See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/350824466

AI and 6G Security: Opportunities and Challenges

Conference Paper · June 2021

DOI: 10.1109/EuCNC/6GSummit51104.2021.9482503

CITATION READS

1 1,726

4 authors:

Yushan Siriwardhana Pawani Porambage

University of Oulu University of Oulu
9 PUBLICATIONS   103 CITATIONS    59 PUBLICATIONS   1,259 CITATIONS   

SEE PROFILE SEE PROFILE

Madhusanka Liyanage Mika Ylianttila

University College Dublin University of Oulu
183 PUBLICATIONS   2,731 CITATIONS    217 PUBLICATIONS   4,973 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

5G-CHAMPION : 5G Communication with a Heterogeneous, Agile Mobile network in the Pyeongchang wInter Olympic competitioN View project

Blockchain Technology View project

All content following this page was uploaded by Madhusanka Liyanage on 21 April 2021.

The user has requested enhancement of the downloaded file.

 AI and 6G Security: Opportunities and Challenges
Yushan Siriwardhana∗ , Pawani Porambage† , Madhusanka Liyanage‡ , Mika Ylianttila§
∗†‡§ Centre for Wireless Communications, University of Oulu, Finland
‡ Schoolof Computer Science, University College Dublin, Ireland
Email: ∗†‡§ [firstname.lastname]@oulu.fi, ‡ [email protected]

Abstract—While 5G is well-known for network cloudification “sub-networks” such as in-body networks will emerge, which
with micro-service based architecture, the next generation net- generally operate in a standalone fashion while benefiting from
works or the 6G era is closely coupled with intelligent network the wide area network.
orchestration and management. Hence, the role of Artificial
Intelligence (AI) is immense in the envisioned 6G paradigm. 6G will take the network softwarization/cloudification into
However, the alliance between 6G and AI may also be a double- network intelligentization, revolutionizing wireless networks
edged sword in many cases as AI’s applicability for protecting from connected things to “connected intelligence” [1], [2].
or infringing security and privacy. In particular, the end-to- Hence, AI becomes an integral part of the network, which
end automation of future networks demands proactive threats
plays a crucial role. The distributed heterogeneous networks
discovery, application of mitigation intelligent techniques and
making sure the achievement of self-sustaining networks in 6G. require ubiquitous AI services to ensure the fulfilment of
Therefore, to consolidate and solidify the role of AI in securing 6G goals. Intelligent wireless communications, closed-loop
6G networks, this article presents how AI can be leveraged in optimization of networks, big data analytics for 6G emphasize
6G security, possible challenges and solutions. the use of AI in diverse aspects of 6G networks.
Index Terms—6G, 6G Security, Artificial Intelligence, Machine
Learning, Intelligent Security
Beyond 2030 wireless applications will demand much
higher data rates (up to 1 Tb/s), extremely low end-to-
I. I NTRODUCTION TO 6G end latency (< 1 ms), extremely high end-to-end reliabil-
ity (99.99999%) [1], [2]. Moreover, 6G networks will comprise
As the realization of digital transformation expected with a collection of heterogeneous dense networks embedded with
5G networks has already begun and continue to evolve over connected intelligence and utilize hyper-connected cloudifica-
this decade, the 6G communication era envisions how hu- tion. Service provision for extreme requirements with complex
mans will interact with the digital virtual worlds beyond 6G networks requires sophisticated security mechanisms. The
2030. Future networks must possess novel technologies that security systems designed for 5G using the concepts of SDN
enable the digital virtual worlds with connected intelligence, and NFV should be further improved to cater to the security
to address the communication and networking challenges demands in 6G [5]. The end-to-end automation of future
beyond 2030. While the conventional applications such as networks demands proactive threats discovery, intelligent mit-
multimedia streaming will remain, literature envisions new igation techniques, and self-sustaining networks in 6G. Hence,
application domains for 6G systems such as multisensory the end-to-end security design leveraging AI techniques is
extended reality (XR) applications, Connected Robotics and essential to autonomously identify and respond to potential
Autonomous Systems (CRAS), and wireless Brain-Computer threats based on network anomalies rather than cryptographic
Interactions (BCI) [1]. Holographic telepresence, eHealth in- methods. This paper discusses the role of AI in the security
cluding in-body networks are a few other 6G use cases [2] provision of 6G networks.
that demand extremely high data rates, ultra-low latency and The remainder of the paper is organized as follows: Section
ultra-reliability. II outlines the security threat landscape in 6G. Section III
The evolution of 6G application domains calls for an discusses the role of AI in 6G security and privacy. Section IV
innovative network architecture beyond current network de- highlights the issues and possible solutions. Finally, Section V
signs [3]. An open and distributed reference framework for concludes the paper.
6G architectural building blocks defined by Nokia Bell Labs
comprises four major interworking components [4]. These are II. S ECURITY T HREAT L ANDSCAPE OF 6G
platform, functional, specialized, and orchestration, covering
the physical layer to the service layer with the following distin- This section presents the security issues of 6G networks.
guishing features. The “het-cloud” is a heterogeneous cloud Pre-6G security explains the security issues inherited from
environment that eases the creation, placement and scaling 5G to 6G. The section also discusses security issues resulting
of dynamic cloud services. 5G core network service-based from architectural changes and novel technologies used in 6G.
architecture will extend to the Radio Access Network (RAN), Figure 1 illustrates possible attacks on different layers of the
named “RAN-core convergence”. It will harmonize the RAN 6G architecture. Figure 2 depicts a summary of how this paper
and core functions to create simpler networks [4]. Even smaller presents the security threat landscape of 6G.
 Intelligent
AI-enabled Functions Smart City Transportaion Smart Industry Smart Health

