nodes are considered as two
dropper, the UAV executes the verification process communicating parties. One sensor node sends the
to check the authenticity of the data before perform- public key encrypted message; while the other sen-
ing the final action. On the sender side, the 164 bits sor node decrypts the encrypted message with the
hash is generated using the SHA-1 algorithm. The private key and its own generated random number.
hash is then encrypted using the public key before Similarly, the second party sends the encrypted mes-
sending it to the other node. At the receiver end, the sage with the public key. The first party decrypts the
hash will be decrypted using the private key and then cipher message with the private key and with its own
the hash of the original message is calculated by the generated random number. If the decrypted message
receiver. In the last step, the receiver will compare is exactly the replica of the encrypted message, the
both hashes, and if there is no difference between sensors will be declared as an authentic entities and
them, the received message is deemed verified and will continue communicating.
was not modified by an unauthorized person. In [40], Valentin et al. have proposed a trust-based
The asymmetric security protocols can enhance protocol for the security of the UAVs. To check the
the security of the Automatic Dependent Surveil- correctness and accuracy of the data, trust values
lance Broadcast (ADS-B) [36]. The ADS-B is an air are assigned to the sensors which are determined
traffic surveillance protocol which is unacceptably by the UAV. The proposed methodology consists of
insecure. The ADS-B is used to detect the other three modules: i) a direct trust value determination
UAVs flying in the surrounding area. There are some phase, ii) an indirect trust value determination phase
major problems with the ADS-B which include the and iii) the final trust value determination phase,
lack of built-in security mechanisms such as authen- calculated by the UAV. In the whole environment,
tication codes and encryption modules that protect each sensor will determine its own trust values.
against the tampering of data and eavesdroppers, The UAV will use its own trust values as well as
respectively. It is critical that the technology used indirect trust values generated from the sensors.
in the ADS-B must meet the security requirements. The sensors in the environment can be placed by
In [36], Wesson et al. have raised a question; “Can the attackers and those sensors will also generate
an asymmetric cryptographic security protocol en- their own trust values. The final trust value will
hance the security of ADS-B?”. To address this be determined by the UAV. The trust values that
question, the existing cryptographic security proto- will be received by the UAV are compared with the
cols that are designed for the information exchange values placed in the log file of the UAV. If the final
between the GCS and the UAV are evaluated on determined trust value is negative, the sensors are
VOLUME 4, 2016 7

not trusted and the UAV will avoid taking further is a match, it implies that the message was not
data from those sensors, and hence, the attacker will changed by an external entity (attacker). This pro-
not be able to interfere with the UAV. In contrast, posed methodology implements the authentication.
if the final determined trust value is positive, the However, it is not feasible to keep the keys secret.
communication between the UAV and the sensors
will be enabled. Whereas, zero trust value means 3) Lightweight Authentication Protocols for UAV
that the UAV requires more information to decide Another way to conceal confidential information
whether the sensors are trusted or not. from the attackers is by using lightweight en-
In [41], Yoon et al. have proposed an authenti- cryption and authentication protocols. The use of
cation protocol to detect whether the information these lightweight schemes might help in encod-
received by the UAV is sent from the ground station ing the information in less time. It also does
or the attacker. In the proposed protocol, the UAV not consume heavy program memory which al-
sends encrypted random stream to the GCS. After lows the UAV to perform actions faster. In [43],
receiving the data, the GCS decrypts the random a lightweight encryption protocol was proposed
stream using the private key, then it encrypts the that works appropriately with frequent context
data a second time using its public key, and finally switching in a heavily multi-tasked environment.
sends it to the UAV. The UAV will compare the A lightweight blockchain-based stable routing al-
received data with the data maintained in its in- gorithm for swarm unarmed aerial systems (UAS)
dexes. If the authentication is successful, the UAV networking was proposed in [44]. Wang et al. have
is ready to take off. On the other hand, if the used the lightweight blockchain as a bargaining
authentication is unsuccessful during the exchange chip to strengthen the routing of swarm UAS net-
of information between the UAV and the ground working that uses 5G cellular network technology.
station, it indicates that the attacker is trying to The lightweight blockchain algorithm is different
take control of the UAV. Therefore, it disconnects than traditional routing algorithms as it can easily
the communication channel. No further information avoid the vindictive connections from the attackers,
is exchanged between the sender and the receiver identify malicious UAS, and reduce the intensity
after that. In this protocol, only the encryption of of the attacks from spiteful UASs. The suggested
a random message and the comparison method are algorithms were swarm UAS pitched that strives to
used to check whether the UAV has been hijacked expand the swarm UAS deployment networking on
or not. The proposed scheme provides information a wide range.
authentication. However, when large sized data is The low-cost devices can be consolidated into
sent to the UAV for authentication purposes, it re- UAVs to secure the data from the attackers by us-
quires a high bandwidth, cost, and processing time. ing the Internet of Things (IoT). To minimize the
To overcome these issues, we can use a small hash effects of cyber-attacks, the data should be encoded
for authentication instead of encrypting the whole with the use of session keys familiar to the specific
message. participating nodes. On the other hand, the embod-
In [42], to provide the authenticity and security to iment of the required abilities for both generations
the data which is stored in the UAV memory chips, of secure session keys and encoding/decoding of
Steinmann et al. have proposed a key negotiation the secret information is very tough in low-cost IoT
mechanism. In this method, the basic theme is to installations because of the performance limitations.
make an algorithm that continuously changes the In [45], Demeri et al. have applied a combined
random keys. For instance, the data encryption can secure and public key data transfer system with
be done by the one-time pad technique in which a low-cost aerial platform that combines different
the key-size is equal to the original message length. cryptographic accelerators. The components are in-
Now, if the attacker explores one key, the original corporated with the use of moldable and extensible
message can easily be revealed. So, the generation application programming interference (API) in a
of random keys may enhance the security of the keys software-hardware approached design that resulted
and the original message. In the proposed method- in costless drones. With the latest enhancement in
ology, the first node sends the public key encrypted the wireless communication system and miniatur-
data along with the hash code to the second node. ization of all the electronic devices, UAVs are of-
The second node decrypts the data with its private fering a great relaxation to the public. Moreover,
key, calculates the hash value, and then compares the UAV cybersecurity is getting more attention due
the calculated and received hash codes. If there to upcoming security issues, strategic and financial
8 VOLUME 4, 2016

information, and the importance associated with confidentiality attack, and the man in the middle
Aerial applications. attack can be performed by the attackers. To prevent
In order to provide security and authentication the UAVs from such attacks, authentication is quite
to the communication parties and to ensure the urgent to be established before the UAVs start to
privacy of the data, a lightweight authentication communicate with each other and guaranteeing that
protocol was suggested in [46] to offer secure com- an authentic drone in the network is the priority
munication between UAVs and ground stations. The of UAV network security. Whereas, the standard
proposed scheme also mentioned a packet capture authentication system that contains a username/-
(PCAP) to ensure the secure communications be- password or dynamic key is not significantly secure.
tween two parties. The basic idea of the PCAP is RSA certification requires a long-lasting session key
that the UAV and the ground station use the seed that is not able to fulfill the lightweight requirement
values of the chaotic maps that randomly shuffle the in the UAV infrastructure.
original message according to the generated chaotic In [50], a lightweight recognition authentication
sequence [46]. However, with the advancement in mode backed by ECC (Elliptic Curve Cryptogra-
the remote environments and the availability of low phy) is suggested that has three steps: i) ECC cer-
resources, the UAVs are doubted for device captur- tification initiation, ii) identity authentication, and
ing and dabble attacks. This increases the risk of iii) key compatibility verification. Teng et al. have
the data stored in UAVs to be stolen by adversaries. mentioned that the first two steps are fully com-
In [47], Haque et al. have focused only on the patible with the two-way authentication. Whereas,
secure transmission of information that UAVs send the last step verifies the consistency of the verifi-
to the base station. In [47], the data security and cation key. In contrast with the traditional authenti-
lightweightness was discussed and a new framework cation modes in the UAV network, the perspective
was proposed to achieve the desired tasks. For the suggested in [50] is based on short keys and less
lightweightness of the system, specific encryption computing workload. Barka et al. suggested a ca-
is performed. In the proposed scheme, apart from pable lightweight communication plan for the aerial
the cryptography, watermarking is also incorporated Named Data Networking (NDN) [51]. The proposed
to increase the integrity and confidentiality of the technique can sustain the NDN security and it pre-
data. The purpose of doing selective encryption is to dicts with 80% accuracy while reducing the end-
provide the stabilization between the UAVs under to-end delay to less than 1 second in the worst-
limited resources. Selective encryption may also case scenario. The proposed scheme also reduces
have advantages especially in real-time applications the average use of energy.
where fast processing is required. In [52], a novel authentication scheme for UAV
To highlight information insecurity and authen- is suggested. Since the UAV is supported by small-
tication issues, a two-phase lightweight mutual au- sized batteries and contains limited memory, the
thentication protocol was introduced in [48]. In the lightweight security methods are perfectly suited for
proposed system, a well suited software-defined them. In [52], Srinivas et al. suggested a temporal
networking (SDN) is supported with multi UAV credential-based anonymous lightweight authenti-
network installed in the required spying areas. In cation scheme (TCALAS) for the Internet of Drone
addition, the security evidence of protocol was also (IoD) networks. Contrary to the IoD surveillance
introduced to emphasize its security features. To framework suggested by Srinivas et al., their scheme
apply the authentication protocol more smartly in can perform only in the situation when there is only
UAVs, a Smart Internet of Drone (S-IoD) supported one flying zone that is not extensible. Moreover,
framework for a UAV environment was proposed in despite their declaration of robustness, the investiga-
[49]. The proposed scheme collects all the required tion done in [53] proves that Srinivas et al.’s scheme
information independently. For the sake of reduc- does not stand with traceability and availability at-
ing the computational cost of the authentication tacks. With the use of lightweight symmetric key
protocol, a lightweight privacy-preserving scheme primitives and temporal credentials, an upgraded
(L-PPS) was presented in [49]. The L-PPS offers scheme (uTCALAS) was suggested by Ali et al. in
robustness between the IoT devices with an appro- [53]. The suggested scheme offers security against
priate authentication time. several attacks that include traceability and avail-
Furthermore, due to the limited availability of ability attacks while keeping the lightweightness.
the resources and risky environment around the It enhances extendibility and can perform in the
UAV, different attacks that include wireless attack, area where several flying zones are available in the
VOLUME 4, 2016 9

IoD network. Furthermore, Ali et al. have success- on the training. The accuracy of the training model
fully achieved computationally fast authentication is evaluated in the testing phase and can be improved
that takes 2.29ms to accomplish the authentication by using different strategies. The learning based
process. techniques can be implemented in UAVs for the
intrusion detection by pattern recognition. Once the
B. PHYSICAL LAYER SECURITY IN UAVS UAV is trained, it is able to recognize the pattern of
One extensively adopted performance metric in the the intrusion.
physical layer security design is the so-called se- In [61], a deep reinforcement learning and a
crecy rate [54], at which the information can trans- weighted least squares algorithm [62] is incorpo-
mit securely. Traditional encryption protocols have rated to estimate the power of the jamming signal
vulnerabilities in the key distribution and high pro- with a convolution neural network (CNN) [63]. In
cessing time. The analysis of the physical properties the first step of the proposed approach, a relay power
of cellular channels can support secure transmis- factor is selected based on the bit error rate (BER)
sion. Physical layer security (PLS) is frequently and the channel gain. To initialize the weights which
used to achieve the maximum secrecy rate of trans- will be equal to the anti-jamming relays, a con-
mitted data between the two different nodes. In volutional neural network is used. These weights
fact, it is essential for all security controls and are updated by using a stochastic gradient descent
communication devices mounted in the UAV. Unlike algorithm [64]. After that, the UAV receives the
the conventional cryptographic security approaches, BER value from the ground station. If the learning
PLS takes advantage of the characteristics of cellu- parameter is greater than the power factor of the re-
lar channels such as fading, interference, and noise, lay power, the device chooses a random relay power.
to boost the signal reception at the legitimate re- If it becomes greater than zero, the UAV sends the
ceiver and reduce the received signal quality at the message with the randomly selected value of the
eavesdropper [55], [56]. PLS can be achieved by power by using a reinforcement learning. Note that
incorporating the cryptographic protocols. Several the randomly chosen relay power can increase the
cryptographic security protocols are presented in the error rate. Although the algorithm can prevent the
literature that provides a significant level of security UAVs and communications from jamming to some
but there is a no framework that offers an ideal extent, it can be very costly in case of a high error
security. Therefore PLS is gaining serious attention. rate.
To enhance and maximize the secrecy rate of In [65], an attack detection technique was pro-
wireless communication in the UAVs, a variety of posed in which two different machine learning al-
work have been proposed on PLS [57]–[59]. In the gorithms such as Support Vector Machine (SVM)
past few decades, static relay based communication and K-mean clustering are used. These algorithms
systems were deployed to improve the existing PLS learn from the data and make decisions for the
schemes. With the fascinating development in au- upcoming samples. There are two phases in the
tonomous vehicles such as UAVs, a new model of proposed technique. First, two parties send a signal
relying technique known as UAV-enabled mobile to the UAV, and second, the UAV transmits the
relaying has become a valuable technology. In [60], received signal to the third party for the detection
the authors have proposed an improved version of of deviation. For this purpose, machine Learning
a PLS scheme using UAV enabled mobile relying. (ML) is incorporated. After receiving the signal by
To improve the security of communication system, the third party, it is required to build a data set by
buffer-aided mobile relay is deployed which allows which the ML algorithms classify the output labels.
data to arrive independently and more quickly which The third-party will continuously receive the signal
is useful for real-time applications. and it will find the mean and standard deviation of
each received signal. As a result, the mean and the
C. LEARNING-BASED INTRUSION DETECTION standard deviation are the two feature points of the
A digital machine can perform different tasks based data set. After completing the training of the model,
on some instructions given by the user. To accom- when a third party received a new sample, it is fed
plish the automation of the tasks, machine learn- into the machine learning algorithm and according
ing (ML), deep learning, and neural networks are to the calculated values of both the features, it gets
frequently used. ML algorithms have two phases, assigned the label 1 (Attacked) or label 2 (Not
training and testing. In the training phase, the model attacked).
learns from the data and predicts future events based In the machine learning algorithms, there must be
10 VOLUME 4, 2016

