Transfer Learning-Empowered Physical Layer Security in Aerial Reconfigurable Intelligent Surfaces-Based Mobile Networks

Abstract

This paper investigates the enhancement of physical layer security (PHY security) in Reconfigurable Intelligent Surfaces (RIS)-aided terrestrial and non-terrestrial networks (TN/NTN), focusing on the challenges posed by node mobility. In the context of next-generation mobile networks, ensuring secure communication is critical, especially under varying channel conditions caused by mobility. We explore different mobility models, including random walk, Gauss-Markov, and reference point group mobility, to assess their impact on key security metrics such as secrecy capacity and average secrecy rate. To address these challenges, we develop robust algorithms for optimizing the phase-shift configurations of RIS. Additionally, we employ Artificial Intelligence (AI) and Machine Learning (ML) techniques, specifically Deep Neural Networks (DNN), for performance prediction of PHY security metrics. We also leverage transfer learning to enhance model robustness across different mobility scenarios through domain adaptation. Our results demonstrate the effectiveness of our proposed methods in maintaining high levels of PHY security despite the dynamic nature of the channel conditions and the mobility of nodes. The proposed phase-shift configuration algorithms and ML-based solutions ensure secure and resilient communication in aerial RIS-aided TN/NTN, contributing to the advancement of secure mobile networks.