UAV-enabled aerial active RIS with learning deployment for secured wireless communications

Reconfigurable Intelligent Surfaces (RISs) enable programmable wireless environments and thus have great potential for enhancing physical layer security. However, the security gain of conventional passive RISs is often limited by the “multiplicative fading” effect through reflection links, which becomes severe in the case of double reflections and significantly degrades the security performance. In this paper, we consider a wireless system that consists of a fixed passive RIS and an Unmanned Aerial Vehicle (UAV)-mounted active RIS, where the UAV-enabled aerial amplification and reflection are exploited to compensate for the multiplicative fading effect. We formulate the problem to maximize the secrecy rate by jointly considering the optimal deployment of the UAV-based active RIS and the reflection coefficients at both the passive and active RISs. To enable efficient algorithm design, we decompose the problem into two layers: the outer layer optimizes the UAV deployment through deep reinforcement learning, while the inner layer solves the beamforming and reflection design using a block coordinate descent framework. Simulation results demonstrate the convergence of the proposed learning procedure, and indicate that the active RIS with learned deployment can effectively enhance the reflection and significantly improve the secrecy rate.