Study Paper

On

Artificial Intelligence (AI)

and Big Data for Telecom

FN Division, TEC
K.L. Bhawan, Janpath, New Delhi-01

March 2019
 Table of Contents
Page No.
1. Introduction …………………………………………………………………...…… 4
2. Trends in Communication Network and Services ………………………. …..……. 4
2.1 Characterized requirements …………………………………………………………...…... 4
2.2 Multimedia services ….......................................................................................................... 5
2.3 Precision management ………………………………………………………………....….. 5
2.4 Predictable future ……………………………………………………………………….….. 5
2.5 Intellectualization ……………………………………………………………………….….. 5
2.6 More attention to security and safety ………………………………………….……...……. 6
2.7 Trends of mobile network ……………………………………………………………...…... 6
2.8 Big data for development and ICT monitoring ………………………………….….….…… 6

3. Advantages on AI ….................................................................................................... 7
3.1 Abilities of learning ……………………………………………………….…………….….. 7
3.2 Abilities of understanding and reasoning ………………………………….…………...…... 7
3.3 Ability of collaborating …………………………………………………………….…….… 7

4. AI: Progress & Possible uses in Communications ………………….………………. 8

4.1 AI in SDN ……………………………………………………………….……………….…. 8
4.2 AI in NFV ……………………………………………………………….…………….……. 9
4.3 AI and Network monitor, security and reliability ……………………….………….….….... 10
4.4 AI and QoS ……………………………………………………………….…………….…... 12
4.5 ITU Work/ Focus Group on Machine Learning 5G …………………….………………...… 12

5. AI Challenges and Opportunities …............................................................................ 13

6. Ethical use of AI ………………………………………………………….…….….... 14
7. Conclusion ……………………………………………………………….….………. 15
Abbreviation …………………………………………………………………...……. 16
References ………………………………………………………….......…….……… 17

Figure 1: AI use cases ……………………………………………………….………. 8

Figure 2: Security architecture for AI with business decision making ……………………... 11

2
 Abstract

Artificial Intelligence (AI) is the ability of machines and systems to acquire and apply
knowledge and carry out intelligent behavior. AI with learning abilities is a revolutionary
technology, which the communication industry is exploring, with the aim of introducing it into
communication networks to provide new services/ applications and to improve network
efficiency and user experience. The telecom industry has been a fertile field of application for
AI. The development opportunities created by AI technologies will certainly benefit the entire
industry as well as users.
The paper discusses various aspects of AI, its progress and its application to the telecom
sector including possible use cases. Explains briefly role of AI in Network monitoring, security
and reliability. Also examines the work at ITU on AI such as Focus Group on Machine
Learning 5G etc. Finally, identifies possible challenges in AI implementation and need of
following certain ethics/ ethical values in doing so.

Disclaimer: The outcomes/conclusions drawn and recommendations made thereof in this study paper
are of academic interest only and view of the writers only and should in no case be
considered as an official stand or formal view of TEC.

3
1. Introduction
Artificial Intelligence (AI) is the ability of machines and systems to acquire and apply
knowledge and carry out intelligent behavior. AI can be viewed as a set of associated
technologies and techniques that can be used to complement traditional approaches, human
intelligence and analytics and/or other techniques. In recent years, with the development and
enrichment of technologies such as cloud computing, big data and deep learning, the
industrialization of artificial intelligence is being developed accordingly. With new
developments and worldwide excitement, AI has attracted more and more attention for
introducing it into a number of areas.

Communications sector with a variety of consumer demands on individualization

requirements, multimedia services and precision management, its network security has become
more and more important. With AI’s advantages in learning, understanding, reasoning and
cooperating and with the invent of software-defined networks (SDN) and network functions
virtualization (NFV), technologies of deep packet inspection and service aware networks, the
intellectualization of communication networks and services has become possible. Furthermore,
operators/industry have a keen interest in AI, which may decrease capital expenditure (CAPEX),
and operating expense (OPEX).