Application Layer
Smart
Control Layer

Parameter Resource Centered Management Distributed

SDN
Intelligent

Optimization Management Cloud Center Cloud

Task Policy
Scheduling Learning
Intelligent Edge

Dimension Abnormal Storage Data Compute

Reduction Data Filtering Center Server
Layer

at Edge
Knowledge Feature
Discovery Extraction
Sensing Layer

Data Statue
Intelligent

Collection Detection

Environment
Measurement
Monitoring

Fig. 1: Intelligent 6G Architecture [6], and 6G Security and Privacy Issues.

A. Pre-6G Security The local 5G network deployments usually provide services

Network softwarization technologies in 5G such as for verticals such as industries, healthcare, education. 6G
Software-Defined Networking (SDN), Network Function Vir- further expands the concept by allowing even smaller networks
tualization (NFV), Multi-access Edge Computing (MEC), and like in-body networks, swarms of drones, environmental sen-
network slicing are still applicable for 6G systems; thus, their sor networks with longer battery life [4]. These local networks
security issues would remain in 6G. SDN related prominent operate as a standalone network and interwork with wide area
security issues are attacks on SDN controller, attacks on connectivity when needed. In contrast to the well-defined local
northbound and southbound interfaces, inherent vulnerabilities 5G networks, many stakeholders implement local 6G networks
of platforms used to deploy SDN controllers/applications [7]. with different embedded security levels. A local 6G network
NFV related security issues are attacks targeting Virtual Ma- with minimal security provides intruders with the opportunity
chines (VM), Virtual Network Functions (VNF), hypervisor, to initiate attacks. And then infiltrate the networks which trust
VNF manager, NFV orchestrator[8]. Due to the massively dis- the compromised network.
tributed nature of 6G systems, MEC in 6G is subjected to phys-
ical security threats, Distributed Denial of Service (DDoS), B. Security of 6G Architecture
man-in-the-middle attacks[9]. Potential attacks for network 6G cells will shrink from small cells to “tiny cells” [1], and
slicing are DoS attacks, information theft via compromised denser deployment of cells, mesh networks, multi-connectivity,
slices[10]. Attacks on network softwarization technologies fail and Device-to-Device (D2D) communications will be a norm.
the 6G network from achieving the promised dynamicity and Malicious parties have a better potential to attack a dis-
full automation. tributed network with more vulnerable devices, each having
6G envisions the realization of the Internet of Every- mesh connectivity, thereby increasing the threat surface. The
thing (IoE), a collection of billions of heterogeneous devices. definition of sub-networks requires a change in the security
The fundamental device security model relying on SIM cards strategy. Security provision for the massive number of de-
is not a practical deployment for IoE in 6G, especially with vices within each sub-network by the wide area network is
the small form factor devices such as in-body sensors. Key far from practical. A hierarchical security mechanism that
distribution and management functions are highly inefficient distinguishes the sub-network level communication security
in such a massive network [11]. The resource-constrained IoT and sub-network to wide area network security would be a
devices cannot afford complicated cryptography to maintain better approach in 6G. The RAN-Core convergence makes
strong security [12], making them a primary target of the higher layer RAN functions more centralized and coexist
attackers. These devices can be compromised and potentially with the distributed core functions as User Plane Micro Ser-
used to initiate attacks. Data collection by hyper-connected vices (UPMS) and Control Plane Micro Service (CPMS) [4],
IoE to serve 6G applications raises privacy issues. Data theft possibly at the edge. Attackers can target UPMS and CPMS,
by exploiting resource-constrained IoT devices will affect data affecting multiple radio units served by micro-services. 6G
privacy, location privacy, and identity privacy. networks will coexist with frameworks like Zero-touch net-
work and Service Management (ZSM) architecture [13] to III. AI FOR S ECURITY AND P RIVACY IN 6G
enable short time-to-market of services, low operational cost, Intelligent security and privacy provision is a part of AI’s
and reduced human error. Full automation equipped with self- role in 6G systems. This section presents AI’s use in pre-
learning causes the attacks to propagate in the closed-loops. 6G security, security of 6G architectures, security of 6G
Data privacy protection is challenging due to the vital need technologies, and AI for 6G privacy. Figure 3 illustrates AI’s
for automation with lesser human intervention in zero-touch role as a defender for 6G applications.
