TrustV2X: A Hybrid Hierarchical and Transfer Federated Learning Framework for Trustworthy Next-Generation V2X Network Communication

초록

The increasing complexity of vehicular communication ecosystems, such as Vehicle-to-Everything (V2X) systems, introduces multifaceted cybersecurity challenges. While traditional intrusion detection systems (IDS) rely on centralized architectures, they fail to meet latency and privacy constraints of distributed vehicular networks. This paper presents TrustV2X, a novel hybrid Hierarchical Federated Learning (HFL) and Federated Transfer Learning (FTL) framework for V2X intrusion detection. The design integrates edge-level training cycles, transfer-phase adaptation schedules, and a rule-based client vetting mechanism to achieve privacy-preserving and intrusion-resilient learning at Roadside Units (RSUs) and Multi-access Edge Computing (MEC) aggregators. Experimental validation on the CICIoV2024 dataset demonstrates that the proposed framework achieves an F1-score of 100% and reduces training latency by 27.4% compared to conventional FL-based IDS models. The proposed hybrid architecture provides a scalable pathway for trustworthy and adaptive V2X security systems.