Secure Federated Learning-Based Privacy Preserving in V2V Communications

Vehicular Ad-Hoc Networks (VANETs) have emerged as a promising technology to improve road safety, traffic efficiency, and provide various intelligent transportation services. However, the widespread deployment of VANETs raises concerns about data privacy and security, as it contains sensitive data about vehicles and their drivers. Traditional approaches for sharing and analysing VANET data centrally could compromise individual privacy and expose users to various security risks such as position falsification, man-in-the middle attacks and data poisoning attacks. To address these privacy challenges, a secure Federated Learning (FL)-based privacy preserving in V2V communications is proposed. FL is a decentralized machine learning that allows multiple vehicles to collectively build a global model while storing the data on their devices. It facilitates model training without sharing raw data, thereby mitigating privacy risks and ensuring confidentiality. To ensure data integrity and security during the FL process, cryptographic technique is incorporated to detect and mitigate the model poisoning and data poisoning attacks. By leveraging cryptographic techniques, and applying the enhanced federated averaging, the proposed scheme attains improved performance in terms of accuracy, precision and recall over the existing models while promoting collaboration and knowledge sharing in intelligent vehicular networks.