Misbehavior Detection System in V2X 5G Edge Networks based on Transformer

With the increasing reliance on V2X communications for Cooperative Intelligent Transport Systems, ensuring the integrity and reliability of transmitted data has become a critical challenge. Misbehavior detection systems (MDS) play a vital role in identifying malicious activities that can compromise road safety and network efficiency. Traditional approaches, such as LSTM-based models, face limitations in capturing complex patterns and distinguishing attacks that replicate legitimate behavior. To address these challenges, this paper proposes a Transformer-based model integrated within a 5G Edge Misbehavior Detection System. Leveraging the self-attention mechanism of Transformers, the model effectively captures nuanced temporal and spatial dependencies in vehicle movement data. The framework employs federated learning to maintain data privacy while achieving scalability, and includes personalization techniques to enhance performance. Evaluations on the VeReMi-extension dataset reveal that the Transformer-based model achieves an accuracy of 99.37%, a recall of 98.69%, and a F1-score of 99.30% demonstrating its robustness across various deployment scenarios and attack types.