more features in the data set to increase the accuracy which is used to detect the malicious nodes and
of the model in order to avoid any declassification protect the authentic UAVs from the attacks.
of any event. In this mode, only two features are In [67], intrusion detection and malicious node
used, as a result, the accuracy of the model gets ejection issues are addressed. A new intrusion detec-
compromised. tion technique was proposed by using the Bayesian
game model [68] to detect the intruders more ac-
D. RULES-BASED INTRUSION DETECTION curately. The main focus was to detect the internal
To make a device intelligent, some instructions intruders and to eject the node which can be harmful
should be given to that device. In rule-based tasks, for the UAV network. In the first step, the intrusion
the user must define some rules. Based on those detection mode is activated by the different nodes.
rules, the device takes the decision and sends the To perform this task, the IDS first computes the
command to the base station. In the case of UAVs, misbehavior rate (MR) of any other UAV which is in
for each task, different rules are fed into the chip the UAV network. If the MR exceeds the threshold
of the UAV and threshold levels of the acceptance value, then the IDS starts monitoring the neighbor-
of each rule are set. For example, if the threshold ing nodes and activates the detection system. Other-
is 80%, it means that if the UAV finds the true wise, the IDS does not perform any task. Similarly,
condition of the rules equal or greater than 80 per- the intrusion ejection system computes the MR of
cent, the UAV performs the specific function and the node. If the MR exceeds the threshold value
vice versa. In [66], a new intrusion detection system of the intrusion ejection mechanism, it declares the
was proposed based on the specific behavior rules node as malicious and rejects it from the network.
to minimize the false negative predictions. In the As far as the intrusion ejection is concerned, the
proposed detection method, seven different attacks ejection of a node from the network immediately is
were discussed which are related to the availability, not a suitable approach. The node may misbehave
confidentiality, and integrity attacks. When the UAV for a certain period due to the environmental condi-
experiences anything from these seven attacks, the tion. If we eject the node immediately on the bases
UAV takes measures to defend itself. First, when the of the misbehavior rate, the false positive rate may
UAV reaches outside to the safe space, it activates increase.
the weapons to defend itself against the attack. Sec- Table 2 and 3 shows the comparison of security
ond, when the sensors readings are different from protocols used for secure communication and intru-
the trusted node, actions are taken. Third, when bad sion detection in UAVs, respectively.
recommendations are received regarding the trusted We have analyzed some statistical results of ma-
node and good recommendation are received regard- chine learning based techniques for UAVs which
ing misbehaving UAV, appropriate actions are taken. either incorporate rule-based or learning-based tech-
The fourth indicator handles the situation when niques [65], [69]–[74]. The results analyzed are
UAV deploys landing gear in an inappropriate area. displayed in Table 4. Four important factors are used
These four attacks correspond to integrity attacks. to analyze these techniques which are True Positives
The fifth indicator is activated when the UAV starts (TP), True Negatives (TN), False Positives (FP) and
sending data to unauthorized parties. This attack False Negatives (FP). The aforementioned factors
corresponds to the confidentiality attack. The sixth are used to find the statistical parameters such as
attack indicator occurs when without analyzing any accuracy, precision, recall and F1-score. Accuracy
attack the UAV uses its countermeasures. Seventh tells us how many correct predictions are made by
and last attack is when the UAV uses more thrust any model. The higher number of correct predic-
to cross the limited altitude which is defined by the tions will result in higher accuracy. The accuracy
authorized person. Sixth and seventh attack corre- can be calculated as follows:
spond to the availability attack. These seven attacks No. of correct predictions
are taken into account and after detecting the attack, Accuracy = (1)
Total number of predictions
the UAV defends initiates a defense phase to protect
itself against the above-mentioned attacks. OR
Intrusion detection systems (IDS) are also used TP + TN
to detect the deviations that occur in the network. Accuracy = (2)
TP + TN + FP + FN
Henceforth, the IDSs remove the effect of the attack
in order to prevent the systems from hazards. An The second parameter Precision is calculated by
IDS is a major mechanism in the UAVs network taking the ratio between the true positives and the
VOLUME 4, 2016 11

sum of true positives and true negatives. High pre- In this section, we will discuss how an attacker can
cision is required for a better model. The precision breach the security of the protocols by using dif-
for the model can be calculated as follows: ferent attacking strategies. Vulnerabilities and their
TP countermeasures for the security protocols are high-
Precision = (3) lighted in Table 7 and the schematic diagram for the
TP + FP
The third statistical parameter Recall refers to possible vulnerabilities is shown in Figure 4.
the sensitivity of the system, and is calculated as
follows: A. GLOBAL POSITIONING SYSTEM (GPS)
TP SPOOFING
Recall = (4)
TP + FN GPS spoofing is categorized as a cyber-attack by
The last parameter is called F-Score and it can which an attacker transmits a fake GPS signal with
be calculated from the recall and precision. In other slightly higher power to mislead the reception of
words, F1-score is the weighted average of recall the UAVs. Due to the wireless connection between
and precision.The range of F1-score lies between the GCS and the UAVs, the vulnerability factor
0 and 1, whereas 1 indicates perfect precision and inclines. Without the integration of complex checks,
recall and 0 indicates either the precision or recall is i.e., whether the signal is received from the GCS or
0. F1-score can be calculated as follows: the attacker, the UAVs may tend to perform actions
−1 from an unauthorized signal as well. To recognize
(Recall)-1 + (Precision)-1

F1-Score = the authority of a legitimate signal, a log-likelihood
2 (5)
radio test method is adopted in [87] to fight against
Precision × Recall the signal spoofing attack. Prior information of the
=2
Precision + Recall received signal frequency from the GCS assists the
It can be seen from Table 4, the accuracy of UAVs to identify the signal transmitter’s informa-
the work proposed in [72] is higher in comparison tion and breakdown the legitimacy of the received
to other schemes. This means that the percentage signal. In [87], a decision threshold methodology
of TP predictions for the scheme proposed in [72] was designed using the Neyman-Pearson criterion
is comparatively higher. However, the accuracy for [88]. To select the appropriate threshold value, the
[71] is slightly less than the scheme proposed in false alarm rate (FAR) value is set to be fixed.
[72]. For the selection of an appropriate FAR value, a
Moreover, we have also analyzed some schemes cumulative distribution function is estimated, which
discussed in section II in terms of applications and helps in the detection of spoofed signals.
used technologies in Table 5 and 6. As different In [89], Sedjelmaci et al. proposed a new method-
types of UAVs are used for different applications ology for the detection of spoofed signals. A rule-
such as wireless coverage, remote sensing, real-time based detection technology, when incorporated with
monitoring, search and rescue operations, surveil- the protocols, helps in achieving better accuracy.
lance, and delivery of goods, it is important to In this methodology, a comparison between a spe-
choose the right UAV and the suitable scheme for cific threshold value and the transmitted signals is
each specific application. For instance, in Table 5, it performed. If the value of the transmitted signal
can be seen that the scheme which was developed in (transmitted by the attacker) becomes greater than
[75] is suitable for wireless coverage and surveil- the threshold value, the UAV detects that the re-
lance. Similarly, Table 6 provides an analysis of ceived signal is spoofed. In [90], Qiao et al. pro-
the existing schemes in terms of used technologies. posed the use of a vision system based technique to
For example, the scheme presented in [75] supports detect the GPS spoofing for UAVs in which inertial
secure communication. By incorporating the image measurement unit (IMU) sensors are used to find
processing technology, one can securely communi- the instantaneous acceleration and velocity of the
cate by sending the encrypted digital data such as UAV. To make the model of the proposed system,
images. Qiao et al. considered three different coordinate sys-
tems which include the body frame coordinates,the
III. VULNERABILITIES IN UAVS ground coordinate system, and the image coordi-
Several methodologies have been proposed to en- nates system. To find the GPS spoofing, it compares
hance the security of UAVs. However, there are the two kinds of velocities which are measured by
shortcomings in the proposed protocols which have the Lucas-Kanade (LK) method [91] with root mean
made UAVs vulnerable to certain security threats. square error (RMSE) values. If the value of RMSE
12 VOLUME 4, 2016

TABLE 2: Comparison of different protocols used for secure communication in UAVs

Categories Techniques Features Vulnerabilities/

Weaknesses

Cryptographic OTP for securing Secure technique due Large key More bandwidth is required for
symmetric communication links [31] size sharing the secrete key
security protocols

Data security protocols More randomness due to Lorenze Absence of confusion part which makes
(Symmetric) [34] machine the system insecure

Data authentication Shorter hash length Hash is in encrypted form, so, the
using asymmetric Computationally fast eavesdropper may try to perform attacks
security protocol [36] to retrieve the original hash.

Analysis of Declares that the elliptic curve More costly and time inefficient
cryptography in ADS-B [37] digital signature algorithm is eavesdropper may try to perform attack
viable

Public-key exchange Nodes communicate with each Choose only one random number as a
security protocols in others after the authentication of public key to encrypt the message that can
UAVs [39] the received commands from the easily be judge by the attacker.
ground station (GS)

Trust-based security Authenticate whether the sensors Packets which are sent by the UAV
protocols for the UAVs placed in the UAV network are to the GS are in the original form which
and the sensors [40] trusted or not. can be fabricated or stolen by the
attacker.
Cryptographic
asymmetric
security protocols Authentication scheme, When someone sends a large data to the
whether the data Protects the information even after UAV for authentication purposes,
received by the UAV is hijacking the UAV by the it requires a high bandwidth, high cost
from the authentic or attacker and the sending or processing time
unauthentic user [41] will also be high.

Key negotiation
mechanism for Continuously changes the
providing authenticity random keys Not feasible to keep the keys secret
and security to the data
stored in UAS chips [42]

Reduces overhearing effects, Takes more time to execute all the

Convex optimization [76] Enhances the trajectory and complex computations to find whether
technique transmit power of the UAVs the system converges or diverges
simultaneously

Lightweight blockchain
algorithm is way far
lightweight Supports 5G cellular different than traditional
blockchain-based [44] network algorithms

Authentication using Components are Slower than traditional

Lightweight public key data incorporated with the authentication schemes
Authentication transfer system [45] use of moldable and because of asymmetric
protocols extensible API keys
for UAV

Authentication using use the challenge Less dimensional chaotic

packet capture (PCAP) [46] response pair of map is used
physical unclonable
function

SDN based Combination of Computationally

authentication [48] security and SDN inefficient

Authentication
S-IoD supported collects all the Low computational
framework [49] required information cost
independently

VOLUME 4, 2016 13

TABLE 3: Comparison of different protocols used for intrusion detection in UAVs

IDS type Used techniques Advantages Disadvantages

Deep reinforcement Estimate the power of the jamming High time complexity and error rate
learning technique [61] signal
Learning-based
Intrusion
Detection Attack detection Fast process due to the Less features selected due to which
technique [65] incorporation of ML algorithm accuracy is compromised

Intrusion detection Less false-negative prediction High number of rules, due to which
technique [66] detection, defends the UAV by the processing time can be high delaying
false information injected the decision taken by the UAV
Rule-based
Intrusion
Detection Use of a Bayesian game model. If Instead of immediately ejecting nodes,
Intrusion detection and we eject the node immediately on more rounds are needed. Then take
malicious node ejection [67] the bases of misbehavior, the false the average after taking the final decision
positive rate may further increase of the ejection of the malicious node.

