Decentralized Trust Management Model to Detect Malicious Nodes in Internet of Vehicles

Abstract

With the rapid expansion of the Internet of Vehicles, ensuring security and trust among nodes has emerged as one of the fundamental challenges in this domain. The open, dynamic, and distributed nature of these networks creates a suitable environment for malicious nodes, which can compromise communication integrity and overall system security by disseminating false or misleading information. This research aims to present a hybrid and decentralized trust management model that, through a multilayer approach, can effectively detect and analyze malicious nodes in connected vehicular networks. The proposed framework adopts a two-layer structure: in the first layer, vehicles compute short-term local trust scores of their peers based on interaction data using the proposed LTrustAssess algorithm; while in the second layer, roadside units model the network as a graph and employ the proposed deep learning model, TemporalGATwithLSTM, to predict and update the global and long-term trust scores of nodes over time. Experimental evaluation on a dataset generated from simulated vehicular interaction logs demonstrates that the proposed model achieves higher accuracy and efficiency in trust scores distribution and detecting malicious nodes compared to existing baseline approaches. Overall, by providing a scalable and adaptive mechanism, the proposed model enhances the security, trust and efficiency of vehicular networks and represents a significant step toward realizing future intelligent and safe transportation systems.