2. Trends in Communication Networks and Services:

Artificial Intelligence with learning abilities is a revolutionary technology, which the
communication industry is exploring, with the aim of introducing it into communication
networks to provide new services, to improve network efficiency and user experience.
Communication is a sector with heavy ICT use, dealing with a variety of consumer demands
on individualization requirements, multimedia services and precision management. Current
major trends in communication networks and services are described below: -

2.1 Characterized requirements:

With the increasing number of users and the expanding size of the communication networks
and services, habits, preferences differences and the information needs of individual users and
enterprises are gradually exposed. The customized networks and services now being provided
for enterprise users, which makes the demand for specialized businesses stronger. In near future
special service package for each user, and even a special network may be prioritizing. Such
complex requirements would be unimaginable without an intellectual tool, such as AI.

4
2.2 Multimedia services
Internet users, now have also become information producers, as well as information consumers,
and are producing more and more information in multimedia. User-generated content increases
Internet traffic exponentially. Under these circumstances, both storage and transmission of
data/ information are a great challenge. The inclusion of AI can bolsters the abilities to handle
these challenges.

2.3 Precision management

The use of smartphones makes it inevitable that the various dimensions and granularities in
today's wireless traffic system should be considered in the networks. With the development of
the technologies of network function virtualization (NFV) and software-defined networks
(SDN), the management of the network has become more precise. Virtualization is not only at
the level of network elements, but also at the level of components such as the CPU, memory,
port, bandwidth, etc. AI-based technologies allow operators to set up such on-demand networks
for special users.

2.4 Predictable future

The increasing numbers of users and expansion of business requirements has meant that the
gap between the peaks and troughs of network usage is becoming greater. In this case, operators
are required to predict the future status of networks more accurately to satisfy users’ demand
and improve their experience. AI-driven predictive analytics are helping telecoms provide
better services by utilizing data, sophisticated algorithms and machine learning techniques to
predict future results based on historical data.

2.5 Intellectualization
Networks are becoming more heterogeneous as users often uses a variety of equipment with
different wireless access technologies such as 2G, 3G, 4G, and Wi-Fi. Adoption of 5G will
further reshape telecommunication networks in the near future. The network management
becoming more difficult to maintain with an acceptable quality of service (QoS) due to the
increase in network equipment and user terminals, the expansion of network size, the increase
in the number of users, and the increasing complexity of the network. As well as expanding
capacity by introducing more equipment, the operators are expected to raise their network
performance with smart tools and intelligence technologies. This includes introducing more
intelligence into networks and management to meet customer needs, make more profits, reduce
operating costs, and improve network performance.

5
2.6 More attention to security and safety
As AI develops, security and security will be significant factors for everyone involved in these
technologies. Security incidents are growing and becoming more severe. These events have
resulted in significant commercial consequences, including broken networks, economic losses,
etc. AI can be used to establish strong security protection and behavioral analysis based on
machine learning, will significantly improve the ability of network detection attacks, automatic
analysis of data, and the identification of relationships between isolated behaviors.

2.7 Trends of mobile network

Today, the application of virtualized network functions (NFV) to mobile core networks is in
progress. For years, various network functions, such as the conventional Evolved Packet Core
(EPC), have been provided in their dedicated hardware (HW) such as Advanced Telecom
Computing Architecture (ATCA) hardware. With the introduction of NFV, software will be
able to run on a virtualized operating system (OS) of generic Intel architecture (IA) servers and
be provided separately from hardware. Furthermore, the NFV architecture enables integrated
management and control (orchestration) of network services and resources, interworking with
Management and Orchestration. The application of AI will enable to respond to the above-
mentioned problems in the planning and maintenance processes quickly and efficiently.

