Physical Layer Authentication in UAV-enabled Relay Networks Based on Manifold Learning
基于流形学习的无人机中继网络物理层认证
多様体学習に基づくUAV対応リレーネットワークの物理層認証
매니폴드 학습 기반 UAV 지원 릴레이 네트워크의 물리 계층 인증
Autenticación de capa física en redes de retransmisión habilitadas para UAV basadas en aprendizaje múltiple
Authentification de la couche physique dans les réseaux de relais compatibles UAV basée sur l'apprentissage multiple
Аутентификация физического уровня в ретрансляционных сетях с поддержкой БПЛА на основе многообразного обучения
Shida Xia 夏仕达 ¹, XiaoFeng Tao 陶小峰 ¹, Na Li 李娜 ¹ ², Shiji Wang 王施霁 ¹, Jin Xu 徐瑨 ¹
¹ The National Engineering Lab for Mobile Network Technologies, Beijing University of Posts and Telecommunications, Beijing 100876 China
中国 北京 北京邮电大学 移动互联网安全技术国家工程实验室
² Beijing University of Posts and Telecommunications Research Institute, Shenzhen, 518000, China
中国 深圳 北京邮电大学深圳研究院
Science China Information Sciences, 20 January 2022
Abstract
Unmanned aerial vehicle (UAV) relay network is a promising solution in the next-generation wireless networks due to its high capacity and unlimited geography. However, because of the openness of wireless channels and UAV mobility, it is remarkably challenging to guarantee the secure access of UAV relay. In this paper, we investigate the physical layer authentication (PLA) to verify the identity of the UAV relay for preventing unauthorized access to users' information or network service.
Unlike the most existing PLA methods for UAV, the proposed PLA scheme fully considers the time-varying of physical layer attributes caused by UAV mobility, and transforms the authentication problem into recognizing nonlinearly separable physical layer data. Particularly, we propose a manifold learning-based PLA scheme that can authenticate the mobile UAV relay in real-time by establishing the local correlation of physical layer attributes. The Markov chain of physical layer data in the time domain is established to evaluate UAV state transition probability through the proposed diffusion map algorithm. The legitimate UAV and spoofing attackers can always be authenticated by the different motion states.
Performance analysis offered a comprehensive understanding of the proposed scheme. Extensive simulations confirm that the performance of the proposed scheme improves over 18% in resisting the intelligent spoofing UAV compared to the traditional methods.