networks. Implementing automated machine ethics is also an
open question in fully automated 6G networks. A. Use of AI to Identify/Mitigate Pre-6G Security Issues
Multilayered intrusion detection and prevention using deep
C. Security of 6G Technologies
reinforcement learning and Deep Neural Networks (DNN) is
6G rely on AI to enable fully autonomous networks. viable in SDN/NFV-enabled networks [18]. They effectively
Therefore, attacks on AI systems, especially Machine Learn- defend against IP spoofing attack, flow table overloading
ing (ML) systems, will affect 6G. Poisoning attacks, data attack, DDoS attack, control plane saturation attack, and host
injection, data manipulation, logic corruption, model evasion, location hijacking attack compared to several conventional
model inversion, model extraction, and membership inference approaches. ML approaches such as Decision Trees and
attacks are potential security threats against ML systems [14]. Random Forest, are also proved useful for detecting DDoS
The collection of more features allows AI systems to perform attacks in SDN environments due to their short processing time
better. Attacks on collected data, and the unintended use of and accuracy, respectively [19]. ML-based adaptive security
private data, lead to privacy issues as the data processing approaches are effective against attacks on SDN/NFV as the
is usually not visible to the users. Blockchain is also a 6G networks expect dynamic placement of virtual functions
pivotal technology to unleash the potential of 6G systems. on-demand. The attacks also evolve using AI techniques to
Blockchain is suitable for decentralized resource management, learn vulnerabilities in a vastly distributed network. Hence,
spectrum sharing, service management in massively large rule-based detection systems are ineffective.
and distributed 6G networks [15]. 51% attacks are feasible On-device resource limitations, the difficulty of key man-
with quantum computers to destabilize the blockchain. As agement in massive scale heterogeneous networks, the vast
the blockchain networks stores data publicly, privacy preser- amount of device data generation make the conventional
vation is challenging. The current 5G standard does not authentication/authorization systems insufficient for adequate
concern the security issues due to quantum computing; rather, security provision in large-scale IoT. Anomaly-based intrusion
it depends on traditional cryptography like Elliptic Curve detection systems in Industrial IoT (IIoT) [20] detect mali-
Cryptography (ECC) [5]. The present security mechanisms cious packets based on their behaviour. These learning-based
based on asymmetric key cryptography are vulnerable against detection systems utilize various features of the data as the
quantum computer based attacks as the 6G era will mark the input, therefore, suitable for detecting zero-day attacks. The
presence of quantum computers [16]. Thus, the secure 5G use of communication link attributes and user behaviours with
communications enabled with asymmetric key cryptography machine learning for authentication and authorization [21],
may be no longer useful for post-quantum security without [22] is a better approach in future networks for resource-
the design of quantum-safe cryptographic algorithms. Visible constrained devices. In that way, devices do not utilize their
Light Communication (VLC) is also a technology suitable for limited resources to provide additional complex security.
indoor based systems such as positioning systems and outdoor The sub-networks in 6G, which can be considered an
systems such as vehicle-to-vehicle communication. Common expansion of local 5G networks beyond vertical domains,
attacks against VLC systems such as eavesdropping, jamming, can benefit from learning-based security techniques within the
node compromise systems [17] prevents the safe use of VLC. sub-network and between different sub-networks. ML-based
algorithms deployed at the perimeter can capture the behaviour
Pre-6G Security Issues of other sub-networks and detect malicious traffic from those
¨Issues of 5G technologies inherited to 6G
¨Security and Privacy Issues of IoE
sub-networks. Massive data transfer from one sub-networks to
¨Security Issues of Local 6G Networks another might be of no use as these networks usually operate
standalone. A sub-network can share only the learned security
Security Threats of 6G Architecture
¨Tiny cells, Mesh Networks, and Multi-Connectivity intelligence to another for communication efficiency [23]. A
Analysis of 6G ¨Sub-networks second sub-network can use the shared intelligence, feed it
Security Threat ¨RAN-Core Convergence
Landscape ¨Zero-touch networks into its ML models, determine the malicious traffic of other
¨Privacy Issues networks, and apply dynamic policies.

