Differentially private federated adversarial learning–based intrusion prevention for V2X communications

Abstract

Vehicular networks are a continuation of vehicle-to-everything (V2X) communication, which is becoming the foundation of intelligent transportation systems and allows vehicles, roadside units, and grid infrastructure to communicate reliably. Nevertheless, the decentralized aspect of V2X renders it very susceptible to adversarial cyberattacks, both on a large scale such as DDoS attacks, spoofing, and Sybil attacks. Conventional centralized intrusion detection systems (IDSs) have limitations in terms of latency, bandwidth overhead, and privacy risks. In this study, we introduce a Differentially Private Federated Adversarial Learning (DP-FAL) model that integrates federated learning to train decentralized models with adversarial defence schemes and differential privacy tools to address gradient leakage. The proposed DP-FAL framework achieves a detection accuracy of 94.2% and a communication overhead reduction of up to 23%. This shows that it can be used in bandwidth-limited and latency-sensitive V2X systems because of its ability to ensure high detection reliability with low communication costs. The suggested DP-FAL model is a privacy-conserving, scalable, and robust intrusion prevention system that can be used real-time V2X conditions. These results indicate that DP-FAL has the potential to be deployed safely, reliably, and sustainably in next-generation transportation systems.