Federated Learning-Based Architecture for Detecting Cyber Vehicular Attackers in Intelligent Transportation Systems

Abstract

The Internet of Vehicles (IoV) concept within Intelligent Transportation Systems (ITS) integrates automobiles, transportation, information sharing, and traffic infrastructure management to enhance road safety. IOV can collaborate to create models through federated learning, improving performance, enhancing data privacy, and ensuring the security of local vehicle data. This paper introduces a novel Distillation-based Semi supervised (DS-FL) Model for intrusion detection. This model was demonstrated using the datasets to address the heterogeneity and diversity of devices and malicious samples. Experimental results show that the proposed system achieved 99.48%, 99.75%, 99.83%, and 99.93% accuracy in detecting various types of attacks on the ISCXIDS2012, CIC-IDS2017, CSE-CIC-IDS2018, and Car-Hacking datasets, respectively, outperforming other intrusion detection techniques. It highlights the model's effectiveness in securing intelligent transportation system networks against cyber-attacks.