Network Defense Strategy Model Based on Deep Reinforcement Learning and Evolutionary Game Theory
Zhi‐Wei Sun, Hai Yang, Rui Ma · 2025
To address the limitations of deep reinforcement learning (DRL) in modeling adversarial strategy evolution and the rigidity of evolutionary game theory in dynamic scenarios, this paper proposes a Dual-layer Cooperative Evolutionary Framework (DCEF). DCEF integrates the Asynchronous Advantage Actor-Critic (A3C) algorithm for real-time defense generation with an evolutionary game layer for global optimization, leveraging an incremental payoff matrix and adaptive triggering mechanisms. Experiments on the CICIDS2017 dataset demonstrate DCEF’s superiority: 86.2% defense success rate (5.6% higher than A3C), 38ms response latency, and robust adaptability to multi-stage attacks. The framework effectively balances short-term responsiveness with long-term strategic stability.