Multi-agent Reinforcement Learning for Cybersecurity: Approaches and Challenges

In the face of the rapidly evolving threat landscape, traditional security measures often lag behind with sophisticated cyber attacks. Through a review of existing literature, we examine the shortcomings of conventional cybersecurity methods, highlighting the need for Reinforcement Learning based methods. Our study classifies various RL approaches in cybersecurity, aimed to enhance detection, mitigation, and response capabilities, along two dimensions: the RL technique used, and the network configuration. Moving forward, we emphasise the importance of further research and development to address challenges such as model complexity, sample efficiency, and vulnerabilities to adversarial attacks.