IRIS-DRL FOR SECRECY IN RIS-ASSISTED UAV NETWORKS

Abstract

This paper introduces a novel deep reinforcement learning (DRL) framework Intelligent Reconfigurable Integrated Security (IRIS) designed to enhance the secrecy and efficiency of communication in Reconfigurable Intelligent Surface (RIS) assisted Unmanned Aerial Vehicle (UAV) networks. IRIS jointly optimizes UAV flight trajectories, RIS phase configurations, and power allocation strategies to strengthen physical-layer security in environments susceptible to eavesdropping and signal interference. Unlike traditional (DRL) methods such as Proximal Policy Optimization (PPO), which often struggle with high-dimensional optimization tasks in dynamic wireless systems, IRIS employs an adaptive exploration-exploitation mechanism tailored for secure UAV operations. The framework dynamically responds to environmental changes, maximizing the secrecy rate while minimizing energy consumption and latency. Simulation results, conducted using MATLAB, demonstrate that IRIS significantly outperforms conventional approaches across multiple performance indicators, including secrecy rate, convergence speed, and energy efficiency. A comprehensive sensitivity analysis of key hyperparameters further validates the model's robustness across various deployment scenarios. The results highlight IRIS as a promising algorithm for secure communication in UAV-enabled applications such as disaster relief, critical infrastructure monitoring, and next-generation IoT deployments.