Cooperative Proactive Eavesdropping based on Deep Reinforcement Learning

There is illegitimate transmission of information between suspicious users, whereas single legitimate monitor (LM) has finite capacity to satisfy eavesdropping. This letter studies a cooperative proactive eavesdropping(CPE) scheme, where two LMs eavesdrop on multiple suspicious links simultaneously with a cooperative jamming method under finite power constraint. Specifically, two LMs collaborate to emit jamming signals for the purpose of influencing the rate of suspicious links and facilitate successful eavesdropping at each LM. However, how to make jamming power decision over multiple suspicious links to maximize cumulative sum eavesdropping energy efficiency (EEE) in a long term for each LM under dynamic environment and continuous action space is a huge challenge. To solve the dynamic decision-making problem, we use multi-agent deep deterministic policy gradient (MADDPG) algorithm to solve the cooperative jamming problem. In simulation, the results show that our proposed CPE scheme can obtain the effective sum EEE compared with DDPG-based scheme and Random policy scheme.