Security Threats of 6G Technologies B. Use of AI to Mitigate Security Issues of 6G Architecture

¨AI/ML
¨Blockchain 6G will primarily depend on edge intelligence compared to
¨Quantum Computing present centralized cloud-based AI systems. The distributed
¨Visible Light Communication
¨Privacy Issues nature enables the execution of edge-based federated learn-
ing for network security in the massive device and data
Fig. 2: Analysis of 6G Security Threat Landscape.
regime [24], ensuring communication efficiency. 6G architec- to feed intelligent models, users would prefer different pri-
ture envisages connected intelligence and uses AI at different vacy levels on different applications. AI-based service-oriented
levels in network hierarchy [6]. AI at the tiny cell level has the privacy-preserving policy updates [29] is a potential solution to
potential to block DoS attacks on cloud servers at the lowest support fully automated 6G networks with preserved privacy.
level. The Multi-connectivity of a device in a mesh network
IV. I SSUES AND C OUNTER - MEASURES
allows several base stations to evaluate the behavior of a device
using AI classification algorithms, and collectively decide the Despite the importance of AI in 6G systems, AI has its
authenticity using weighted average schemes as in [25]. The security, privacy, and ethical issues. Moreover, AI can be an
behaviour based approaches reduce the overhead caused by instrument to launch intelligent attacks. This section presents
frequent key exchanges as the tiny cells and multiple access the privacy, security, ethical issues in AI and potential solu-
technologies cause frequent handovers. Different levels of tions to overcome them. Figure 3 also illustrates AI’s role as
authorization are possible for sub-network level and wide area a defender for 6G applications.
network level with federated learning. The trust score learned A. Security Issues in AI
within the sub-network level can be shared outside only when 1) Issues: 6G achieves connected intelligence via AI-
external communication is needed. Learning based intrusion enabled functions, especially with ML systems that are sub-
detection approaches [18], [19] can be good candidates to jected to security threats. Poisoning attacks influence the
prevent attacks on CPMS and UPMS as the edge already learning phase of a ML system, which lead the model to learn
possess the data for intelligent service provision. Frameworks inaccurately. For example, data injection, data manipulation,
like ZSM are equipped with domain analytics and domain and logic corruption are some of the poisoning attacks. Eva-
intelligence services for zero-touch management of networks, sion attacks try to avoid the model during the inference phase
predominantly based on AI. AI model assessment, AI engine using carefully crafted adversarial examples. Model extraction,
for API security are key security functional components that model inversion, and membership inference are API based
enhance the security of ZSM reference architecture [26]. attacks on ML models [14].
2) Solutions: Potential countermeasures such as adversar-
C. Use of AI to Mitigate Security Issues of 6G Technologies ial machine learning and moving target defence can create
The predictive analytics using AI can predict attacks such resilient AI systems. Input validation and robust learning
as 51% attacks on blockchain before the attack occurs. A against poisoning attacks, adversarial training and defensive
quantum computer may threaten asymmetric key cryptogra- distillation against evasion attacks, and differential privacy
phy. However, they can provide exponential speed-ups for and homomorphic encryption against API based attacks are
AI/ML algorithms to perform tasks much faster and realize other defence mechanisms. The balance between the increased
previously impossible tasks. Hence, quantum machine learning defence and performance degradation is a design challenge
