Reinforcement Learning-Driven Secrecy Energy Efficiency Maximization in RIS-Enabled Communication Systems

Reconfigurable intelligent surfaces (RISs) are becoming an innovative technology for sixth-generation (6G) wireless networks, providing improved coverage and spectral efficiency. Nonetheless, incorporating RIS into 6G systems brings forth considerable security challenges, especially given the potential for multiple eavesdroppers. This study explores the use of physical layer security (PLS) to facilitate secure communications in RIS-assisted networks, tackling the important issues of safeguarding sensitive information while also addressing limited energy resources. This study examines the optimization of RIS passive beamforming, active beamforming, and power budgets with the aim of enhancing secrecy energy efficiency (SEE). Given the intricate nature of this problem, we utilize artificial intelligence, particularly deep reinforcement learning. By treating the problem as a Markov decision process (MDP), we make it easier to make decisions in real-time by creating specific states and actions, along with a reward system designed to balance privacy and energy efficiency. We introduce an innovative framework that utilizes the deep deterministic policy gradient (DDPG) algorithm to address the challenges of the MDP. Our extensive simulations show that our method works much better than current techniques using deterministic policy gradients, like the proximal policy optimization (PPO) framework, in improving SEE. Our results demonstrate the ability of DDPG to improve secure and energy-saving communications with RIS, offering a robust and flexible solution for the changing and tough conditions expected in future 6G networks.