Dynamic Optimization of Defense Architecture Based on Deep Reinforcement Learning

With the development of internet technology and the increasing number of cyber threats, traditional static network defense strategies have shown numerous deficiencies in dealing with complex and variable attacks. To address this issue, this paper proposes a dynamic optimization method for defense architecture based on deep reinforcement learning, aiming to enhance the intelligence and adaptability of network defense systems. This method first constructs a defense architecture model based on reinforcement learning, transforming the network defense problem into an optimization problem of state space, action space, and reward function. Subsequently, the Dueling DQN algorithm is employed to achieve dynamic deployment and adjustment of defense strategies. By separating state values and action advantages, this algorithm more effectively learns the optimal defense strategy. Experimental results indicate that, compared to traditional greedy algorithms, the proposed method significantly improves defense efficiency and asset protection.