TABLE 4: Statistical results for learning and rules-based intrusion detection

References Accuracy (%) Recall Precision F1 score
Sun et al. [69] 93.5 95.8 0.94 0.94
Hoang et al. [65] 88.8 0.98 0.91 0.94
Anwar et al. [70] 89.1 0.98 0.90 0.93
Shoufan et al. [71] 96.47 0.91 0.95 0.92
Ezuma et al. [72] 96.83 0.93 0.97 0.94
Li et al. [73] 66.87 0.81 0.76 0.78
Alipour et al. [74] 90 0.94 0.95 0.94
Han et al. [77] 74.7 0.045 0.81 0.085

FIGURE 4: Vulnerabilities in UAVs.

becomes greater than the threshold value then it will threshold value, the UAV will remain in the normal
be declared that the UAV is spoofed else it is in a state and if the error is greater than the threshold
normal state. In [82], another scheme is proposed to value, it means that the UAV is hijacked. As this
detect GPS spoofing. It is supposed that the UAV has technique is so simple, it can be significantly im-
an inertial measurement unit (IMU) sensor which proved by placing the jammers in the UAV model,
is helpful in monitoring the behavior of the UAV. i.e., those jammers will only block the signals which
The proposed GPS spoofing detection is very simple are greater than the threshold value. By two this, two
to implement because it just compares the strength goals can be achieved: the first is the detection of
of the signals and finds the error rate. Based on spoofing attacks which is already done in [82], and
the error, it declares whether the UAV is spoofed the second is the protection of the UAV from the
or not. The concept is very much similar to the spoofing attacks.
idea proposed in [90]. There is a fixed threshold
value and it compares with the GPS signal strength
and finds the error. If the error is smaller than the
14 VOLUME 4, 2016

TABLE 5: Analysis of existing schemes in terms of applications

Schemes Providing Wireless Remote Real-Time Search and Delivery Surveillance

Coverage Sensing Monitoring Rescue of Goods

Faraji et al. [75] X X

Houng et al. [78] X X X
Challita et al. [79] X X
Li et al. [80] X X X
Hudson et al. [81] X X X X
Ma et al. [33] X X X
Kirichenko et al. [34] X X
Pan et al. [37] X X X X
Valentin et al. [40] X X X X
Yoon et al. [41] X X X X
Li et al. [54] X X
Zeng et al. [57] X X
Choi et al. [58] X X
Li et al. [59] X X X
Wang et al. [60] X X X X
Hong et al. [65] X X X X
Sedjelmaci et al. [67] X X X X
Zou et al. [82] X X X X X
Fotohi et al. [83] X X X
Li et al. [84] X X X X X
Amelin et al. [85] X X X X
Hooper et al. [86] X X X X

TABLE 6: Analysis of existing schemes in terms of used technologies

Schemes Collision Avoidance Free Space Cloud Machine Image

Optical Computing Learning Processing

Faraji et al. [75] X

Houng et al. [78] X X
Challita et al. [79] X X
Li et al. [80] X X
Hudson et al. [81] X X
Ma et al. [33] X X
Kirichenko et al. [34] X X X
Pan et al. [37] X X X X
Valentin et al. [40] X X X
Yoon et al. [41] X X X
Li et al. [54] X X
Zeng et al. [57] X X
Choi et al. [58] X X X
Li et al. [59] X X X
Wang et al. [60] X X X
Hong et al. [65] X X X X X
Sedjelmaci et al. [67] X X X X
Zou et al. [82] X X X X
Fotohi et al. [83] X X X
Li et al. [84] X X X X X
Amelin et al. [85] X X X
Hooper et al. [86] X X X X

B. FALSE DATA INJECTION control of the UAVs. In Abbaspour et al.’s proposed

system, neural networks are incorporated to increase
False data injection is a technique by which an the learning ability that will detect the FDI. Specifi-
unauthorized person sends the clone data to the cally, Abbaspour et al. have focused on IMU sensors
UAV in order to take control. UAVs are prone to which provide the information of angular and linear
vulnerable data based on the technique implemented movement of the UAVs. Instead of direct applica-
on their security and cannot differentiate between tion of neural network, a neural network adaptive
the authentic and unauthentic data. To ensure se- structure (NNAS) was implemented. To improve the
cure operations, UAVs must be able to detect the accuracy and time efficiency, an Embedded Kalman
false data injections. To address this issue, in [10], Filter (EKF) was used to update the neural network
Abbaspour et al. proposed a protocol for the detec- parameters. In [92], the ability to transmit fake in-
tion of cyber-attacks, known as false data injections formation to take control on the UAV was explored.
(FDI). To perform a false data injection, intruders To detect the fake inserted information, a threshold
normally use their own sensors to inject the fake based technique was adopted. The threshold values
information in the UAVs sensors in order to take
VOLUME 4, 2016 15

of the position and speed of the UAV are kept spe- the receiver. In wireless communication, jamming is
cific to determine whether the signal is fake or orig- a well-known research area. Many defense mecha-
inal. Information is flagged as fake when the values nisms are proposed by the researchers. In [94], full-
of the velocity and position become larger than the duplex eavesdropping [95] is considered in which
threshold values. Detection of fake information is jamming and eavesdropping can occur simulta-
not enough but it is important to equip the UAVs to neously. To avoid the malicious jamming signals
take the right action against such information. Major which are sent by the attackers, a new mathematical
issue with this type of techniques is that the UAV model was proposed in which the receiver values
only detects the fake information when the values of fall below the threshold. In the proposed model, it
the received signal raises above or decreases below was assumed that the source and the eavesdropper
the specified threshold value. have a line of sight (LoS) path [96] towards UAVs.
Researchers are trying to resolve the issue of The probability of the LoS increases with the height
insecure transmission of the data over an insecure of the UAVs. To manage the LoS probability, path
channel using strong wireless communication secu- loss exponents are formulated for the UAV. In addi-
rity protocols [83], [84]. Although, the wireless con- tion, for the secure transmission of the information,
nection (WiFi or Radio) security has now improved an artificial noise (AN) is additionally sent with
to some extent but more advancements are required the original information for effective confusion. The
in this area of research. eavesdropper will receive the jamming signals with
the original from the UAV simultaneously. Hence,
C. WIFI INSECURITY AND JAMMING ATTACK a simultaneous transmission of the original signal
UAVs can be connected via wireless channels such along with the noise can make it harder for the
as WiFi. WiFi based UAVs are vulnerable to the hacker to extract important information.
wireless attacks and can be hacked by interrupting Most of the researchers are using AN frequently
the communication between the UAV and the GCS. over the last few decades to protect the sensitive in-
For the secure communication using WiFi, it can formation. Protection of sensitive information from
be protected via unique passwords. The analysis attackers has become a great challenge for the re-
of the vulnerabilities of micro-air-vehicle commu- searchers. In order to sort out these issues, a new
nication (MAVLink) protocol was included in [93]. methodology was proposed in [75] for the secure
In the MAVLink protocol, two specific issues can transmission of sensitive information from the UAV
arise. First, the MAVLink protocol does not en- to the GS. A protocol was proposed by Faraji et al.
crypt the message which is to be sent. As a result, in [75] to protect the information from malicious
integrity attacks can surely happen. Secondly, if UAVs launched by the attacker in the communica-
the MAVLink encrypts the message, a delay can tion network, and the transmission of information
occur due to the encryption and decryption of the from the UAV to the GS. The proposed technique
messages, which can lead to availability attacks. In defines rules to spot malicious UAVs, so that the in-
the proposed methodology in [93], it was supposed formation can only be sent to the authorized nodes.
that the attacker has already hacked the network and The purpose to spot the malicious UAV was to get
gathered all the relevant information and can send rid of these UAVs in accordance to their behavior
this fabricated information to the host. Based on in the whole UAV network. Hence, the exchange of
the fabricated information, the attackers can easily fake information passed on to the ground stations
identify the position of the UAV. In this methodol- could be prevented. Though the presented work is
ogy, although a counter attack against hackers can fully capable of removing the malicious UAVs from
be carried out, but the supposition can affect the the network in order to keep the information secure,
overall efficiency of the technique. There should be it is not able to keep the information secure form the
a feasible protocol without pre-supposing any kind attackers due to the unencrypted nature of data. This
of attacks. Apart from the fabricated information, means the information could be hijacked. Therefore,
jamming attacks are also often performed by the the algorithm needs to be improved and it requires
eavesdropper to hack the information which is sent the incorporation of a powerful cryptographic pro-
by the UAV to the GS. tocol.
To distract the UAVs, jamming attacks are fre- In [78], a novel scheme was suggested to secure
quently incorporated by the attackers. A jamming the data and to get rid of both the security and
attack can also be planned to discontinue the com- efficiency issues. While, the purpose was to uncover
munication between the authentic transmitter and the security and wireless challenges that stood up
16 VOLUME 4, 2016

in the context of a UAV-based information transfer computer and, the navigation sensors. To estimate
system [79]. To expose these challenges, a solution the parameters such as the control parameters of the
based on ANN was suggested that enables UAVs UAVs, a recursive least square (RLS) method [85]
to adaptively exploit the wireless system resources was adopted that takes a certain input seed value
while safeguarding the operation in real-time. In and generates the corresponding output values. The
[80], to prevent confidential signals and information RLS method detects the divergence of the system
from the cyber-attacks and hijacking, an artificial control parameter values by continuously compar-
noise (AN) was incorporated. By using the AN with ing the predicted values from the previously known
the original signal, the resulting signal is transmitted values. If there is a significant divergence of the
in a disordered form which makes it difficult for control parameters, a fail-safe protocol is executed
the hacker to decrypt it and extract the original which allows the UAVs to return on the designated
information from it. To secure the information from spot safely. The controllers are responsible for the
the attackers by incorporating friendly jamming in security of the UAVs, but it can be hacked by simple
[66]. A friendly jamming signal is a noisy signal, fuzzing attacks. It is another common attack that can
which is transmitted with the original signal. Noisy be used to hack the controller of the UAVs.
signal is named friendly jamming signal because it In [86], the authors have proposed a methodology
does not disturb the original signal, and its purpose to secure the UAVs from the basic attacks such as
is to provide the security to the original signal. For Daniel of service (DoS) [99] and buffer overflow at-
the friendly jamming signal, another reference UAV tacks [100] by using a fuzzy technique. WiFi-based
UAVj is considered which transmits this noisy signal UAVs can become easy targets of such attacks due
with the original signal. However, to maximize the to the wireless communication links. Moreover, it is
secrecy rate, a block successive upper bound min- also demonstrated in [86] that the protocols which
imization technique (BSUBM) [97] was developed are used to create a link between the controller and
in [66]. BSUMB is used to identify the user schedul- the UAVs are insecure. To solve this problem, three
ing, in which when one user is scheduled, the additional mechanisms were used which are: watch-
others are unscheduled. The UAV sends messages dog timer, hardline input data filtering, and anti-
only to the scheduled users to protect the trans- spoofing mechanisms [101], [102]. The watchdog
mitted messages from other unknown users. Once timer provides security against the DoS attack and
the scheduling process is completed, the transmitted it works in the domain of operating system (OS).
power of UAVm (which transmits the message) and This ensures that the non-navigational processes are
UAVj (which transmit the friendly jamming signal) at low priority and allows access to the CPU for only
is optimized just to send the information in a more a definite period of time. The second mechanism,
sophisticated way. The secure communication with i.e., hardline input data filtering, is able to decline
the friendly jamming signals is guaranteed for the the non-authenticated process. It protects the UAVs
scheduled users. What if the jamming signal is from Buffer overflow attacks that limit the data
trapped by the eavesdroppers? There should also which is to be sent to the UAVs. Lastly, an anti-
be a protocol to secure the friendly jamming signal spoofing mechanism is incorporated to prevent the
as well. Moreover, in the proposed algorithm, the UAVs from the Address Resolution Protocol (ARP)
collision of UAVm and UAVm is also ignored. So, attack. This secures the network from unauthentic
another problem can be faced when UAVm and information.
UAVj collide with each other. In the Fuzzy logic based UAVs, landing issues
require special attention. In [103], an issue of safety
D. CONTROL SYSTEM VULNERABILITIES AND landing is discussed and an algorithm is proposed
FUZZING ATTACKS using Fuzzy logic. In this algorithm, a speed control
Besides the choice of a suitable UAV type, it is mechanism is used for a safe landing. Vertical speed
also important to analyze the control system which and altitude are the two inputs. Throttle positions are
is mounted in the UAVs. Most of the movements considered as an output parameter. The most impor-
of UAVs are dependent on the control system. In tant factor that is “in ground effect” is considered as
[98], Birnbaum et al. focused on two major aspects. a threshold value. Based on a threshold value, the
First, to develop such a system that controls the UAV can land safely. The major advantage of this
performance and detects the hardware failures of fuzzy logic-based methodology is that the process-
the UAVs. Secondly, to detect the different types of ing time is less and the system can operate quickly.
attacks such as attacks against the flight control, the
VOLUME 4, 2016 17