for network security is a potential defence technique against with these defence mechanisms [14].
quantum computer based attacks [27]. Intelligent beamforming B. Privacy Issues in AI
techniques based on Reinforcement Learning (RL) provide op-
1) Issues: Due to AI’s ability of large scale data analysis
timal beamforming policy against the eavesdropper attacks in
combined with the speed of future computers and automation
VLC systems [28]. Jamming resembles DoS attacks; therefore,
needs of future networks, AI can easily compromise privacy.
anomaly-based detection systems equipped with AI is a possi-
6G requires a massive amount of user data collected via
ble solution to detect jamming attacks. AI-based authentication
billions of devices, and the users no longer foresee how
and authorization systems are also suitable for preventing node
external systems handle their data. For example, the pro-
compromise attacks [21].
posed intelligent authentication systems depend on physical
attributes [22], may use private user data. Insecure IoT devices
D. Use of AI for 6G Privacy
(ex: low powered sensors) feeding personal data to AI systems,
Multi-connectivity, mesh networks with tiny cells in 6G are a potential target for data theft. Model inversion attacks
allow simultaneous communication for devices via multiple on ML to retrieve the training data could also be a source for
base stations. Edge-based ML models could be used for privacy violation [30].
dynamic detection of privacy-preserving routes, rank them, 2) Solutions: Edge-based federated learning preserves data
and allow devices to transfer data via privacy-preserving routes privacy by imposing a physical control to maintain data closer
based on the ranking. Federated learning [23] keeps data in the to the user [30]. Homomorphic encryption, which allows
user’s proximity compared to cloud-based centralized learning performing mathematical operations without decrypting data,
to enhance data privacy and location privacy. The 6G sub- imposes a technical control for privacy preservation [31]. Fur-
network level AI allows privacy preservation within the sub- ther research on homomorphic encryption is needed to ensure
network and share only the learned intelligence outside to learning with encrypted data produces the same output as with
minimize privacy risks. Confining data within the network is plain data. Differential privacy techniques can add random
suitable for applications like in-body networks. With the vast noise to the training data and prevent private information
number of applications in 6G and the massive data collection disclosure towards learning models [32].
 Adversaries Compromised
Learn using AI Cloud AI Autonomous
3 1 2 Cars
Stolen Data,
Learned AI V2V
1 2
Models Communication
6G Tiny
- Detect Vulnerabilities Edge Cell Centralized Edge AI Attacked
- Predict next Attacks Server Converged - Poisoning Attacks
6G Core RAN-Core - Model Evasion
- Model Extraction
Adversaries 1 2 - Model Inversion
AI based Security Deployed Converged
Attack 6G/AI
Intelligent - Authentication/Authorization
Models RAN-Core
Services - Anomaly Detection with AI
Poisoning Data Cross-Domain Domains
Connectivity Resilient AI Systems Deployed Transport
Extracted AI Models 6G Tiny Automated Network and - Moving Target Defense
Edge
Cell Service Management
- Adversarial ML
- Attacks on Learning Phase Server - Input Validation
- Attacks on Inference Phase - Differential Privacy Converged
- Model's API based Attacks 1 AI in 6G as an Enabler Delivery
RAN-Core 5
Adversaries 2 AI in 6G as a Defender Intelligent Transportation
3 AI as an Offender
4 AI in 6G as a Target Extended Reality (XR) Industry 5.0 Brain Computer
AI in 6G AI for 6G Adversaries AI in 6G Resilient Interactions (BCI)
Operation Security Attacked AI in 6G 5 Resilient AI in 6G Other 6G Applications
using AI

Fig. 3: AI’s role as an Enabler, Defender, Offender and a Target in 6G Intelligent Transportation.

