Secure Data Sharing Method Based on Federated Learning in IoVs
Shujia Fan, Jia-Wen Xue · 2024
In the realm of the Internet of Vehicles (IoVs), a broad spectrum of information encompassing vehicle operation, traffic conditions, and incidents can be exchanged between vehicles and infrastructure. This data plays a pivotal role in enhancing decision-making processes and holds considerable importance for the advancement of autonomous driving and traffic management. Utilizing machine learning (ML) methods, vehicles can assimilate insights from road and environmental data, subsequently sharing trained model parameters. Addressing concerns related to data security and user privacy during the data-sharing process in IoVs, a framework called DBL is introduced. This framework leverages blockchain and federated learning to validate local model updates. It facilitates the updating and exchange of local ML models among multiple participants, with the option to upload these models to Road Side Units (RSUs). Subsequently, RSUs scrutinize the local ML models, identifying and discarding any malicious ones while aggregating the legitimate ones to generate the global ML model. Simulation results demonstrate that the proposed DBL framework not only safeguards user data privacy but also enhances model accuracy.