E. MALICIOUS UAV DETECTION F. VULNERABILITIES IN PACKET

FORWARDING AND ROUTING PROTOCOLS
As far as the cost is concerned for the detection of Flying Ad-hoc Networks (FANETs) is a type of
malicious UAVs, a vision-based detection system network in which different UAVS are connected in
for UAVs with radar sensors and cameras can be an ad-hoc manner. UAVs are organized into teams
fairly expensive. In [73], detection of malicious to achieve high level goals. To establish a reliable
UAVs within a given time interval to classify the communication between the UAVs specialized rout-
data, and a machine learning-based technique are ing and packet forwarding protocols are needed.
proposed. There are two major concerns while using
Routing protocols are the set of defined rules used
these algorithms: i) the selection of the right features
by the routers to distribute the information between
and ii) the right identification of the event. In the
different nodes. They are also used to update the
data set of the detection of UAVs, eight different
routing tables so that the routing decisions can be
types of UAVs are used. In order to identify the
made. Updating the routing table depends on the
UAV, the features such as packet size and arrival
type of the used routing protocol and the adopted
time of traffic are first extracted. After that, packet
forwarding technique. For instance, in static routing
sizes for different packets and arrival times, in the
protocols, an administrator manually assigns the
data set, are defined. The different sizes of the pack-
path from the source to the destination. Moreover,
ets will take different time to reach the destination.
other than the administrator, no one can add/update
Based on the defined data set, UAVs are classified
the routes. Whereas in dynamic routing protocols,
as secure or insecure. For the detection of malicious
a different route is chosen dynamically in case if
UAVs, the data set is used in this technique con-
a link goes down. As the routing protocols are
tains only two features. We can further enhance the
also responsible for updating the routing tables ef-
accuracy and minimize the declassification rate by
ficiently, it is necessary to ensure the right route is
selecting more features.
selected to forward the packets to the destination. In
In a UAV network, several UAVs exist to per- addition, before forwarding the packets through any
form a specific task. Within the UAVs network, the route it is important to ensure the integrity, confi-
possibility of an external attack is always there. In dentiality, repudiation, availability, and authenticity
[104], a WiFi-based fingerprint technique is pro- of the forwarded messages. In the absence of the
posed to detect the authentication of information. aforementioned security services when using a week
In the WiFi-based channel, there are several types routing protocol, the consequences are devastating
of traffics. In order to observe the classification of and may be in favor of the by allowing them to hack
traffics, data mining algorithms were incorporated into the forwarded packets.
[105] which help to detect the unauthentic UAVs. In Apart from the security protocols such as cryp-
the proposed technique, fingerprint is considered as tography and intrusion detection, routing protocols
a feature vector. The technique starts by capturing have their own significance in FANETs. When data
the traffic flow of the packets. From each flow, one is sent by the UAV to the destination, it follows a
feature vector is extracted with different features, specific route. Packet forwarding and routing pro-
i.e., if the total number of captured flows is ten, then tocols are a major building block of modern UAVs.
the total feature vectors will be ten, in which some However, the initial design of these protocols do not
of the feature vectors are used for training purposes consider security and vulnerability, hence, making
while others are used for testing purposes. Different it an attractive target for the attackers. In this sec-
features are extracted from feature vectors and are: tion, we will analyze the vulnerabilities in the rout-
the average packet length, the root mean square ing protocols and discuss their countermeasures.
value of the packet length, the total duration of each A comprehensive summary of routing protocols,
flow, the average inter-arrival time of the packet, the vulnerabilities and their countermeasures is given in
root mean square value of the packets inter-arrival Table 8.
times, and the transmitter and receiver addresses. Securing the routing protocols has become a chal-
After capturing the traffic flow, pre-processing is lenging task, due to the excessive usage of wireless
performed on the prepared data. Finally, based on connections. Typically, there are several motivations
the extracted features authenticity of the UAV is for the eavesdropper to attack the routing protocols
predicted. [112]. Routing protocols in UAVs are vulnerable
to different cyber-attacks because of various rea-
sons [112]–[114]. For instance, they rely heavily on
18 VOLUME 4, 2016

TABLE 7: Vulnerabilities and their countermeasures for the security protocols used in UAVs

Categories Techniques Details Countermeasures

False alarm rate (FAR) is set

A log-likelihood radio test to be fixed, FAR should be updated
scheme [87] Gives the approximate value with time
of FAR

GPS spoofing attack Comparison of different High threshold may increase the
detection [90] values accuracy of the attack detection

The phenomenon of
Global Positioning Rule-based detection comparing the threshold There should be flexibility while
System (GPS) techniques are incorporated for values with the statistical comparing the values
GPS spoofing attacks [89] values of the transmitter is for the GPS spoofing
detection used. Comparing the values attack detection
may need high accuracy purpose

Monitors the behavior of

GPS spoofing attack UAVs while detecting the It can be significantly improved by
detection [82] spoofing by comparing placing the jammer in
the strengths of the the UAV model
signals

Based on Neural Networks

with embedded Kalman Should update the Neural
False data Protection of UAVs from filter (EKF) and only able to Network parameters continuously to
injection the sensor-spoofing attacks [10] detect FDI, not accurate and make it time-efficient
time inefficient

Analysis of the vulnerability of MAVLink protocol does not

WiFi Insecurity micro-air-vehicle encrypt the message due to —
communication (MAVLink) the time inefficiency
protocol is performed [93]

Artificial noise is added

Methodology for preventing with the original A plaintext can be added by applying
the UAVs from malicious information which can be the watermarking technique in order to
jamming signals [94] detected and separated by make it a meaningful text
the attackers

Jamming attack
Improvement of By introducing the updating
Physical layer security Artificial noise addition, height parameters to work on UAVs having
transmitter-receiver [106] of UAVs is assumed to be fixed no fixed height or position

A recursive least square (RLS)

method is used to detect the
Control system Controls the divergence of the system control Friendly jamming can be placed to
security performance detection parameter, overcome the non-detection problem
vulnerability of hardware failure [98] Not feasible to detect Spoofing of spoofing attacks
attacks

Three different mechanisms are By optimizing three mechanism and

Fuzzing attacks Attack detection [86] used due to which the overall then used as a one mechanism
overhead time is increased

Malicious UAV Use more relevant features to

detection using Less accuracy enhance the malicious UAV
machine learning [73] detection accuracy
Malicious UAVs
detection
WiFi-based fingerprint technique is
Detect the authentic used, in which for each flow, a Instead of registering the fingerprint,
UAV [104] different fingerprint is registered makes other feature vectors different
which causes more processing time from fingerprint

VOLUME 4, 2016 19

TABLE 8: Vulnerabilities and their countermeasures for the routing protocols used in UAVs security

Categories Techniques Details Countermeasures

Static routing [107] Information Authentication precools

disclosure must be incorporated

Routing for Dynamic UAV must be intelligent to

data packets [108] topology identify the redundant
traffic

Packet forwarding Load-carry-and-deliver Implemented a relay

and routing protocols (LCAD) static routing [109] Large latency issues function in an ad-hoc
network [110]

Static routing [110] Does not allow to change the can replace with it
routing tables during the mission dynamic routing.

uses delay tolerant forwarding UAV must be enough

Hybrid Packet and end-to-end routing, intelligent, it can be
forwarding [111] which is not efficient possible by incorporating
ML techniques

wireless connections which are highly vulnerable cache poisoning, or modifying the control packets
to attacks such as data tampering, DoS attacks, [120], [121].
and eavesdropping [115]. Furthermore, because of The attacker can also aim to breach existing
the dynamic topology of FANETs, it is very hard routing protocols to degrade the performance of a
to distinguish between a legitimate and malicious network or to modify its topology [108]. The per-
node. A legitimate node may also misbehave for a formance can be degraded by perturbing the rout-
short period of time due to poor connection quality ing algorithm or by launching a DOS attack. The
or loss of route [116]. Moreover, the UAV launched attacker can add redundant traffics to the system to
by the attacker may behave properly for some time increase the load thus decreasing the performance.
to gain trust, but simultaneously, it may also create The attacker can also modify the topology of a
inconsistencies in the routing. For example, the at- network by introducing non-existent nodes into the
tacker may broadcast a non-existing link or produce routing tables, forging a route link or by performing
a new routing message to mislead the other nodes. a packet modification attack.
Such attacks are very hard to tackle as the malicious
node may be a legitimate entity. 2) Attacks on Routing Protocols
Due to the security issues discussed in the previous
1) Analysis of Security Issues in Routing Protocols section, routing protocols are exposed to different
Without proper security mechanisms, routing proto- types of threats. The goal of the attacker is to control
cols may be subject to information disclosure. The the network traffic, disrupt the routing functions, or
attacker may collect information related to the net- inject malicious nodes. The threats can be classi-
work topology, the position of UAVs, the commands fied based on the basic routing functionalities into
and controls, and the traffic payload. The confiden- various types of attacks such as route discovery
tiality of the system is not protected if an attacker attacks, route maintenance phase attacks, and data
can obtain all that information by eavesdropping. forwarding phase attacks [122], as described below.
Most of the routing protocols have this vulnerability • The goal of the attacker while performing a
[117]–[119]. route discovery attack is to modify the net-
Due to lack of authentication mechanisms, the work topology by adding malicious nodes or
attacker may also collect information related to the by invalidating the routes. During the route
commands and controls, and the data traffic [107]. discovery process, a sender node searches for a
If the attacker can successfully attract the control route to a destination node. The sender broad-
packets during the discovery of a route, it can per- casts a route request and waits for a route
form many attacks such as, disconnecting a spe- response. The discovery process of the route
cific link, rejecting the legitimate routing messages, is very crucial as it conditions other routing
20 VOLUME 4, 2016

processes. A reliable route will be found if the 1) With the exponential growth in UAVs tech-
packets are properly exchanged without any nology and their applications, the variety of
manipulation by the attacker. Otherwise, the algorithms associated with the security of
attacker can establish a false route containing UAVs have been proposed. The applications
malicious node [123]. of UAVs include the information transmission
• The attacks during the route maintenance from UAVs to the ground station, the data
phase are conducted after route loss or when capturing, and the transfer of goods. The secure
a link breaks due the node movement [124]. transmission of information between different
To reduce the excessive overhead required in components of UAVs is very important. Sev-
discovering a new route and to achieve sta- eral protocols are presented in the literature
bility, routing maintenance is necessary. It is to secure the transmitted data so that attackers
done by exchanging the beacon messages pe- cannot access the sensitive information. Apart
riodically. The objective of the attacker is to from secure transmission, it is also very impor-
degrade the performance by adding redundant tant for the security protocols used in UAVs to
nodes, irrelevant traffic to increase the routing verify the authenticity of the data. |In fact, it is
load and add processing delay. Normally, in a common for attackers to send spoofed signal-
routing protocol, an error message is generated s/information to trap the UAVs. The suitability
to publish the broken routes. A malicious node of different authentication protocols was inves-
may exploit that functionality by broadcasting tigated and applied over the years. As UAVs
false route error messages, and hence, prevents have limited resources in terms of memory
the source node from communicating with the and processing power, it is often recommended
destination. to use lightweight authentication to make the
• During the message forwarding phase attacks, protocols efficient.
the goal of the attacker is to fail the mission by 2) Although there are several security protocols in
disrupting the forwarding of the payload traffic the literature to secure the communication be-
[125]. A malicious node may drop, replay, or tween the different components of UAVs, most
modify the packets. In addition, the attacker of them can be attacked. Several weaknesses
can also delay/disrupt the time sensitive com- in the system can be exploited to compromise
munication by delaying the packets to their the UAV system using attacks such as GPS
respective next-hop destinations. signal spoofing, false data injection attacks,
WiFi insecurities, jamming attacks, etc. The
3) Security Solutions for Routing Protocols spoofing attacks can be performed using dif-
To ensure the security of the routing protocols, it ferent techniques. Commonly, the attackers use
is necessary to preserve the reliability, efficiency, the spoofed signals with greater frequency than
and accuracy within the malicious attacker’s envi- the actual signals sent by ground station to the
ronment. To preserve the integrity and confiden- UAV.
tiality of the routing packets, traditional methods 3) Routing and packet forwarding protocols are a
such as symmetric and asymmetric cryptography major part of FANETs. However, most of these
can be used as discussed in Section II-A. The hash- protocols did not consider security in the initial
ing mechanism or a digital signature can also be design. Security features in the routing proto-
incorporated to achieve the desired task. However, cols have not yet been explored in depth within
it adds to the computational complexity. the context of FANETs. The routing protocols
used in FANETs is an attractive target for the
IV. LESSONS LEARNED
attackers to control the network or disrupt the
normal operation.
This section provides an overall picture of the pro-
posed survey and emphasizes on the key lessons
learned. The survey consists of five major sections.
Sections I highlighted the issues with the security of For the successful deployment of UAVs in critical
UAVs and defined the scope of this work. Section missions, it is very important that all the vulnerabil-
II and Section III were devoted to addressing the ities in the existing security protocols are identified
security protocols used in UAVs and the vulnerabili- and removed.
ties in these protocols, respectively. The key lessons
learned from sections II and III are as follows:
VOLUME 4, 2016 21