C. Ethical Issues in AI D. Use of AI for Attacking 6G

1) Issues: Fully automated AI-based 6G networks need 1) Issue: With the ability to make network-wide intelligent
lesser human interventions in network operations. The way decisions with distributed edge-based architecture, AI itself
machines learn differs from how humans perceive things, and can uncover the patterns within a large volume of data at differ-
machines do not address ethical considerations as humans. AI ent levels (intelligent radio, edge, and cloud). Hence, AI-based
systems behave according to the way they were taught and mechanisms have the potential to uncover vulnerabilities of the
trained. However, they are not capable of behaving against network. For example, AI can learn the most vulnerable IoT
the logic in exceptional situations as done by humans. devices, convert them into bots and initiate DDoS attacks [14]
2) Solutions: “Ethics by Design” [33] approach brings the against a critical node.
debate on ethical implications at the primary stage of the 2) Solution: The countermeasure for AI based intelligent
design of AI systems. Considering ethics at the early stage attacks is also the implementation of more intelligent defence
could be useful in intelligent 6G systems. Guidelines, laws, systems. They can be empowered by AI itself using distributed
and regulations [34] are possible measures to address data intelligence. Moving target defence techniques is a proactive
ethics and ownership in the context of 6G to achieve a proper measure that introduces dynamicity to the network [35], and
balance between risk and benefit [6]. Automated machine weakens the learning process of AI-enabled attackers. Quan-
ethics needs to be defined to ensure data privacy in future tum machine learning could also be used to design advanced
networks. defence techniques to resist AI-based attacks [27].

TABLE I: Significance/Importance of AI based Security/Privacy Solutions and Attacks on different 6G Applications

AI based Security/Privacy Solutions Attacks on AI/ML
Encryption with ML

Algorithm Poisoning
Differential Privacy

Blockchain with AI
Intrusion Detection

Machine Learning
for Authentication
and Authorization

Dataset Poisoning

Model Extraction
Privacy using AI

Model Poisoning

Model Inversion
Service oriented

Adversarial ML
Anomaly based

Model Evasion
Homomorphic
Quantum ML

Potential 6G Applications
Multisensory Extended Reality (XR) H M M H M M L M H M H H M M
Connected Robotics and Autonomous Systems (CRAS) H H H H H M M H H H H H H M
Wireless Brain-Computer Interactions (BCI) M M L L M H H L H M H H H H
Smart Grid 2.0 H M H L M L L H H H H H H M
Industry 5.0 H H M M H M M M H M H H H M