2.8 Big data for development and ICT monitoring

Big Data and AI are set of two amazing modern technologies that empower machine learning,
continuously reiterate and update the data banks, and taking the help of human intervention
and recursive experiments for the same. Data combined with steady advances in the power of
computing are leading to the emergence of data-driven innovation: online activity and
networked things generate “big data” which feed machine learning that enables AI, which in
turn leads to advances in intelligent machines (robotics, automated vehicles) as well as new
techniques in science which can spur further innovation. The growth of the volume, variety
and velocity of data and the ability to analyze and use it is a significant departure from the past
and marks the emergence of a new factor of production that augments traditional capital and
labor, but with unique properties of its own.
One of the richest sources of big data is the data captured by the use of ICTs. This
broadly includes data captured directly by telecommunication operators as well as by Internet
companies and by content providers such as Google, Facebook, Twitter, etc. Big data from the
ICT services industry are already helping to produce large-scale development insights of
relevance to public policy. Collectively, they can provide rich and potentially real-time insights
6
 to a host of policy domains. In some countries and regions, the use of big data, including big
data from the ICT industry, is subject to national regulation.

3. Advantages of Artificial Intelligence:

In the communications industry, whether its network operators, equipment manufacturers or
solution providers, the industry hopes to take advantage of AI to assist in areas in which they
are currently struggling, such as in designing, operating, maintaining and managing
communication networks and services. Some of the advantages of AI are as follows:-

3.1 Abilities of learning

Operators need intelligent decisions to manage complex resources and dynamic traffic. With
capability to describe the network traffic characteristics accurately, AI has entered into the
cognitive age, and deep learning can be used, where the machine system can use the existing
training data to process large amounts of data through data mining. AI can also learn the
characteristics of data traffic, management, controls and other characteristics automatically and
master expert experience of operating, managing and maintaining networks. By these efforts,
the accuracy of analysis can be enhanced, and the intelligent management and services of
communication networks can be realized.

3.2 Abilities of understanding and reasoning

Due to the dynamics of the network system, the state information of a resource may have
changed when it is transmitted to the network management system. Therefore, the network
management can only know the local state information without the knowledge about the system
internal state. Machine learning happens to have the strength to deal with this kind of logic and
uncertainty reasoning. In order to make the classification or prediction easier, deep learning
constructs a multi-hidden layer model and uses the hierarchical network structure to transform
the feature representation of the sample into a new feature space layer by layer.

3.3 Ability of collaborating

Due to the expansion of the network both in scale and size, the structure complexity of
communication networks are increasing quickly. Concepts such as distribution and hierarchy
are often talked about in the network management. Management tasks and controls are
distributed to the entire network. As a result, we have to deal with issues such as tasks’
distribution, communication and collaboration between management nodes. If we introduce

7
 the multi-agent collaboration of distributed AI into the network management, we can expect
the ability to collaborate between network managers distributed in every layer.

4. Artificial Intelligence: Progress & possible uses in Communications

Innovative telecom operators use AI and machine learning to increase network reliability,
improve customer satisfaction & retention, optimize their business processes for higher profit,
and much more. The most visible use case for AI in telecoms is enhanced customer service.
Other than customer service, telecoms use artificial intelligence in network maintenance. Top
AI use cases are shown in Figure 1.
SDN, NFV, network slicing and other technologies, coupled with integrated network
management systems have been able to directly issue orders which can be executed by network
equipment, and DPI systems can be deployed on network equipment, and it is possible to
realize real-time monitoring of networks and services and intelligent management.

Figure 1: AI use cases

4.1 AI in SDN
Software defined networking (SDN) represents a promising networking architecture that
combines central management and network programmability. SDN separates the control plane

8
 from the data plane and moves the network management to a central point, called the controller
that can be programmed and used as the brain of the network, which greatly promote the
capabilities of network-automated management and control. A typical SDN framework is
composed of three layers: infrastructure layer, control layer and application layer. The
infrastructure layer includes some network elements which can provide network traffic, acting
as the object controlled by the SDN controller, as well as a data source of the network resource.
The control layer has the SDN controller, which is the core component of the SDN network
carrying out important tasks of controlling network traffic. The application layer includes
various applications. The southbound interface D-CPI (Data-Controller Plane Interface) is
responsible for exchanging data between the SDN controller and the network element. The
northbound interface A-CPI (Application-Controller Plane Interface) is responsible for
providing the upper-level application with the channel exchange to obtain the underlying
network resource information and send data to the lower-level network. SDN provides a good
interface with its programmability to introduce AI into the communication networks. This is
SDN's biggest advantage. SDN uses the application-programming interface to send powerful
programming instructions to the network device.
With AI, network managers can not only schedule an automated intelligent business
orchestrator, but also program the AI-optimized network strategy and automatically compile
them into the task script, then assign them into the network allocation tasks with the
application-programming interface (API). Network managers can also automatically collect
network statistics information to lay a solid foundation for continuous network optimization.
If necessary, some new functionalities can also be added intelligently through the SDN
application for the network environment.

