Adaptive Honeypot Defense Deployment: A Stackelberg Game Approach with Decentralized DRL for AMI Protection

Honeypot defenses are pivotal for safeguarding the Industrial Internet of Things (IIoT), notably the Advanced Metering Infrastructure (AMI), against cyber threats. The success of AMI defense relies on the strategic deployment of small-scale power suppliers (SPSs) and their interaction with traditional power retailers (TPR). Existing methods require exhaustive information exchange, which is not feasible. Prior studies also neglected the competitive aspect among the SPSs in task allocation. Our paper introduces a Stackelberg game model to address TPR-SPS interactions and SPS competition comprehensively. Our proposed approach stands out by eliminating the need for prior deployment and data sharing. It leverages the multi-agent deep deterministic policy gradient (MADDPG) algorithm, centralized training, and distributed execution, effectively adapting to changing environments. Without relying on historical data, each SPS actively learns from its surroundings. Our simulations validate the efficiency of this novel approach.