Secrecy energy efficiency optimization for IRS-aided multi-user communication system with cooperative jamming

In this thesis we investigate the problems to be solved by intelligent reflecting surface (IRS)-assisted communication system in the face of eavesdropping attacks, and proposed a deep reinforcement learning (DRL)-based anti-eavesdropping strategy that balances energy consumption with secrecy rate. In this system we additionally introduce a cooperative jammer (CJ) who works together with the base station (BS) and the IRS. The IRS is able to reflect both the signal from the BS and the signal from the CJ to increase the signal-to-interference-plus-noise ratio of the legitimate user or to reduce the eavesdropping rate of the eavesdropper, which ultimately achieves the goal of improving the system energy efficiency (SEE). To ensure efficient system operation, we propose the use of a double-twin-delayed deep deterministic policy gradient (DTD3) optimization algorithm. This algorithm can effectively solve the complex joint optimization problem between the active beamforming matrix of the BS, the CJ, and the phase-shift matrix of the IRS. The simulation results show that the SEE of the communication system is greatly improved with the help of CJ and DTD3 optimisation algorithm.