Optimal Strategy Selection for Cyber Deception via Deep Reinforcement Learning
Yuantian Zhang, Feng Liu, Huashan Chen · 2022
Cyber deception has emerged as an effective approach to enhance the security of computer systems. One important factor that affects the security effectiveness of cyber deception is the deployment strategy. In this paper, we develop a honeypot-based defense mechanism to deploy honeypot intelligently. We propose a Stackelberg game model to simulate the interaction between network attackers and defenders. We use attack graph to generate the attacker’s strategy and find the optimal honeypot deployment strategy in a constant confrontation. We develop a reinforcement learning (RL) based approach to train the attack strategy and the honeypot deployment strategy. To cope with the large strategy space problem, we modify the neural network structure for the attacker and the defender according to the deception game rules. We conduct experiments with real network data and various game settings to evaluate the performance of our approach. Experiment results show that our RL-based approach outperforms baselines, and the approach is robust to the uncertain security environment.