V. CHALLENGES AND FUTURE RESEARCH brute force attacks, and to significantly reduce the
DIRECTIONS large bandwidth requirements, the key size should
With the advancement in recent UAV technology, be only large enough to resist the brute force attacks.
the use and applications of UAVs are growing expo-
nentially. There are many open research areas with
elevated levels that should be tackled in an efficient C. AUTHENTICATION OBSTACLES IN UAVS
way in order to produce secure and dependable
Authentication is a key aspect of the UAV network.
future UAV generations. Furnishing new arrange-
In fact, most of the eavesdroppers use fake infor-
ments must be limited with specific necessities and
mation, that comes in the form of signals such as
limitations like low intricacy and unwavering qual-
GPS signals and WiFi signals, to hijack the UAV.
ity. This section discusses the conceivable future
Spoofing techniques are frequently used to distract
directions for security correspondence frameworks.
the UAVs. In addition, the authentication is very
important to identify the right signal received by the
A. CONNECTIVITY INSECURITY
UAVs. Researchers are putting their efforts to tackle
There are two common ways to connect the UAVs this research area. Although a lot of information
to the ground station. The first is the connection authentication protocols have been proposed, there
through WiFi and the second is using the radio sig- is still a need for improving the existing works.
nals. Although wired connections are more secure For instance, in some of the existing authentication
than wireless connections but there are some range protocols, a single random number is used for the
limitations in wired connections. So, most of the authentication process. A single random number
UAVs are connected through wireless connections. can easily be predicted by performing a brute force
As the WiFi is protected with the password, the attack. Therefore, instead of using a single random
attacker can easily break the barrier and can get ac- number, chaotic maps must be used with an appro-
cess to the WiFi connection. The WiFi password can priate key for creating different random numbers
either be in the form of alphabets, special characters, [128]–[131]. Less dimensional chaos cannot work
numeric values, or the combination of all these properly in order to produce more random substi-
three. There are two ways to provide more security tution boxes and more random numbers. One can
to the wireless connections: one is the encryption use high dimensional chaotic maps such as hyper
methodology [126], and the other is the watermark- chaotic map to solve the generation of truly random
ing technology [127]. The novelty can be produced number [132].
in the existing security protocols by deploying Wa-
termarking with cryptography. The reason is that
when anyone secures the text by watermarking, the D. INTRUSION DETECTION SYSTEMS (IDSS) IN
watermarked text retains its meaningful content. UAVS
Hence, there is a possibility that even after hacking For intrusion detection, many machine learning al-
the password, the attacker may apply the cracked gorithms have been used. Machine learning algo-
password before applying the reverse process of the rithms predict future events based on previously
watermarking. learned data. To increase the accuracy of the ma-
chine learning models, the features should be rele-
B. DATA INSECURITY vant. For instance, if the features that are used in
Most of the data capture by the UAVs is in the the data set are irrelevant, the predicted results may
form of images which must be protected before for- be inaccurate. To improve the machine learning-
warding to the destination. Many security protocols based intrusion detection techniques, deep learning
have been presented by the researchers which are should be incorporated that can help in predicting
designed for the protection of the data. However, for the future events more accurately. Increasing the
the secure transmission of the data from the UAV number of features used to detect the intrusions can
to the ground station, robust secure algorithms are also increase the accuracy of the model. Moreover,
required which must be incorporated with different when utilizing a larger number of features, some
transforms such as discrete cosine transform, dis- attributes become useless, and hence, they will not
crete wavelet transform, and discrete Fourier trans- contribute to the accuracy of the machine learning
form. In some of the security algorithms, the key model, and they must be removed using well known
size is kept too large, due to which a large bandwidth methods such as correlation.
is required. As the large key size is required to resist
22 VOLUME 4, 2016

VI. CONCLUSION gies for uavs: A survey,” Journal of Network and Computer
Applications, p. 102739, 2020.
In this paper, we have highlighted the issues with
[12] C. M. W. Basnayaka and D. N. K. Jayakody, “Communica-
existing security protocols for UAVs. Firstly, we tion and networking technologies for uavs: A survey,” 2020.
studied the existing security protocols used in the [13] H. Shakhatreh, A. H. Sawalmeh, A. Al-Fuqaha, Z. Dou,
UAVs. Secondly, we identified the vulnerabilities in E. Almaita, I. Khalil, N. S. Othman, A. Khreishah, and
M. Guizani, “Unmanned aerial vehicles (uavs): A survey on
the existing protocols. Our survey revealed that the civil applications and key research challenges,” Ieee Access,
existing security protocols need significant improve- vol. 7, pp. 48572–48634, 2019.
ments to make the UAVs more secure. Also, the [14] V. Moss, D. Jones, and S. Nwaneri, “Analysis of homeland
security and economic survey using special missions un-
vulnerabilities identified in the existing protocols manned aerial vehicle utilities,” in 2012 IEEE International
need to be addressed in order to secure the next Geoscience and Remote Sensing Symposium, pp. 6154–
generations of UAVs. Furthermore, we have sum- 6157, IEEE, 2012.
marized some future research directions in the area [15] J. M. Hamamreh, H. M. Furqan, and H. Arslan, “Classifica-
tions and applications of physical layer security techniques
of UAV security. We believe that the strong security for confidentiality: A comprehensive survey,” IEEE Commu-
in the UAVs is a fundamental concern and we expect nications Surveys & Tutorials, vol. 21, no. 2, pp. 1773–1828,
that the researchers will show their interest to make 2018.
[16] C. L. Krishna and R. R. Murphy, “A review on cybersecurity
the UAVs more secure in the forthcoming years. vulnerabilities for unmanned aerial vehicles,” in 2017 IEEE
International Symposium on Safety, Security and Rescue
Robotics (SSRR), pp. 194–199, IEEE, 2017.
REFERENCES
[17] Y. Zhi, Z. Fu, X. Sun, and J. Yu, “Security and privacy issues
[1] K. Mabodi, M. Yusefi, S. Zandiyan, L. Irankhah, and R. Fo- of uav: a survey,” Mobile Networks and Applications, vol. 25,
tohi, “Multi-level trust-based intelligence schema for secur- no. 1, pp. 95–101, 2020.
ing of internet of things (iot) against security threats using [18] N. A. Khan, S. N. Brohi, and N. Jhanjhi, “Uav’s applications,
cryptographic authentication,” The Journal of Supercomput- architecture, security issues and attack scenarios: a survey,”
ing, pp. 1–26, 2020. in Intelligent Computing and Innovation on Data Science,
[2] H. Zhu, M. L. Cummings, M. Elfar, Z. Wang, and M. Pa- pp. 753–760, Springer, 2020.
jic, “Operator strategy model development in uav hacking [19] J. McCoy and D. B. Rawat, “Software-defined networking
detection,” IEEE Transactions on Human-Machine Systems, for unmanned aerial vehicular networking and security: A
vol. 49, no. 6, pp. 540–549, 2019. survey,” Electronics, vol. 8, no. 12, p. 1468, 2019.
[3] T. Fox-Brewster, “Maldrone: Watch mal-
[20] A. Cavoukian, Privacy and drones: Unmanned aerial vehicles.
ware that wants to spread its wings kill a
Information and Privacy Commissioner of Ontario, Canada
drone mid-flight,” Forbes Magazine, URL:
Ontario, 2012.
https://www.forbes.com/sites/thomasbrewster/2015/01/27/
[21] M. A. Kafi, Y. Challal, D. Djenouri, M. Doudou, A. Bouab-
malware-takes-down-drone/?sh=65e60af44c92, Visited on
dallah, and N. Badache, “A study of wireless sensor networks
February 27, 2021.
for urban traffic monitoring: applications and architectures,”
[4] E. G. AbdAllah, H. S. Hassanein, and M. Zulkernine, “A
Procedia computer science, vol. 19, pp. 617–626, 2013.
survey of security attacks in information-centric networking,”
[22] P. Peterson, “Technologies for homeland security (hst), 2010
IEEE Communications Surveys & Tutorials, vol. 17, no. 3,
ieee international conference on,” 2010.
pp. 1441–1454, 2015.
[5] J. Li, S. Kamin, G. Zheng, F. Neubrech, S. Zhang, and N. Liu, [23] A. Jones and G. L. Kovacich, Global Information Warfare:
“Addressable metasurfaces for dynamic holography and opti- The New Digital Battlefield. CRC Press, 2015.
cal information encryption,” Science advances, vol. 4, no. 6, [24] A. Zeitlin and M. McLaughlin, “Modeling for uas collision
p. eaar6768, 2018. avoidance,” AUVSI Unmanned Systems North America, Or-
[6] Y. Su, S. Z. F. Phua, Y. Li, X. Zhou, D. Jana, G. Liu, W. Q. lando, 2006.
Lim, W. K. Ong, C. Yang, and Y. Zhao, “Ultralong room tem- [25] F. Barfield, “Autonomous collision avoidance: the technical
perature phosphorescence from amorphous organic materials requirements,” in Proceedings of the IEEE 2000 National
toward confidential information encryption and decryption,” Aerospace and Electronics Conference. NAECON 2000. En-
Science advances, vol. 4, no. 5, p. eaas9732, 2018. gineering Tomorrow (Cat. No. 00CH37093), pp. 808–813,
[7] S. Heron, “Advanced encryption standard (aes),” Network IEEE, 2000.
Security, vol. 2009, no. 12, pp. 8–12, 2009. [26] R. Sharma and D. Ghose, “Collision avoidance between uav
[8] M. P. Arthur, “Detecting signal spoofing and jamming attacks clusters using swarm intelligence techniques,” International
in uav networks using a lightweight ids,” in 2019 Interna- Journal of Systems Science, vol. 40, no. 5, pp. 521–538,
tional Conference on Computer, Information and Telecom- 2009.
munication Systems (CITS), pp. 1–5, IEEE, 2019. [27] L. Yang, J. Qi, J. Xiao, and X. Yong, “A literature review
[9] M. Majidi, A. Erfanian, and H. Khaloozadeh, “Prediction- of uav 3d path planning,” in Proceeding of the 11th World
discrepancy based on innovative particle filter for estimating Congress on Intelligent Control and Automation, pp. 2376–
uav true position in the presence of the gps spoofing attacks,” 2381, IEEE, 2014.
IET Radar, Sonar & Navigation, vol. 14, no. 6, pp. 887–897, [28] J. Awrejcewicz, Numerical analysis: Theory and application.
2020. BoD–Books on Demand, 2011.
[10] A. Abbaspour, K. K. Yen, S. Noei, and A. Sargolzaei, “Detec- [29] A. M. Brandt and M. B. Colton, “Haptic collision avoidance
tion of fault data injection attack on uav using adaptive neural for a remotely operated quadrotor uav in indoor environ-
network,” Procedia computer science, vol. 95, pp. 193–200, ments,” in 2010 IEEE International Conference on Systems,
2016. Man and Cybernetics, pp. 2724–2731, IEEE, 2010.
[11] A. Sharma, P. Vanjani, N. Paliwal, C. M. W. Bas- [30] J. Israelsen, M. Beall, D. Bareiss, D. Stuart, E. Keeney,
nayaka, D. N. K. Jayakody, H.-C. Wang, and P. Muthuchi- and J. van den Berg, “Automatic collision avoidance for
dambaranathan, “Communication and networking technolo- manually tele-operated unmanned aerial vehicles,” in 2014