4.2 AI in NFV
With virtualization technology, Network Functions Virtualization (NFV) can divide network-
level functions and applications, such as routing, customer premises equipment (CPE), mobile
core, IP Multimedia Subsystems (IMS), content delivery networks (CDN), switching elements,
mobile network nodes, home routing operations, set-top box business, tunnel gateway
elements, traffic analysis, service assurance, service level agreement (SLA) monitoring, testing
and diagnosis, next generation network (NGN) signal, aggregation and network range
functions, application optimization, security policy, etc., into several functional blocks, and run
them in software mode respectively. This means that they are no longer limited to the hardware
architecture.

9
 The typical NFV reference architecture includes three layers of the complete
infrastructure layer, the resource management layer, and the business flow orchestrator layer.
NFV helps ISV and telecommunication operators to achieve virtual network functions by
deploying hypervisor at the infrastructure layer to virtualize infrastructure resources such as
commercial general computing, storage and network resources and others. The resource
management layer is in charge of the NFV infrastructure’s management, configuration and
collaboration. The business flow orchestrator layer is a key part of the NFV network function
for network operating; it is used to organize and orchestrate the functions of the NFV network.
It is also in charge of managing and monitoring the global resources across the data center or
the resource pool.
With the virtualization of network functions NFV can realize an on-demand dynamic
network configuration separated from the underlying architecture. As key issues have been
solved, AI can play its full role in critical network management.

4.3 AI and Network monitoring, security and reliability:

To master the real-time information of the communication network, the network must have the
function of initiative uploading. Currently there are many DPI systems. With inspectors, the
deep packet inspection (DPI) system can collect the information such as the running state of
network equipment, the usage of resources and the quality of services. With the big data
obtained from the DPI system, the AI system can rapidly analyse and find if there are or will
be abnormity within the information. For example, if the AI system finds a burst a continuous
traffic, it can doubt a distributed denial of service (DDoS) attack in the network and analyse
the package characteristics immediately, then orchestrate an inspector collaboration task to
drop all packages with the characteristics to avoid the damage. It could write a new record in
the security database in case of the appearance of unknown hack attacks or new virus flooding.
While developing AI systems, attention must be paid to the potential security risks;
strengthen prevention mechanisms and constraint conditions; minimize risks; and ensure AI’s
secure, reliable, and controllable development. When applying AI models, it needs to analyze
and determine the risks in using AI technology based on the characteristics and architecture of
specific services, and design a robust AI security architecture and deployment solution using
security mechanisms involving isolation, detection, failsafe, and redundancy (Figure 2).
By detecting unusual network activity at every level of the network, an AI-enabled platform
can accurately detect existing and day-zero threats. In addition, location technology can be

10
used to accurately locate accidental or malicious rogue devices and provide location-based
access to resources.

(Source: Huawei)
Figure 2 Security architecture for AI with business decision making

i) Isolation:
To ensure stable operation, an AI system analyzes and identifies the optimal solution and sends
it to the control system for verification and implementation. Generally, the security architecture
must isolate functional modules and setup access control mechanisms between modules. The
isolation of AI models can reduce the attack surface for AI inference, while the isolation of the
integrated decision module can reduce attacks on the decision module. The output of AI
inference can be imported into the integrated decision module as an auxiliary decision-making
suggestion, and only authorized suggestions can enter the decision module.

ii) Detection:
Adopting continuous monitoring and with an attack-detection model in the main system
architecture, it is possible to comprehensively analyze the network security status and estimate
the current risk level. When the risk is high, the integrated decision system can reject the
suggestion coming from the automatic system and hand over control to a person to ensure
security under attacks.

