Positional forgery by malicious vehicles can severely disrupt AI-driven autonomous intersection management systems, leading to inaccurate traffic predictions and increased accident risks. Traditional detection methods struggle in multi-tiered vehicular networks. This paper presents a Trust-Based Positional Forgery Detection (TPFD) system that leverages a Hybrid Graph-based Reinforcement Learning (HGRL) framework to combat positional forgery. The system integrates Graph Neural Networks (GNNs) for modelling Vehicle-to-Everything (V2X) communication, Recurrent Neural Networks (RNNs) for analysing temporal data, and Reinforcement Learning (RL) for dynamic trust evaluation. TPFD achieved a detection accuracy of 98.3%, reduced the false positive rate to 3.5%, and scaled efficiently to networks of up to 10,000 vehicles. The system also reduced hazardous traffic gap errors by 30%, enhancing intersection safety. This research strengthens the reliability and security of AI-based traffic systems, contributing to more efficient traffic management and increased public trust in autonomous vehicles.
Trust-based positional forgery detection in AI-driven autonomous intersection management using hybrid graph-based reinforcement learning
09.12.2024
5344006 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Taylor & Francis Verlag | 2025
|ArXiv | 2023
|Auction-based autonomous intersection management
IEEE | 2013
|