MARL-MOTAG: Multi-Agent Reinforcement Learning Based Moving Target Defense to thwart DDoS attacks

The popularity of intelligent methods has expanded the means of DDoS attacks, which has significantly impacted online services. The static defense mechanism lacks the resistance to flooding, and the moving target defense has become an effective method to defend against distributed denial of service (DDoS) attacks. In order to adapt dynamic defense according to network conditions, while reducing resource consumption. In this paper, we propose a multi-agent reinforcement learning system (MARL-MOTAG) based on the MOTAG system, which can adaptively make decisions based on the server status. MA-MOTAG retains the proxy server settings of MOTAG and separates the proxy server into two clusters according to the degree of damage. The resource consumption caused by user migration is reduced through the new shuffling mechanism. At the same time, multi-agent reinforcement learning reduces the complexity of the action space and can quickly feedback and adaptively divide server clusters for complex network environments. Simulation results show that the proposed algorithm can converge better and resist DDoS attacks while reducing migration resource consumption.