Adaptive Cyber Defense Technique Based on Multiagent Reinforcement Learning Strategies

The static nature of cyber defense systems gives attackers a sufficient amount time to explore and further exploit the vulnerabilities information technology systems. In this paper, we investigate problem where multiagent sensing acting in an environment contribute adaptive defense. We present learning strategy that enables multiple agents learn optimal policies using reinforcement (MARL). Our proposed approach is inspired by multiarmed bandits (MAB) technique for cooperate decision making or work independently. study MAB which defenders visit system times alternating fashion maximize their rewards protect system. find game can be modeled from individual player’s perspective as restless problem. discover results when takes form pure birth process, such myopic policy, well providing environments offer necessary incentives required cooperation multiplayer projects.