Federated Learning-Based Predictive Traffic Management Using a Contained Privacy-Preserving Scheme for Autonomous Vehicles

Intelligent Transport Systems (ITSs) are essential for secure and privacy-preserving communications in Autonomous Vehicles (AVs) and enhance facilities like connectivity and roadside assistance. Earlier research models used for traffic management compromised user privacy and exposed sensitive data to potential adversaries while handling real-time data from numerous vehicles. This research introduces a Federated Learning-based Predictive Traffic Management (FLPTM) system designed to optimize service access and privacy for Autonomous Vehicles (AVs) within an ITS. Moreover, a CPPS will provide strong security to mitigate adversarial threats through state modelling and authenticated access permissions for the integrity of vehicle communication networks from man-in-the-middle attacks. The suggested FLPTM system utilizes a Contained Privacy-Preserving Scheme (CPPS) that decentralizes data processing and allows vehicles to train local models without sharing raw data. The CPPS framework combines a classifier-based learning technique with state modelling and access permissions to protect user data against invasions and man-in-the-middle attacks. The proposed model leverages Federated Learning (FL) to enhance data security in collaborative machine learning processes by allowing updates that preserve privacy, enabling joint learning without exposing raw data. It addresses key challenges such as high communication costs, the impact of adversarial attacks, and access time inefficiencies. Using FL, the model reduces communication costs by 23.29%, mitigates adversarial effects by 16.1%, and improves access time by 18.95%, achieving significant cost savings and maintaining data privacy throughout the learning process.