Low Importance Medium Importance High Importance

 V. C ONCLUSIONS [15] T. Hewa, G. Gür, A. Kalla, M. Ylianttila, A. Bracken, and M. Liyanage,
“The Role of Blockchain in 6G: Challenges, Opportunities and Research
AI is a key enabler for the next generation 6G mobile Directions,” in 2020 2nd 6G Wireless Summit (6G SUMMIT). IEEE,
networks and ensuring security is a critical consideration to 2020, pp. 1–5.
realize 6G vision. AI-enabled 6G security provides intelligent, [16] A. Dogra, R. K. Jha, and S. Jain, “A Survey on beyond 5G Network
with the Advent of 6G: Architecture and Emerging Technologies,” IEEE
robust security solutions. This paper provides an overview Access, 2020.
to comprehend the multitude of opportunities and challenges [17] G. Blinowski, “Security of Visible Light Communication Systems — A
of having intelligent security and privacy provision as a Survey,” Physical Communication, vol. 34, pp. 246–260, 2019.
[18] I. H. Abdulqadder, S. Zhou, D. Zou, I. T. Aziz, and S. M. A. Akber,
part of AI’s role in 6G systems. Moreover, it also identifies “Multi-layered Intrusion Detection and Prevention in the SDN/NFV
future research directions by discussing challenges in AI-based enabled Cloud of 5G Networks using AI-based Defense Mechanisms,”
security and privacy provision and suggest viable solutions. Computer Networks, vol. 179, p. 107364, 2020.
[19] R. Santos, D. Souza, W. Santo, A. Ribeiro, and E. Moreno, “Machine
Learning Algorithms to detect DDoS Attacks in SDN,” Concurrency
ACKNOWLEDGEMENT and Computation: Practice and Experience, vol. 32, no. 16, p. e5402,
This work is supported by Academy of Finland in 6Genesis 2020.
[20] M. Zolanvari, M. A. Teixeira, L. Gupta, K. M. Khan, and R. Jain,
Flagship (grant no. 318927) and 5GEAR (Grant No. 319669) “Machine Learning-Based Network Vulnerability Analysis of Industrial
projects. The research leading to these results partly received Internet of Things,” IEEE Internet of Things Journal, vol. 6, no. 4, pp.
funding from the European Union’s Horizon 2020 research 6822–6834, 2019.
[21] H. Fang, A. Qi, and X. Wang, “Fast Authentication and Progressive
and innovation programme under grant agreement no 871808 Authorization in Large-Scale IoT: How to Leverage AI for Security
(5G PPP project INSPIRE-5Gplus). The paper reflects only Enhancement,” IEEE Network, vol. 34, no. 3, pp. 24–29, 2020.
the authors’ views. The Commission is not responsible for [22] H. Fang, X. Wang, and S. Tomasin, “Machine Learning for Intelligent
Authentication in 5G and Beyond Wireless Networks,” IEEE Wireless
any use that may be made of the information it contains. Communications, vol. 26, no. 5, pp. 55–61, 2019.
[23] J. Wang and G. Joshi, “Cooperative SGD: A unified Framework for
R EFERENCES the Design and Analysis of Communication-Efficient SGD Algorithms,”
[1] W. Saad, M. Bennis, and M. Chen, “A Vision of 6G Wireless Systems: 2019.
Applications, Trends, Technologies, and Open Research Problems,” [24] C. Ma, J. Li, M. Ding, H. H. Yang, F. Shu, T. Q. S. Quek, and H. V. Poor,
IEEE Network, vol. 34, no. 3, pp. 134–142, 2019. “On Safeguarding Privacy and Security in the Framework of Federated
[2] C. de Alwis, A. Kalla, Q. V. Pham, P. Kumar, K. Dev, W. J. Hwang, Learning,” IEEE Network, vol. 34, no. 4, pp. 242–248, 2020.
and M. Liyanage, “Survey on 6G Frontiers: Trends, Applications, [25] Z. Chkirbene, A. Erbad, R. Hamila, A. Gouissem, A. Mohamed,
Requirements, Technologies and Future Research,” IEEE Open Journal M. Guizani, and M. Hamdi, “Weighted Trustworthiness for ML based
of the Communications Society, pp. 1–1, 2021. Attacks Classification,” in 2020 IEEE Wireless Communications and
[3] M. Giordani, M. Polese, M. Mezzavilla, S. Rangan, and M. Zorzi, Networking Conference (WCNC). IEEE, 2020, pp. 1–7.
“Toward 6G Networks: Use Cases and Technologies,” IEEE Commu- [26] C. Benzaid and T. Taleb, “ZSM Security: Threat Surface and Best
nications Magazine, vol. 58, no. 3, pp. 55–61, 2020. Practices,” IEEE Network, vol. 34, no. 3, pp. 124–133, 2020.