VOLUME 4, 2016 23

IEEE International Conference on Robotics and Automation [49] B. Deebak and F. Al-Turjman, “A smart lightweight privacy
(ICRA), pp. 6638–6643, IEEE, 2014. preservation scheme for iot-based uav communication sys-
[31] S. Atoev, O.-J. Kwon, C.-Y. Kim, S.-H. Lee, Y.-R. Choi, tems,” Computer Communications, vol. 162, pp. 102–117,
and K.-R. Kwon, “The secure uav communication link based 2020.
on otp encryption technique,” in 2019 Eleventh International [50] L. Teng, M. Jianfeng, F. Pengbin, M. Yue, M. Xindi, Z. Ji-
Conference on Ubiquitous and Future Networks (ICUFN), awei, C. Gao, and L. Di, “Lightweight security authenti-
pp. 1–3, IEEE, 2019. cation mechanism towards uav networks,” in 2019 Interna-
[32] Y. Niu, S. Xu, L. Wu, and W. Hu, “Airborne infrared and vis- tional Conference on Networking and Network Applications
ible image fusion for target perception based on target region (NaNA), pp. 379–384, IEEE, 2019.
segmentation and discrete wavelet transform,” Mathematical [51] E. Barka, C. A. Kerrache, R. Hussain, N. Lagraa, A. Lakas,
Problems in Engineering, vol. 2012, 2012. and S. H. Bouk, “A trusted lightweight communication strat-
[33] Y. Ma, X. Wu, G. Yu, Y. Xu, and Y. Wang, “Pedestrian egy for flying named data networking,” Sensors, vol. 18,
detection and tracking from low-resolution unmanned aerial no. 8, p. 2683, 2018.
vehicle thermal imagery,” Sensors, vol. 16, no. 4, p. 446, [52] J. Srinivas, A. K. Das, N. Kumar, and J. J. Rodrigues,
2016. “Tcalas: Temporal credential-based anonymous lightweight
[34] V. Kirichenko, “Information security of communication authentication scheme for internet of drones environment,”
channel with uav,” Electronics and control systems, no. 3, IEEE Transactions on Vehicular Technology, vol. 68, no. 7,
pp. 23–27, 2015. pp. 6903–6916, 2019.
[35] C. E. Shannon, “Communication theory of secrecy systems,” [53] Z. Ali, S. A. Chaudhry, M. S. Ramzan, and F. Al-Turjman,
The Bell system technical journal, vol. 28, no. 4, pp. 656–715, “Securing smart city surveillance: a lightweight authentica-
1949. tion mechanism for unmanned vehicles,” IEEE Access, vol. 8,
[36] K. D. Wesson, T. E. Humphreys, and B. L. Evans, “Can pp. 43711–43724, 2020.
cryptography secure next generation air traffic surveillance?,” [54] Q. Li, Y. Yang, W.-K. Ma, M. Lin, J. Ge, and J. Lin,
IEEE Security and Privacy Magazine, 2014. “Robust cooperative beamforming and artificial noise design
[37] W.-J. Pan, Z.-L. Feng, and Y. Wang, “Ads-b data authentica- for physical-layer secrecy in af multi-antenna multi-relay
tion based on ecc and x. 509 certificate,” Journal of Electronic networks,” IEEE Transactions on Signal Processing, vol. 63,
Science and Technology, vol. 10, no. 1, pp. 51–55, 2012. no. 1, pp. 206–220, 2014.
[38] E. Bresson, O. Chevassut, and D. Pointcheval, “Provably se- [55] L. Sun and Q. Du, “Physical layer security with its appli-
cure authenticated group diffie-hellman key exchange,” ACM cations in 5g networks: A review,” China Communications,
Transactions on Information and System Security (TISSEC), vol. 14, no. 12, pp. 1–14, 2017.
vol. 10, no. 3, pp. 10–es, 2007. [56] N. Yang, L. Wang, G. Geraci, M. Elkashlan, J. Yuan, and
[39] O. K. Sahingoz, “Multi-level dynamic key management for M. Di Renzo, “Safeguarding 5g wireless communication net-
scalable wireless sensor networks with uav,” in Ubiqui- works using physical layer security,” IEEE Communications
tous Information Technologies and Applications, pp. 11–19, Magazine, vol. 53, no. 4, pp. 20–27, 2015.
Springer, 2013. [57] Y. Zeng, R. Zhang, and T. J. Lim, “Throughput maximization
[40] V. Valentin-Alexandru, B. Ion, and P. Victor-Valeriu, “Energy for uav-enabled mobile relaying systems,” IEEE Transactions
efficient trust-based security mechanism for wireless sensors on Communications, vol. 64, no. 12, pp. 4983–4996, 2016.
and unmanned aerial vehicles,” in 2019 11th International [58] D. H. Choi, S. H. Kim, and D. K. Sung, “Energy-efficient ma-
Conference on Electronics, Computers and Artificial Intelli- neuvering and communication of a single uav-based relay,”
gence (ECAI), pp. 1–6, IEEE, 2019. IEEE Transactions on Aerospace and Electronic Systems,
[41] K. Yoon, D. Park, Y. Yim, K. Kim, S. K. Yang, and M. Robin- vol. 50, no. 3, pp. 2320–2327, 2014.
son, “Security authentication system using encrypted channel [59] K. Li, R. C. Voicu, S. S. Kanhere, W. Ni, and E. Tovar,
on uav network,” in 2017 First IEEE International Conference “Energy efficient legitimate wireless surveillance of uav com-
on Robotic Computing (IRC), pp. 393–398, IEEE, 2017. munications,” IEEE Transactions on Vehicular Technology,
[42] J. A. Steinmann, R. F. Babiceanu, and R. Seker, “Uas se- vol. 68, no. 3, pp. 2283–2293, 2019.
curity: Encryption key negotiation for partitioned data,” in [60] Q. Wang, Z. Chen, W. Mei, and J. Fang, “Improving physi-
2016 Integrated Communications Navigation and Surveil- cal layer security using uav-enabled mobile relaying,” IEEE
lance (ICNS), pp. 1E4–1, IEEE, 2016. Wireless Communications Letters, vol. 6, no. 3, pp. 310–313,
[43] K. Driscoll, “Lightweight crypto for lightweight unmanned 2017.
arial systems,” in 2018 Integrated Communications, Naviga- [61] X. Lu, L. Xiao, C. Dai, and H. Dai, “Uav-aided cellular
tion, Surveillance Conference (ICNS), pp. 1–15, IEEE, 2018. communications with deep reinforcement learning against
[44] J. Wang, Y. Liu, S. Niu, and H. Song, “Lightweight jamming,” arXiv preprint arXiv:1805.06628, 2018.
blockchain assisted secure routing of swarm uas networking,” [62] W.-S. Ra, I.-H. Whang, and J. B. Park, “Robust weighted least
Computer Communications, 2020. squares range estimator for uav applications,” in 2008 SICE
[45] A. Demeri, W. Diehl, and A. Salman, “Saddle: Secure aerial Annual Conference, pp. 251–255, IEEE, 2008.
data delivery with lightweight encryption,” in Science and [63] M. B. Bejiga, A. Zeggada, A. Nouffidj, and F. Melgani,
Information Conference, pp. 204–223, Springer, 2020. “A convolutional neural network approach for assisting
[46] C. Pu and Y. Li, “Lightweight authentication protocol for avalanche search and rescue operations with uav imagery,”
unmanned aerial vehicles using physical unclonable function Remote Sensing, vol. 9, no. 2, p. 100, 2017.
and chaotic system,” in 2020 IEEE International Sympo- [64] A. Nemes and G. Mester, “Unconstrained evolutionary and
sium on Local and Metropolitan Area Networks (LANMAN, gradient descent-based tuning of fuzzy-partitions for uav
pp. 1–6, IEEE, 2020. dynamic modeling,” FME Transactions, vol. 45, no. 1, pp. 1–
[47] M. S. Haque and M. U. Chowdhury, “A new cyber se- 8, 2017.
curity framework towards secure data communication for [65] T. M. Hoang, N. M. Nguyen, and T. Q. Duong, “Detection
unmanned aerial vehicle (uav),” in International Conference of eavesdropping attack in uav-aided wireless systems: Un-
on Security and Privacy in Communication Systems, pp. 113– supervised learning with one-class svm and k-means cluster-
122, Springer, 2017. ing,” IEEE Wireless Communications Letters, vol. 9, no. 2,
[48] T. Alladi, V. Chamola, N. Kumar, et al., “Parth: A two-stage pp. 139–142, 2019.
lightweight mutual authentication protocol for uav surveil- [66] H. Lee, S. Eom, J. Park, and I. Lee, “Uav-aided secure com-
lance networks,” Computer Communications, 2020. munications with cooperative jamming,” IEEE Transactions

24 VOLUME 4, 2016

on Vehicular Technology, vol. 67, no. 10, pp. 9385–9392, munication under uav smart attack with imperfect channel
2018. estimation,” IEEE Access, vol. 6, pp. 76395–76401, 2018.
[67] H. Sedjelmaci, S. M. Senouci, and N. Ansari, “Intrusion de- [85] K. Amelin, S. Tomashevich, and B. Andrievsky, “Recursive
tection and ejection framework against lethal attacks in uav- identification of motion model parameters for ultralight uav,”
aided networks: A bayesian game-theoretic methodology,” IFAC-PapersOnLine, vol. 48, no. 11, pp. 233–237, 2015.
IEEE Transactions on Intelligent Transportation Systems, [86] M. Hooper, Y. Tian, R. Zhou, B. Cao, A. P. Lauf, L. Watkins,
vol. 18, no. 5, pp. 1143–1153, 2016. W. H. Robinson, and W. Alexis, “Securing commercial
[68] J. Xu, Z. Deng, Q. Song, Q. Chi, T. Wu, Y. Huang, D. Liu, wifi-based uavs from common security attacks,” in MIL-
and M. Gao, “Multi-uav counter-game model based on un- COM 2016-2016 IEEE Military Communications Confer-
certain information,” Applied Mathematics and Computation, ence, pp. 1213–1218, IEEE, 2016.
vol. 366, p. 124684, 2020. [87] K.-W. Huang and H.-M. Wang, “Combating the control sig-
[69] M. Z. Anwar, Z. Kaleem, and A. Jamalipour, “Machine nal spoofing attack in uav systems,” IEEE Transactions on
learning inspired sound-based amateur drone detection for Vehicular Technology, vol. 67, no. 8, pp. 7769–7773, 2018.
public safety applications,” IEEE Transactions on Vehicular [88] L. Zhang, G. Ding, Q. Wu, and P. Liu, “Detection of abnor-
Technology, vol. 68, no. 3, pp. 2526–2534, 2019. mal power emission in uav communication networks,” IEEE
[70] A. Shoufan, H. M. Al-Angari, M. F. A. Sheikh, and E. Dami- Wireless Communications Letters, vol. 8, no. 4, pp. 1179–
ani, “Drone pilot identification by classifying radio-control 1182, 2019.
signals,” IEEE Transactions on Information Forensics and [89] H. Sedjelmaci, S. M. Senouci, and N. Ansari, “A hierarchical
Security, vol. 13, no. 10, pp. 2439–2447, 2018. detection and response system to enhance security against
[71] M. Ezuma, F. Erden, C. K. Anjinappa, O. Ozdemir, and lethal cyber-attacks in uav networks,” IEEE Transactions on
I. Guvenc, “Detection and classification of uavs using rf fin- Systems, Man, and Cybernetics: Systems, vol. 48, no. 9,
gerprints in the presence of wi-fi and bluetooth interference,” pp. 1594–1606, 2017.
IEEE Open Journal of the Communications Society, vol. 1, [90] Y. Qiao, Y. Zhang, and X. Du, “A vision-based gps-spoofing
pp. 60–76, 2019. detection method for small uavs,” in 2017 13th Interna-
[72] F. Li, J. Xin, T. Chen, L. Xin, Z. Wei, Y. Li, Y. Zhang, H. Jin, tional Conference on Computational Intelligence and Secu-
Y. Tu, X. Zhou, et al., “An automatic detection method of rity (CIS), pp. 312–316, IEEE, 2017.
bird’s nest on transmission line tower based on faster_rcnn,” [91] S. Kamate and N. Yilmazer, “Application of object detection
IEEE Access, vol. 8, pp. 164214–164221, 2020. and tracking techniques for unmanned aerial vehicles,” Pro-
[73] A. Alipour-Fanid, M. Dabaghchian, N. Wang, P. Wang, cedia Computer Science, vol. 61, pp. 436–441, 2015.
L. Zhao, and K. Zeng, “Machine learning-based delay-aware [92] D. Davidson, H. Wu, R. Jellinek, V. Singh, and T. Risten-
uav detection over encrypted wi-fi traffic,” in 2019 IEEE Con- part, “Controlling uavs with sensor input spoofing attacks,”
ference on Communications and Network Security (CNS), in 10th {USENIX} Workshop on Offensive Technologies
pp. 1–7, IEEE, 2019. ({WOOT} 16), 2016.
[74] J. Han, Z. Yang, H. Xu, G. Hu, C. Zhang, H. Li, S. Lai, [93] Y.-M. Kwon, J. Yu, B.-M. Cho, Y. Eun, and K.-J. Park,
and H. Zeng, “Search like an eagle: A cascaded model for “Empirical analysis of mavlink protocol vulnerability for
insulator missing faults detection in aerial images,” Energies, attacking unmanned aerial vehicles,” IEEE Access, vol. 6,
vol. 13, no. 3, p. 713, 2020. pp. 43203–43212, 2018.
[94] C. Liu, J. Lee, and T. Q. Quek, “Safeguarding uav com-
[75] M. Faraji-Biregani and R. Fotohi, “Secure communication
munications against full-duplex active eavesdropper,” IEEE
between uavs using a method based on smart agents in
Transactions on Wireless Communications, vol. 18, no. 6,
unmanned aerial vehicles,” The Journal of Supercomputing,
pp. 2919–2931, 2019.
pp. 1–28, 2020.
[95] Z. Mobini, B. K. Chalise, M. Mohammadi, H. A. Suraweera,
[76] Y. Li, R. Zhang, J. Zhang, and L. Yang, “Cooperative
and Z. Ding, “Proactive eavesdropping using uav systems
jamming via spectrum sharing for secure uav communica-
with full-duplex ground terminals,” in 2018 IEEE Inter-
tions,” IEEE Wireless Communications Letters, vol. 9, no. 3,
national Conference on Communications Workshops (ICC
pp. 326–330, 2019.
Workshops), pp. 1–6, IEEE, 2018.
[77] W. D. Scheller, “Detecting drones using machine learning,”
[96] R. Rysdyk, “Uav path following for constant line-of-sight,”
2017.
in 2nd AIAA" Unmanned Unlimited" Conf. and Workshop &
[78] W. Hong, L. Jianhua, L. Chengzhe, and W. Zhe, “A provably Exhibit, p. 6626, 2003.
secure aggregate authentication scheme for unmanned aerial [97] M. Hong, T.-H. Chang, X. Wang, M. Razaviyayn, S. Ma, and
vehicle cluster networks,” Peer-to-Peer Networking and Ap- Z.-Q. Luo, “A block successive upper-bound minimization
plications, vol. 13, no. 1, pp. 53–63, 2020. method of multipliers for linearly constrained convex opti-
[79] U. Challita, A. Ferdowsi, M. Chen, and W. Saad, “Machine mization,” Mathematics of Operations Research, 2020.
learning for wireless connectivity and security of cellular- [98] Z. Birnbaum, A. Dolgikh, V. Skormin, E. O’Brien, D. Muller,
connected uavs,” IEEE Wireless Communications, vol. 26, and C. Stracquodaine, “Unmanned aerial vehicle security
no. 1, pp. 28–35, 2019. using recursive parameter estimation,” Journal of Intelligent
[80] A. Li, W. Zhang, and S. Dou, “Uav-enabled secure data & Robotic Systems, vol. 84, no. 1-4, pp. 107–120, 2016.
dissemination via artificial noise: Joint trajectory and com- [99] S. Utsai and R. B. Joshi, “Dos attack reduction by using web
munication optimization,” IEEE Access, 2020. service filter,” International Journal of Computer Applica-
[81] B. Hudson, “Drone attacks are essentially terrorism by joy- tions, vol. 105, no. 14, 2014.
stick,” The Washington Post, 2018. [100] A. Kolichtchak, “Buffer overflow attack detection and sup-
[82] Q. Zou, S. Huang, F. Lin, and M. Cong, “Detection of gps pression,” Jan. 16 2003. US Patent App. 09/904,502.
spoofing based on uav model estimation,” in IECON 2016- [101] A. Crosland, R. May, E. Flaherty, and A. Draper, “Embedded
42nd Annual Conference of the IEEE Industrial Electronics processor with watchdog timer for programmable logic,”
Society, pp. 6097–6102, IEEE, 2016. Mar. 4 2008. US Patent 7,340,596.
[83] R. Fotohi, E. Nazemi, and F. S. Aliee, “An agent-based self- [102] W. Gruszczyński, E. Puniach, P. Ćwiakała,
˛ and W. Matwij,
protective method to secure communication between uavs in “Application of convolutional neural networks for low veg-
unmanned aerial vehicle networks,” Vehicular Communica- etation filtering from data acquired by uavs,” ISPRS Journal
tions, p. 100267, 2020. of Photogrammetry and Remote Sensing, vol. 158, pp. 1–10,
[84] C. Li, Y. Xu, J. Xia, and J. Zhao, “Protecting secure com- 2019.

