Intelligent decision-making framework based on static game and PPO algorithm

Intelligent decision-making of abnormal indicators in network security enterprises is the core challenge of risk management. Traditional methods have limitations in multi-party interest game and dynamic environment adaptability. This paper proposes an intelligent decision-making framework combining static game and Policy Gradient (SG-PPO), which aims to solve the collaborative problem of multi-party strategy interaction and dynamic optimization in network attack and defense scenarios. Firstly, a static game model was constructed to quantify the strategy space and payoff function of defenders, attackers and security regulators, and Nash equilibrium was solved as the decision-making benchmark. Secondly, a reinforcement learning algorithm based on equilibrium constraint was designed, and the game equilibrium solution was integrated into the reward function and strategy exploration process of PPO to realize the adaptive strategy optimization in dynamic environment. Experiments show that the proposed model is significantly better than the single-game model or pure PPO method in terms of the decision profit and strategy stability of anomaly detection. This research provides a solution with both theoretical rigor and practical flexibility for enterprises to balance risk control and operational efficiency, and opens up a new path for the integration application of game theory deep learning and intelligent security decision-making in the adversarial network environment.