FL-SATS: Federated Learning for Sybil Attack Detection in Transportation System

The Intelligent Transportation System (ITS) is advancing with enhanced vehicular networks, making security a critical concern. A major threat to these networks is Sybil attacks, where adversaries forge multiple identities to compromise the system. We propose Federated Learning for Sybil Attack Detection in Transportation Systems (FL-SATS), a mechanism leveraging federated learning for detecting Sybil attacks in ITS. FL-SATS employs a unique three-tier model aggregation at the Roadside Unit, Roadside Controller, and Software-Defined Network Controller, achieving high accuracy. Our results show that FL-SATS outperforms traditional methods with a detection accuracy of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 8. 7 \%}$</tex> in baseline Sybil attacks, and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 8. 5 \%}$</tex> in highdensity traffic. Moreover, the fuzzy logic-based vehicle selection mechanism optimizes localized training, further reducing detection latency to 20 ms and lowering communication overhead by up to 33% compared to centralized learning approaches. These results establish FL-SATS as a robust solution for securing vehicular communication.