Enhancing Security With Hybrid Active- Passive RIS: A DRL Approach Against Eavesdropping and Jamming

Abstract

The security of wireless communications is increasingly threatened by eavesdropping and jamming attacks. This paper proposes a novel framework for enhancing secure transmission using hybrid active-passive Reconfigurable Intelligent Surfaces (RIS). By combining active and passive elements, the hybrid RIS can dynamically adjust the amplitude and phase of reflected signals, providing a robust defense against eavesdropping and jamming. The key challenge lies in optimizing the beamforming at the base station (BS) and the hybrid RIS configuration in real time. To tackle this, we employ a Deep Reinforcement Learning (DRL) approach using the Deep Deterministic Policy Gradient (DDPG) algorithm, enabling efficient and dynamic optimization. Simulation results demonstrate that the proposed DRL-based method significantly improves the secrecy rate compared to conventional passive RIS and benchmark methods. Our results indicate that the system can achieve substantial security gains even with a limited number of active RIS elements, making it a viable solution for next-generation wireless networks.

Keywords

• Deep reinforcement learning (DRL)
• hybrid active-passive RIS
• physical layer security (PLS)