iii) Failsafe:
When a system needs to conduct critical operations such as AI-assisted autonomous driving or
medical surgery, a multi-level security architecture is required to ensure the entire system

11
 security. The certainty of the inference results provided by the AI system must be analyzed.
When the certainty of the result is lower than a certain threshold, the system falls back to
conventional rule-based technologies or manual processing.

iv)Redundancy:
Many business decisions and data are associated with each other. A feasible method to ensure
the security of AI models is to analyze whether the association has been ruined. A multi-model
architecture can be set up for critical applications, so that a mistake in one model does not keep
the system from reaching a valid decision. In addition, the multi-model architecture can largely
reduce the possibility of the system being fully compromised by a single attack, thereby
improving the robustness of the entire system.

4.4 AI and QoS:

The AI module consists of two different parts. The first one is a classifying part, which detects
the type of traffic that is sent through the network. The second part is an estimator that informs
the SDN controller on which kind of action should be executed to guarantee the Quality of
Service (QoS) and Quality of Experience (QoE). Results show that with the actions performed
by the network, problems like jitter and losses can be reduced.

4.5 ITU Work/ Focus Group on Machine Learning 5G:

ITU-T under its Study Group 13 has established a Focus Group on Machine Learning for Future
Networks including 5G, to draft technical reports and specifications for machine learning (ML)
for future networks, including interfaces, network architectures, protocols, algorithms and data
formats. The major objectives of FG-ML5G includes:
 To help adoption of ML in future networks including architecture, interfaces, use cases,
protocols, algorithms, data formats, interoperability, performance, evaluation, security
and protection of personal information;
 To study, review and survey existing technologies, platforms, guidelines and standards
for ML in future networks;
 To recognize and highlight the various perspectives for the future of networks and
computing systems involving ML;
 To identify aspects enabling safe and trusted use of ML frameworks;
 To identify possible requirements on network functionality, interfaces and capabilities to
use ML;
 To identify challenges in the standardization activities for ML in communications;

12
  To establish liaisons and relationships with other organizations which could contribute to
the standardization activities for ML.

5. AI Challenges and Opportunities:

With humans and machines joining forces now, more than ever before, AI is no longer confined
to innovation labs and is being hailed for its immense transformational possibilities. However,
businesses including telecom need to overcome certain challenges before they can realize the
true potential of this emerging technology. The key lies in leveraging the right opportunities in
AI.

(i) Provability:
Organizations involved in AI cannot demonstrate clearly, why it does and what it does. No
wonder AI is a “black box” as of now. People are skeptical about it, as they fail to understand
how it makes decisions. Provability – the level of mathematical certainty behind AI predictions
– remains a grey area for organizations. There is no way they can prove or guarantee that the
reasoning behind the AI system’s decision-making is clear. The solution lies in making AI
explainable, provable, and transparent. Organizations must embrace Explainable AI as a best
practice.

(ii) Data privacy and security:

As AI develops, Privacy and security will be significant factors for everyone involved in these
technologies. Most AI applications rely on huge volumes of data to learn and make intelligent
decisions. Machine Learning systems use/ depend heavily on data, often sensitive and personal
in nature, to learn from them and enhance themselves. This makes it vulnerable to serious issues
like data breach and identity theft. Besides, an emerging method – ‘Federated Learning’, which
is a machine learning setting where the goal is to train a high-quality centralized model with
training data distributed over a large number of clients each with unreliable and relatively slow
network connections– is all set to disrupt the AI paradigm. It will empower data scientists to
develop AI without compromising users’ data security and confidentiality.

