Cyber threats have evolved into more sophisticated forms of coordinated, autonomous multi-agent attacks. Those
are now termed as AI-Predator Swarms (APS), which are decentralized systems of autonomous, self-learning agents that
interact and organize their methods of attack to pursue the most desirable malicious goals without a central controller being
present (Vorobeychik& Kantarcioglu, 2018). Although conventional AI-derived cyber threats, which include automated
phishing and mutating malware, are merely the use of AI, APS is the actual body and practice of swarm intelligence, in
which behaviors of individual agents are harmonized to attain complex and adaptable patterns of attack (Wooldridge, 2009).
The paper describe AI-Predator Swarm threat model, discuss the principles of multi-agent reinforcement learning (MARL),
and propose a defenses model against such highly distributed and low-signal attacks that overcome conventional intrusion
detection systems (IDS) (Cui et al., 2023).
[1] A Aly, “Applied Cryptography and Network Security: Fundamentals and Infrastructure,” IEEE Communications
Surveys & Tutorials, vol. 20, no. 2, 2018.
[2] C. Szegedy et al., “Intriguing Properties of Neural Networks,” in Proc. ICLR, 2014.
[3] D. Silver et al., “Mastering the Game of Go with Deep Neural Networks and Tree Search,” Nature, vol. 529, pp. 484
489, 2016. (For reinforcement learning foundations)
[4] H. Cui et al., “Graph Neural Networks for Cybersecurity: A Survey,” IEEE Transactions on Neural Networks and
Learning Systems, vol. 34, no. 1, pp. 1–20, 2023.
[5] I. Goodfellow, J. Shlens, and C. Szegedy, “Explaining and Harnessing Adversarial Examples,” in Proc. ICLR, 2015.
[6] IEEE Global Initiative on Ethics of Autonomous and Intelligent Systems, Ethically Aligned Design, First Ed., 2019.
[7] J. Hu, G. Wang, and J. Zhang, “Temporal Graph Neural Net-works for Dynamic Pattern Detection,” IEEE Transactions
on Knowledge and Data Engineering, vol. 35, no. 4, pp. 3231–3245, 2023.
[8] L. Busoniu, R. Babuska, and B. De Schutter, “A Comprehensive Survey of Multi-Agent Reinforcement Learning,”
IEEE Transactions on Systems, Man, and Cybernetics – Part C, vol. 38, no. 2, pp. 156–172, 2008.]
[9] M. E. Newman, “Networks: An Introduction,” Oxford University Press, 2010. (Foundational for graph modelling)
[10] M. Wooldridge, An Introduction to Multiagent Systems, 2nd Ed., Wiley, 2009.
[11] MITRE Corporation, “Adversarial Threat Landscape for Artificial-Intelligence Systems (ATLAS),” 2022.
(Authoritative framework for AI adversarial scenarios).
[12] N. Papernot et al., “The Limitations of Deep Learning in Adversarial Settings,” Proc. IEEE European Symposium on
Security and Privacy, 2016.
[13] P. Barford, C. Papadopoulos, and X. Yao, “Anomaly Detection for Network Traffic Using Machine Learning,” IEEE
Transactions on Network and Service Management, vol. 17, no. 4, pp. 2143–2156, 2020.
[14] S. Russell and P. Norvig, Artificial Intelligence: A Modern Approach, 4th Ed., Pearson, 2021.
[15] S. Sengupta et al., “AI in Cybersecurity: A Review of Applications,” IEEE Access, vol. 8, pp. 200936–200957, 2020.
[16] S. Zhong, J. Shi, and W. Zhu, “Security Challenges of Ma-chine Learning in Computer Networks,” IEEE Network, vol.
34, no. 6, pp. 313–319, 2020.
[17] W. Wei et al., “Malware Detection Using Deep Learning Based on Hybrid Features,” IEEE Transactions on Network
and Service Management, vol. 18, no. 3, pp. 2909–2922, 2021.
[18] Y. LeCun, Y. Bengio, and G. Hinton, “Deep Learning,” Nature, vol. 521, pp. 436–444, 2015.
[19] Y. Vorobeychik and M. Kantarcioglu, Adversarial Machine Learning, Morgan & Claypool, 2018.
[20] Z. Qin, G. Wang, and C. J. Hsieh, “Generative Adversarial Networks: A Survey Towards Data Augmentation and
Adversarial Attack/Defense,” IEEE Transactions on Neural Net-works and Learning Systems, 2021.
