Knowledge Guided Two-player Reinforcement Learning for Cyber Attacks and Defenses
Published in 21st IEEE International Conference on Machine Learning and Applications (ICMLA), 2022
Recommended citation: A. Piplai, M. Anoruo, K. Fasaye, A. Joshi, T. Finin, and A. Ridley, "Knowledge Guided Two-player Reinforcement Learning for Cyber Attacks and Defenses," in 21st IEEE International Conference on Machine Learning and Applications (ICMLA), Nassau, Bahamas, 2022, pp. 1342-1349, doi: 10.1109/ICMLA55696.2022.00213. https://doi.org/10.1109/ICMLA55696.2022.00213
The paper introduces a two-player game-based reinforcement learning (RL) environment designed for cyber defense exercises. It leverages recent advancements in cyber battle simulation platforms to train RL-based autonomous agents, enhancing both attacker and defender performance, with accelerated convergence achieved through guidance from expert knowledge extracted from Cybersecurity Knowledge Graphs on attack and mitigation steps. The proposed approach has been implemented and integrated into the CyberBattleSim system.