VOLUME 4, 2016 25

[103] M. Talha, F. Asghar, A. Rohan, M. Rabah, and S. H. Kim, attacks,” in 2017 2nd international conference on communi-
“Fuzzy logic-based robust and autonomous safe landing for cation and electronics systems (ICCES), pp. 505–510, IEEE,
uav quadcopter,” Arabian Journal for Science and Engineer- 2017.
ing, vol. 44, no. 3, pp. 2627–2639, 2019. [121] K. Wang, C. Pan, H. Ren, W. Xu, L. Zhang, and A. Nal-
[104] I. Bisio, C. Garibotto, F. Lavagetto, A. Sciarrone, and S. Zap- lanathan, “Packet error probability and effective through-
patore, “Blind detection: Advanced techniques for wifi-based put for ultra-reliable and low-latency uav communications,”
drone surveillance,” IEEE Transactions on Vehicular Tech- IEEE Transactions on Communications, 2020.
nology, vol. 68, no. 1, pp. 938–946, 2018. [122] V. Desnitsky, N. Rudavin, and I. Kotenko, “Modeling
[105] W.-H. Au, K. C. Chan, and X. Yao, “A novel evolutionary and evaluation of battery depletion attacks on unmanned
data mining algorithm with applications to churn prediction,” aerial vehicles in crisis management systems,” in Interna-
IEEE transactions on evolutionary computation, vol. 7, no. 6, tional Symposium on Intelligent and Distributed Computing,
pp. 532–545, 2003. pp. 323–332, Springer, 2019.
[106] Y. Zhou, P. L. Yeoh, H. Chen, Y. Li, R. Schober, L. Zhuo, [123] S. Abbas, M. Faisal, H. U. Rahman, M. Z. Khan, M. Merabti,
and B. Vucetic, “Improving physical layer security via a et al., “Masquerading attacks detection in mobile ad hoc
uav friendly jammer for unknown eavesdropper location,” networks,” IEEE Access, vol. 6, pp. 55013–55025, 2018.
IEEE Transactions on Vehicular Technology, vol. 67, no. 11, [124] R. H. Jhaveri, A. D. Patel, J. D. Parmar, B. I. Shah, et al.,
pp. 11280–11284, 2018. “Manet routing protocols and wormhole attack against aodv,”
[107] N. Vanitha and G. Padmavathi, “A comparative study on International Journal of Computer Science and Network Se-
communication architecture of unmanned aerial vehicles and curity, vol. 10, no. 4, pp. 12–18, 2010.
security analysis of false data dissemination attacks,” in 2018 [125] A. Kamble, V. S. Malemath, and D. Patil, “Security attacks
International Conference on Current Trends towards Con- and secure routing protocols in rpl-based internet of things:
verging Technologies (ICCTCT), pp. 1–8, IEEE, 2018. Survey,” in 2017 International Conference on Emerging
[108] J.-A. Maxa, M. S. B. Mahmoud, and N. Larrieu, “Survey on Trends & Innovation in ICT (ICEI), pp. 33–39, IEEE, 2017.
uaanet routing protocols and network security challenges.,” [126] V. Kriz and P. Gabrlik, “Uranuslink-communication protocol
Adhoc & Sensor Wireless Networks, vol. 37, 2017. for uav with small overhead and encryption ability,” IFAC-
[109] C. Pu, “Link-quality and traffic-load aware routing for uav PapersOnLine, vol. 48, no. 4, pp. 474–479, 2015.
ad hoc networks,” in 2018 IEEE 4th international conference [127] M. P. Marcinak and B. G. Mobasseri, “Digital video wa-
on collaboration and Internet computing (CIC), pp. 71–79, termarking for metadata embedding in uav video,” in MIL-
IEEE, 2018. COM 2005-2005 IEEE Military Communications Confer-
[110] B.-S. Kim, K.-I. Kim, B. Roh, and H. Choi, “A new routing ence, pp. 1637–Vol, IEEE, 2005.
protocol for uav relayed tactical mobile ad hoc networks,” [128] Z. Cheng, Y. X. Tang, and Y. L. Liu, “3-d path planning for
in 2018 Wireless Telecommunications Symposium (WTS), uav based on chaos particle swarm optimization,” in Applied
pp. 1–4, IEEE, 2018. Mechanics and Materials, vol. 232, pp. 625–630, Trans Tech
Publ, 2012.
[111] C. Pu and L. Carpenter, “To route or to ferry: A hybrid packet
[129] M. Rosalie, G. Danoy, S. Chaumette, and P. Bouvry, “Chaos-
forwarding algorithm in flying ad hoc networks,” in 2019
enhanced mobility models for multilevel swarms of uavs,”
IEEE 18th International Symposium on Network Computing
Swarm and evolutionary computation, vol. 41, pp. 36–48,
and Applications (NCA), pp. 1–8, IEEE, 2019.
2018.
[112] M. Islabudeen and M. K. Devi, “A smart approach for in-
[130] M. Rosalie, G. Danoy, S. Chaumette, and P. Bouvry, “From
trusion detection and prevention system in mobile ad hoc
random process to chaotic behavior in swarms of uavs,” in
networks against security attacks,” Wireless Personal Com-
Proceedings of the 6th ACM Symposium on Development
munications, vol. 112, no. 1, pp. 193–224, 2020.
and Analysis of Intelligent Vehicular Networks and Applica-
[113] A. Mondal and S. Mitra, “Security issues in vehicular ad
tions, pp. 9–15, 2016.
hoc networks for evolution towards internet of vehicles,” in
[131] J. Sun, W. Wang, L. Kou, Y. Lin, L. Zhang, Q. Da, and
Connected Vehicles in the Internet of Things, pp. 253–307,
L. Chen, “A data authentication scheme for uav ad hoc
Springer, 2020.
network communication,” The Journal of Supercomputing,
[114] A. K. Malhi, S. Batra, and H. S. Pannu, “Security of vehicular vol. 76, no. 6, pp. 4041–4056, 2020.
ad-hoc networks: A comprehensive survey,” Computers & [132] P. S. Gohari, H. Mohammadi, and S. Taghvaei, “Using
Security, vol. 89, p. 101664, 2020. chaotic maps for 3d boundary surveillance by quadrotor
[115] N. M. Rodday, R. d. O. Schmidt, and A. Pras, “Exploring se- robot,” Applied Soft Computing, vol. 76, pp. 68–77, 2019.
curity vulnerabilities of unmanned aerial vehicles,” in NOMS
2016-2016 IEEE/IFIP Network Operations and Management
Symposium, pp. 993–994, IEEE, 2016.
[116] M. O. Kalinin and A. Minin, “Security evaluation of a
wireless ad-hoc network with dynamic topology,” Automatic
Control and Computer Sciences, vol. 51, no. 8, pp. 899–901, ARSLAN SHAFIQUE received the
2017. B.E. and M.S degree in Mechatronics
[117] J.-A. Maxa, M. S. B. Mahmoud, and N. Larrieu, “Perfor- and Electrical Engineering from Wah
mance evaluation of a new secure routing protocol for uav Engineering College and Heavy Indus-
ad hoc network,” in 2019 IEEE/AIAA 38th Digital Avionics tries Taxila Education City (HITEC)
Systems Conference (DASC), pp. 1–10, IEEE, 2019. University, Pakistan in 2014 and 2017
[118] M. J. Alala, K. I. Khorzom, and W. Y. Aljuneidi, “Ef- respectively. He has five journal publi-
fects of communication channel on aodv performance within
cations with accumulative impact fac-
uaanets,” International Journal of Communications, vol. 3,
tor of 14.54. He is currently pursuing
2018.
a Ph.D. degree and also serving as a
[119] H. Nawaz and H. M. Ali, “Implementation of cross layer
design for efficient power and routing in uav communication
Research Associate in the Faculty of Engineering and Applied
networks,” Stud. Informat. Control, vol. 29, no. 1, pp. 111– Sciences at Riphah International University, Islamabad, Pakistan.
120, 2020. His research interests include cryptography, secure communica-
[120] D. Sasirekha and N. Radha, “Secure and attack aware routing tion, and machine learning.
in mobile ad hoc networks against wormhole and sinkhole

26 VOLUME 4, 2016

ABID MEHMOOD received his Ph.D.

degree in computer science from
Deakin University, Australia. He is cur-
rently an Assistant Professor with Abu
Dhabi University. His research inter-
ests include information security and
privacy, data mining, machine learning
and cloud computing.

MOURAD ELHADEF is a professor

of Computer Science in the College of
Engineering at Abu Dhabi University,
UAE. He has a Ph.D. degree in Com-
puter Science from the University of
Sherbrooke, Québec, Canada, and his
B.Sc. and M.Sc. degrees in Computer
Science from the Institute Supérieur de
Gestion, Tunis, Tunisia. His current re-
search interests include fault tolerance
and fault diagnosis in distributed, wireless and ad-hoc networks,
cloud computing, artificial intelligence, and security. Dr. Elhadef
has over 50 publications in refereed journals and conference
proceedings. He is an active reviewer for various international
conferences and journals such as IEEE transactions on Parallel
and Distributed Systems, Journal of Parallel and Distributed
Computing.

