Poster: Secrecy rate analysis of RIS-assisted UAV networks using DDQN Network

Abstract

Physical layer security (PLS) is a critical research challenge in unmanned aerial vehicle (UAV)-assisted wireless networks due to the intrinsic vulnerability of a UAV's line of sight (LoS) to eavesdroppers' attacks. In this paper, a Double Deep Q-Network (DDQN) network model is proposed to optimize the UAV trajectory for a reconfigurable intelligent surface (RIS)-assisted network, with an aim of maximizing the Secret Key Rate (SKR). SKR is calculated against multiple eavesdroppers. Two network scenarios, including fixed-RIS, UAV and no-RIS-UAV, are compared and demonstrated in this paper. The proposed DDQN algorithm automatically learns UAV trajectory to gain maximum secrecy reward while taking 10 ground terminals (GT) with minimum data requirements. The validity of the proposed DDQN network is verified by simulating the SKR based on the number of GT's and eavesdroppers. Moreover, the presentation of 2D and 3D UAV trajectories. Simulation results show that RIS-assisted UAV always provide greater SKR than no-RIS UAV systems, which proves efficacy of the designed scheme.