Abstract

Vehicular Ad Hoc Networks (VANETs) are a cor-
nerstone of modern Intelligent Transportation Systems (ITS),
enabling real-time communication among vehicles and infrastruc-
ture. However, the open and dynamic nature of VANETs exposes
them to a wide range of cybersecurity threats such as spoofing,
Sybil attacks, and denial-of-service (DoS). This paper introduces
a novel Federated Learning (FL) framework designed to enhance
VANET security by enabling distributed and privacy-preserving
intrusion detection across the network. By leveraging local model
updates instead of centralized data aggregation, our proposed
FL approach mitigates privacy risks, reduces communication
overhead, and offers robust detection of cyber threats. The paper
presents a comprehensive analysis including system architecture,
threat modeling, security properties, performance evaluation,
and real-world applicability. Extensive simulations show that our
model achieves detection accuracy of up to 96.2%, with minimal
latency and low model convergence time, outperforming existing
centralized and traditional machine learning models.

Key words: Federated Learning, VANET, Intrusion Detec- tion System, Cybersecurity, Distributed AI, Privacy Preservation, Edge Computing