Enhanced intrusion strategy learning for security systems using an optimized PPO algorithm

In response to the limitations observed in critical infrastructure protection systems, particularly their lack of adaptabilit y in dynamic and complex environments, this paper introduces an intrusion behavior strategy analysis method utilizing an enhanced Proximal Policy Optimization (PPO) algorithm. This method accounts for the decision-making behavior of attackers who have access to partial information, enabling it to effectively simulate authentic intrusion strategies. By leveraging feedback fro m environmental interactions as a learning mechanism, the approach facilitates agent strategy learning to accurately simulate optimal intrusion strategies. This advancement aids in the strategic deployment and enhancement of defenses at critical infrastructure vulnerabilities. Test results demonstrate that the improved deep reinforcement learning algorithm exhibits rapid convergence and stable computational outcomes. The proposed intrusion behavior strategy analysis model displays robust learning capabilities and strong generalizability, offering a novel framework for analyzing attack and defense strategies and advancing security defenses in physical protection systems against novel intrusion challenges.