Cyber Agents for Security Testing and Learning Environments (CASTLE) CompetitionThis project addresses the pressing need to develop the capability to defend computer networks from sophisticated adversaries and advanced persistent threats. As the scale and complexity of cyber attacks grows, network operators need the ability to understand and defend against these attacks in real-time. Autonomous cyber agents offer the promise of automating the defense against such complex attacks, but developing agents capable of reasoning across multiple layers of the computing stack is a challenging task. This project investigates the use of reinforcement learning and AI in developing autonomous cyber agents capable of defending a computer network against such attacks.