VOLUME 4, 2016 27

State-Of-The-Art and Future Research Challenges in UAV Swarms
No ratings yet
State-Of-The-Art and Future Research Challenges in UAV Swarms
23 pages
Enabling Drones in The Internet of Things With Decentralized Blockchain-Based Security
No ratings yet
Enabling Drones in The Internet of Things With Decentralized Blockchain-Based Security
10 pages
Fusion Queries
100% (2)
Fusion Queries
4 pages
Journal of Network and Computer Applications: Hassan Jalil Hadi Yue Cao Khaleeq Un Nisa Abdul Majid Jamil Qiang Ni
No ratings yet
Journal of Network and Computer Applications: Hassan Jalil Hadi Yue Cao Khaleeq Un Nisa Abdul Majid Jamil Qiang Ni
26 pages
Sensors: Drone Secure Communication Protocol For Future Sensitive Applications in Military Zone
No ratings yet
Sensors: Drone Secure Communication Protocol For Future Sensitive Applications in Military Zone
25 pages
A Survey of Cyberattack Countermeasures For Unmann
No ratings yet
A Survey of Cyberattack Countermeasures For Unmann
20 pages
Drone_Secure_Communication_Protocol_for_Future_Sen
No ratings yet
Drone_Secure_Communication_Protocol_for_Future_Sen
25 pages
Intrusion Detection Systems For Networked Unmanned Aerial Vehicles A Survey
No ratings yet
Intrusion Detection Systems For Networked Unmanned Aerial Vehicles A Survey
6 pages
Communication Security of Unmanned Aerial Vehicles
No ratings yet
Communication Security of Unmanned Aerial Vehicles
6 pages
Electronics: Software-Defined Networking For Unmanned Aerial Vehicular Networking and Security: A Survey
No ratings yet
Electronics: Software-Defined Networking For Unmanned Aerial Vehicular Networking and Security: A Survey
25 pages
A Multifractal Analysis and Machine Learning Based Intrusion Detection System With An Application in A UAS-RADAR System
No ratings yet
A Multifractal Analysis and Machine Learning Based Intrusion Detection System With An Application in A UAS-RADAR System
16 pages
Defense Techniques Against Cyber Attacks On Unmanned Aerial Vehicles
No ratings yet
Defense Techniques Against Cyber Attacks On Unmanned Aerial Vehicles
7 pages
Electronics 10 00999
No ratings yet
Electronics 10 00999
30 pages
Csit89708 PDF
No ratings yet
Csit89708 PDF
11 pages
Electronics 10 01549 v2
No ratings yet
Electronics 10 01549 v2
28 pages
Deep Learning For Unmanned Aerial Vehicles Detection - A Review
No ratings yet
Deep Learning For Unmanned Aerial Vehicles Detection - A Review
26 pages
Security Threats and Mitigation Techniques in UAV Communications A Comprehensive Survey
No ratings yet
Security Threats and Mitigation Techniques in UAV Communications A Comprehensive Survey
40 pages
2105.01347v1
No ratings yet
2105.01347v1
31 pages
Drone PDF
No ratings yet
Drone PDF
18 pages
A Drone-Aided Blockchain-Based Smart Vehicular Network
No ratings yet
A Drone-Aided Blockchain-Based Smart Vehicular Network
11 pages
1 s2.0 S0140366419319358 Main
No ratings yet
1 s2.0 S0140366419319358 Main
10 pages
TCALAS: Temporal Credential-Based Anonymous Lightweight Authentication Scheme For Internet of Drones Environment
No ratings yet
TCALAS: Temporal Credential-Based Anonymous Lightweight Authentication Scheme For Internet of Drones Environment
14 pages
2022 - Akshita - Flying Through The Secure Fog A Complete Study On UAV Fog in Heterogeneous Networks
No ratings yet
2022 - Akshita - Flying Through The Secure Fog A Complete Study On UAV Fog in Heterogeneous Networks
41 pages
A Swarm Anomaly Detection Model for IoT UAVs Based on a Multi-modal Denoising Autoencoder and Federated Learning
No ratings yet
A Swarm Anomaly Detection Model for IoT UAVs Based on a Multi-modal Denoising Autoencoder and Federated Learning
22 pages
1-s2.0-S2772671124000536-main
No ratings yet
1-s2.0-S2772671124000536-main
15 pages
Tutorial On UAVs PDF
No ratings yet
Tutorial On UAVs PDF
28 pages
Application of Blockchain in UAV_2
No ratings yet
Application of Blockchain in UAV_2
24 pages
A Review of Cyber Security Challenges and Solution-1
No ratings yet
A Review of Cyber Security Challenges and Solution-1
21 pages
218 857 2 PB
No ratings yet
218 857 2 PB
5 pages
Survey On Anti-Drone Systems Components Designs An
No ratings yet
Survey On Anti-Drone Systems Components Designs An
25 pages
RozehaARashid2021_DevelopmentofanAutonomousIoTBasedDrone
No ratings yet
RozehaARashid2021_DevelopmentofanAutonomousIoTBasedDrone
7 pages
Blockchain-Based_Trust_and_Authentication_Model_fo
No ratings yet
Blockchain-Based_Trust_and_Authentication_Model_fo
13 pages
Countering Large-Scale Drone Swarm Attack by Efficient Splitting
No ratings yet
Countering Large-Scale Drone Swarm Attack by Efficient Splitting
13 pages
UAV Payload and Misssion Control HW/SW Architecture
No ratings yet
UAV Payload and Misssion Control HW/SW Architecture
6 pages
Security and Privacy - 2023 - Mekdad - A Comprehensive Security and Performance Assessment of UAV Authentication Schemes
No ratings yet
Security and Privacy - 2023 - Mekdad - A Comprehensive Security and Performance Assessment of UAV Authentication Schemes
24 pages
Attacks Against Mobile Ad Hoc Networks Routing Protocols: S. A. Razak, S. M. Furnell, P. J. Brooke
No ratings yet
Attacks Against Mobile Ad Hoc Networks Routing Protocols: S. A. Razak, S. M. Furnell, P. J. Brooke
6 pages
Mee 789
No ratings yet
Mee 789
19 pages
RF-Enabled Deep-Learning-Assisted Drone Detection and Identification - An End-To-End Approach
No ratings yet
RF-Enabled Deep-Learning-Assisted Drone Detection and Identification - An End-To-End Approach
18 pages
Review 3
No ratings yet
Review 3
24 pages
Threats To UAVs (20210812, Flashpoint)
No ratings yet
Threats To UAVs (20210812, Flashpoint)
8 pages
Sustainability 14 04861
No ratings yet
Sustainability 14 04861
17 pages
Building Trust For Smart Connected Devices: The Challenges and Pitfalls of TrustZone
No ratings yet
Building Trust For Smart Connected Devices: The Challenges and Pitfalls of TrustZone
46 pages
Securing The Skies
No ratings yet
Securing The Skies
22 pages
Security in Internet of Drones A Comprehensive Review
No ratings yet
Security in Internet of Drones A Comprehensive Review
81 pages
2019 Secure UAV Communication Networks Over 5G
No ratings yet
2019 Secure UAV Communication Networks Over 5G
7 pages
Abbreviated UAV Cloud Integration Proposal - (Katende Marvin and Ayebare Anil Pelice)
No ratings yet
Abbreviated UAV Cloud Integration Proposal - (Katende Marvin and Ayebare Anil Pelice)
5 pages
Remotesensing 16 00879 v2
No ratings yet
Remotesensing 16 00879 v2
42 pages
sensors-23-03215-v2
No ratings yet
sensors-23-03215-v2
29 pages
Detection of UAS Paper
No ratings yet
Detection of UAS Paper
31 pages
Internet of Drones
No ratings yet
Internet of Drones
15 pages
Contingency Clarification Protocols For Reliable C
No ratings yet
Contingency Clarification Protocols For Reliable C
11 pages
Forensic UAV
No ratings yet
Forensic UAV
29 pages
Drones 06 00147
No ratings yet
Drones 06 00147
27 pages
A Comprehensive Survey On 5G-And-Beyond Networks With UAVs Applications Emerging Technologies Regulatory Aspects Research Trends and Challenges
No ratings yet
A Comprehensive Survey On 5G-And-Beyond Networks With UAVs Applications Emerging Technologies Regulatory Aspects Research Trends and Challenges
41 pages
IoT and Machine Learning Based UAV For Surveillance Along India's International Borders
No ratings yet
IoT and Machine Learning Based UAV For Surveillance Along India's International Borders
10 pages
Drone Hacking. Mirai in The Sky. Clara and Jason Andress
No ratings yet
Drone Hacking. Mirai in The Sky. Clara and Jason Andress
7 pages
Survey On Anti-Drone Systems Components Designs and Challenges
No ratings yet
Survey On Anti-Drone Systems Components Designs and Challenges
25 pages
Drones 06 00009 With Cover
No ratings yet
Drones 06 00009 With Cover
29 pages
Tee 3444
No ratings yet
Tee 3444
14 pages
Network Security Traceback Attack and React in the United States Department of Defense Network
From Everand
Network Security Traceback Attack and React in the United States Department of Defense Network
Edmond K. Machie
No ratings yet
Drone Technology
From Everand
Drone Technology
Zuri Deepwater
No ratings yet
[Soal Materi] Main Ideas Paket 1
No ratings yet
[Soal Materi] Main Ideas Paket 1
5 pages
Mazhar Abbas
No ratings yet
Mazhar Abbas
2 pages
CONTEMPORARY WORLD Reviewer
No ratings yet
CONTEMPORARY WORLD Reviewer
5 pages
pEOPLE VS. IBANEZ - TORTS
No ratings yet
pEOPLE VS. IBANEZ - TORTS
2 pages
Form 4 Biology Notes Latest
No ratings yet
Form 4 Biology Notes Latest
104 pages
Guidelines For Scanning and Uploading of Photograph & Signature
No ratings yet
Guidelines For Scanning and Uploading of Photograph & Signature
3 pages
Seangio Vs Reyes PDF Will and Testament Intestacy
No ratings yet
Seangio Vs Reyes PDF Will and Testament Intestacy
1 page
Metric Conversion
No ratings yet
Metric Conversion
5 pages
Assignment 3
No ratings yet
Assignment 3
9 pages
Fordyce MW 1983
No ratings yet
Fordyce MW 1983
46 pages
CBE Online Application System (COAS)
No ratings yet
CBE Online Application System (COAS)
1 page
Experiment 4 - Orientation Effect
No ratings yet
Experiment 4 - Orientation Effect
5 pages
Stavanger, Norway: Finn Tengs Christensen
No ratings yet
Stavanger, Norway: Finn Tengs Christensen
6 pages
Grain Size DHP
No ratings yet
Grain Size DHP
6 pages
Conventional Paper-I - EC
No ratings yet
Conventional Paper-I - EC
8 pages
Python Question Bank With Answers - End Semester (By Ur Engineering Friend)
No ratings yet
Python Question Bank With Answers - End Semester (By Ur Engineering Friend)
111 pages
A Desertion: by Stephen Crane
No ratings yet
A Desertion: by Stephen Crane
4 pages
2 Exer chap6 probability
No ratings yet
2 Exer chap6 probability
2 pages
Practical and Symbolic Geometry in Borromini's San Carlo Alle Quattro Fontane
No ratings yet
Practical and Symbolic Geometry in Borromini's San Carlo Alle Quattro Fontane
30 pages
Rani Mam Practice Book 2024 Edition
No ratings yet
Rani Mam Practice Book 2024 Edition
123 pages
BATE
No ratings yet
BATE
5 pages
2m Tanker Cal. Solved
100% (1)
2m Tanker Cal. Solved
21 pages
Soft Tissue Augmentation Techniques in Implants Placed and Provisionalized Immediately. A Systematic Review.
No ratings yet
Soft Tissue Augmentation Techniques in Implants Placed and Provisionalized Immediately. A Systematic Review.
13 pages
CCMT 2024
No ratings yet
CCMT 2024
5 pages
Punch
No ratings yet
Punch
14 pages
Antidiabetic Potential of Plants Used in Cuba
No ratings yet
Antidiabetic Potential of Plants Used in Cuba
11 pages
PV A Directory Jan Revision
No ratings yet
PV A Directory Jan Revision
131 pages
How To Say Nothing in Five Hundred Words
No ratings yet
How To Say Nothing in Five Hundred Words
3 pages
6th Mscshealth and Diet Notes
No ratings yet
6th Mscshealth and Diet Notes
4 pages