Secure Communication With UAV-Enabled Aerial RIS: Learning Trajectory With Reflection Optimization

Reconfigurable intelligent surfaces (RISs) have manifested huge potential in enhancing information security by actively intervening the wireless propagation, yet the security gain may still be limited depending on the RIS deployment. In this paper, we propose to employ an unmanned aerial vehicle (UAV) mounting a RIS to enable on-demand reflection, noted as an aerial RIS (ARIS). The ARIS is then exploited to assist the anti-eavesdropping communications established through a conventional fixed-deployed RIS to further enhance the wireless secrecy. The secure communication is investigated by jointly optimizing the reflection at both RISs as well as the trajectory of the ARIS to maximize the average secrecy rate during the flight. To facilitate effective algorithm design, the formulated security problem is decomposed and solved in a double-layer framework. The outer layer tackles the flying trajectory through deep reinforcement learning while the inner layer solves for reflection phase shift design with manifold optimization. Finally, simulation results demonstrate the learned trajectory in various topologies as well as the superior performance of our proposal in terms of security provisioning.