Autonomous network defense in cloud data center environments based on reinforcement learning
Jianping Zhao, Yichuan Wang, Huan Wang, Hua Li, Yunlong Tang · 2024
The internal networks of cloud data centers are vast in scale, rich in data value, and high in traffic, imposing stringent demands on network security. Manual security operations for such network resources are highly resource-intensive and prone to errors. To address this, this paper proposes an autonomous network defense method for internal networks of cloud data centers based on reinforcement learning. We construct a leader-follower game model for network attack and defense, depicting the NP-hard interactive behavior in network attack and defense; we design a reinforcement learning defense agent that autonomously iterates learning, continuously feedback control, and adapts to environmental changes, training to solve for autonomous network defense strategies. Experiments demonstrate that our reinforcement learning agent acquires an excellent autonomous network defense capability, effectively resisting network attacks.