Robust Secure UAV-RIS-Aided ISAC Against Multiple Intelligent Eavesdroppers

Abstract

This paper proposes a secure downlink vehicle-toeverything integrated sensing and communication architecture aided by an unmanned aerial vehicle (UAV) equipped with a reconfigurable intelligent surface (RIS). An adversarial interplay framework is introduced in which a base station simultaneously performs communication and target sensing, while the legitimate UAV serves a dual purpose by providing passive beamforming and transmitting active jamming signals to suppress multiple intelligent UAV eavesdroppers. To counter dynamic eavesdroppers, a non-cooperative game is formulated to maximize the system’s secrecy energy efficiency under a bounded channelstate-information error model, subject to minimum sensing-rate guarantees, robust secure-rate thresholds, power budgets, and flight trajectories. To solve this highly coupled optimization problem, a sample average approximation-based multi-agent deep deterministic policy gradient algorithm is proposed and demonstrated to achieve a 125.19% performance improvement over the no-RIS scheme.