(iii) Algorithm bias

An inherent problem with AI systems is that they are only as good – or as bad – as the data
they are trained on. Bad data is often laced with racial, gender, communal or ethnic biases.
Proprietary algorithms are used to determine who is called for a job interview, who has granted
bail, or whose loan is sanctioned. If the bias lurking in the algorithms that make vital decisions

13
 goes unrecognized, it could lead to unethical and unfair consequences. For instance, Google
Photos service uses AI to identify people, objects and scenes. However, there is a risk of it
displaying wrong results, such as when a camera missed the mark on racial sensitivity, or when
a software used to predict future criminals showed bias against black people.
In the future, such biases will probably be more accentuated, as many AI systems will
continue to be trained using bad data. Hence, the need of the hour is to train these systems with
unbiased data and develop algorithms that can be easily explained.

(iv) Data scarcity

It is true that organisations have access to more data today than ever before. However, datasets
that are relevant for AI applications to learn are indeed rare. The most powerful AI machines
are the ones that are trained on supervised learning. This training requires labeled data – data
that is organised to make it ingestible for machines to learn. Labeled data is limited. In the not-
so-distant future, the automated creation of increasingly complex algorithms, largely driven by
deep learning, will only aggravate the problem. There is a ray of hope though. As a trend that
is fast catching up, organisations are investing in design methodologies, trying to figure out
how to make AI models learn despite the scarcity of labeled data. ‘Transfer learning’,
‘Unsupervised/Semi-Supervised Learning’, ‘Active Learning’, and so on are just a few
examples of the next-generation AI algorithms that can help resolve this.

6. Ethical use of AI
Artificial Intelligence is seen as a great transformative tech and the possibilities seem almost
limitless to what it can eventually do. With these disruptive developments, questions arises
about the functional capabilities to the ethics behind creating such powerful and potentially
life-consequential technologies. As such, it makes sense to spend time considering what we
want these systems to do and make sure we address ethical questions now so that we build
these systems with the common good of humanity in mind. Following 9 ethical issues have
been identified at World Economic Forum–ASEAN Summit 2018 titled “Top 9 ethical issues
in artificial intelligence: -
(i) Unemployment- what happens after the end of jobs
(ii) Inequality -how do we distribute the wealth created by machines
(iii) Humanity –how do machines affect our behavior and interaction
(iv) Artificial stupidity – how can we guard against mistakes
(v) Racist robots-how do we eliminate AI bias

14
 (vi) Security –how do we keep AI safe from adversaries
(vii) Evil genies-how do we protect against unintended consequences
(viii) Singularity –how do we stay in control of a complex intelligent system
(ix) Robot rights - how do we define the humane treatment of AI

7. Conclusion:
AI and Big Data are two of the emerging technologies that are used in Telecommunications
sector extensively, helping CSPs manage, optimize and maintain not only their infrastructure,
but their customer support operations as well. This paper has highlighted an AI-based network
framework to introducing AI in communication networks and services, with SDN/ NFV
collaboratively deployment. Technology is already a core part of the telecommunications
industry, and as Big Data tools and applications become more available and sophisticated, AI
can be expected to continue to grow in this space.

15
Abbreviations:

AI: Artificial Intelligence

SDN: Software Defined Networks
NFV: Network Function Virtualization
QoS: Quality of Service
QoE: Quality of Experience
IMS: IP Multimedia Subsystems
ICT: Information & Communication Technologies
DPI: Deep Packet Inspection
DDoS: Distributed Denial of Service (DDoS)
CDN: Content Delivery Networks
NGN: Next Generation Network (NGN)
EPC: Evolved Packet Core
OS: Operating system
FG-ML5G: Focus Group- Machine Learning 5G

16
 References:

[1] http://www.oecd.org
[2] https://www.itu.int/en/journal/001/Documents/ITU2017-4.pdf
[3] https://www.itu.int/en/journal/001/Documents/itu2017-7.pdf
[4] https://www.weforum.org/agenda/2016/10/top-10-ethical-issues-in-artificial-intelligence/
[5] http://www.forbesindia.com/blog/business-strategy/artificial-intelligence-key-challenges-and-
opportunities/
[6] https://www-file.huawei.com/-/media/corporate/pdf/cyber-security/ai-security-white-paper-
en.pdf?la=en&source=corp_comm
[7] https://www.itu.int/en/ITU-T/focusgroups/ml5g/Pages/default.aspx

*********

17