Research on the Evolution of Oil and Gas Pipeline Network Attack and Defense Based on Reinforcement Learning and Game Theory

Abstract

As a critical national infrastructure, the oil and gas pipeline network faces more complex and dynamically evolving cybersecurity risks during the process of informatization and intelligentization. Attackers usually have the advantage of information asymmetry and can flexibly adjust the attack path in the multi-stage chain structure, which poses higher requirements for the adaptability of the pipeline network defense system. The article proposes an intelligent defense modeling method that integrates game modeling, deep reinforcement learning, and strategy evolution mechanisms. A dual-agent confrontation structure is constructed to simulate real attack and defense behaviors, enabling the self-optimization of defense strategies in a dynamic environment. The model introduces the replicator dynamic equation in evolutionary game theory to enhance the stability of strategy updating and the expressive ability of diversity, thus making up for the robustness deficiency of traditional methods in incomplete information scenarios. The experimental results show that this method reduces the attack success rate to 17.2% and enhances system stability to 89.1% in typical attack and defense scenarios, exhibiting superior strategy adaptability and robustness in complex environments.