[4] V. Ziegler, H. Viswanathan, H. Flinck, M. Hoffmann, V. Räisänen, and [27] J. Biamonte, P. Wittek, N. Pancotti, P. Rebentrost, N. Wiebe, and
K. Hätönen, “6G Architecture to Connect the Worlds,” IEEE Access, S. Lloyd, “Quantum Machine Learning,” Nature, vol. 549, no. 7671,
vol. 8, pp. 173 508–173 520, 2020. pp. 195–202, 2017.
[5] M. Ylianttila, R. Kantola, A. Gurtov, L. Mucchi, I. Oppermann, Z. Yan, [28] L. Xiao, G. Sheng, S. Liu, H. Dai, M. Peng, and J. Song, “Deep
T. H. Nguyen, F. Liu, T. Hewa, M. Liyanage et al., “6G White Paper: Reinforcement Learning-enabled Secure Visible Light communication
Research Challenges for Trust, Security and Privacy,” arXiv preprint against Eavesdropping,” IEEE Transactions on Communications, vol. 67,
arXiv:2004.11665, 2020. no. 10, pp. 6994–7005, 2019.
[6] K. B. Letaief, W. Chen, Y. Shi, J. Zhang, and Y.-J. A. Zhang, “The [29] M. Liyanage, J. Salo, A. Braeken, T. Kumar, S. Seneviratne, and
Roadmap to 6G: AI Empowered Wireless Networks,” IEEE Communi- M. Ylianttila, “5G Privacy: Scenarios and Solutions,” in 2018 IEEE
cations Magazine, vol. 57, no. 8, pp. 84–90, 2019. 5G World Forum (5GWF), 2018, pp. 197–203.
[7] I. Ahmad, T. Kumar, M. Liyanage, J. Okwuibe, M. Ylianttila, and [30] Y. Sun, J. Liu, J. Wang, Y. Cao, and N. Kato, “When Machine Learning
A. Gurtov, “Overview of 5G Security Challenges and Solutions,” IEEE meets Privacy in 6G: A Survey,” IEEE Communications Surveys &
Communications Standards Magazine, vol. 2, no. 1, pp. 36–43, 2018. Tutorials, vol. 22, no. 4, pp. 2694–2724, 2020.
[8] R. Khan, P. Kumar, D. N. K. Jayakody, and M. Liyanage, “A survey [31] J. Li, X. Kuang, S. Lin, X. Ma, and Y. Tang, “Privacy Preservation for
on security and privacy of 5g technologies: Potential solutions, recent Machine Learning Training and Classification based on Homomorphic
advancements, and future directions,” IEEE Communications Surveys & Encryption Schemes,” Information Sciences, vol. 526, pp. 166–179,
Tutorials, vol. 22, no. 1, pp. 196–248, 2019. 2020.
[9] P. Ranaweera, A. D. Jurcut, and M. Liyanage, “Survey on Multi-Access [32] T. Zhang, T. Zhu, P. Xiong, H. Huo, Z. Tari, and W. Zhou, “Correlated
Edge Computing Security and Privacy,” IEEE Communications Surveys Differential Privacy: Feature Selection in Machine Learning,” IEEE
Tutorials, pp. 1–1, 2021. Transactions on Industrial Informatics, vol. 16, no. 3, pp. 2115–2124,
[10] S. Wijethilaka and M. Liyanage, “Survey on network slicing for internet 2020.
of things realization in 5g networks,” IEEE Communications Surveys & [33] M. d’Aquin, P. Troullinou, N. E. O’Connor, A. Cullen, G. Faller,
Tutorials, 2021. and L. Holden, “Towards an ’Ethics by Design’ Methodology for AI
[11] N. Wang, P. Wang, A. Alipour-Fanid, L. Jiao, and K. Zeng, “Physical- Research Projects,” in Proceedings of the 2018 AAAI/ACM Conference
Layer Security of 5G Wireless Networks for IoT: Challenges and on AI, Ethics, and Society, 2018, pp. 54–59.
Opportunities,” IEEE Internet of Things Journal, vol. 6, no. 5, pp. 8169– [34] E. Bird, J. Fox-Skelly, N. Jenner, R. Larbey, E. Weitkamp and A.
8181, 2019. Winfield, “The Ethics of Artificial Intelligence: Issues and Initiatives,”
[12] Y. Yang, L. Wu, G. Yin, L. Li, and H. Zhao, “A Survey on Security and European Parliamentary Research Service, March 2020. [Online].
Privacy Issues in Internet-of-Things,” IEEE Internet of Things Journal, Available: https://www.europarl.europa.eu/RegData/etudes/STUD/2020/
vol. 4, no. 5, pp. 1250–1258, 2017. 634452/EPRS_STU(2020)634452_EN.pdf
[13] ETSI, “Zero-touch Network and Service Management (ZSM),” ETSI GS [35] J.-H. Cho, D. P. Sharma, H. Alavizadeh, S. Yoon, N. Ben-Asher, T. J.
ZSM 002 - Reference Architecture, Aug 2019. Moore, D. S. Kim, H. Lim, and F. F. Nelson, “Toward Proactive,
[14] C. Benzaid and T. Taleb, “AI for Beyond 5G Networks: A Cyber- Adaptive Defense: A Survey on Moving Target Defense,” IEEE Com-
Security Defense or Offense Enabler?” IEEE Network, vol. 34, no. 6, munications Surveys & Tutorials, vol. 22, no. 1, pp. 709–745, 2020.
pp. 140–147, 2020.

View publication stats