Enhanced Secure Resource Allocation in Eavesdropped V2X Networks based on Multi-Agent Reinforcement Learning

The advancement of communication technologies has established vehicular networks as essential infrastructure for intelligent transportation systems. However, the inherent broadcast nature of wireless communications exposes transmitted data to eavesdropping attacks, creating significant security vulnerabilities. This paper addresses the challenge of secure resource allocation in eavesdropped V2X networks. Specifically, we propose a cooperative jamming scheme that strategically utilizes candidate cooperative vehicles to transmit artificial noise to effectively degrade the eavesdropping capacity without severe effects on communication quality of legitimate links, aimed at enhancing the security while maintaining Quality of Service (QoS) of vehicular networks. To solve the complex resource optimization problem encompassing power control for legitimate transmissions, channel allocation among vehicles, and power distribution for cooperative jammers in highly dynamic environments, we develop a Multi-Agent Deep Reinforcement Learning (MADRL) algorithm based on DQN. Experimental results show that our method significantly outperforms baseline strategies in both communication quality and security effectiveness.