RepBFL: Reputation Based Blockchain-Enabled Federated Learning Framework for Data Sharing in Internet of Vehicles

AbstractInternet of Vehicles (IoV) enables the integration of smart vehicles with Internet and collaborative analysis from shared data among vehicles. Machine learning technologies show significant advantages and efficiency for data analysis in IoV. However, the user data could be sensitive in nature, and the reliability and efficiency of sharing these data is hard to guarantee. Moreover, due to the intermittent and unreliable communications of various distributed vehicles, the traditional machine learning algorithms are not suitable for heterogeneous IoV network. In this paper, we propose a novel reputation mechanism framework that integrates the IoV with blockchain and federated learning named RepBFL. In this framework, blockchain is used to protect the shared data between the vehicles. The Road Side Units (RSU) select the high reputation vehicular nodes to share their data for federated learning. To enhance the security and reliability of the data sharing process, we develop the reputation calculated mechanism to evaluate the reliability of all vehicles in IoV. The proposed framework is feasible for the large heterogeneous vehicular networks and perform the collaborative data analysis in distributed vehicles. The experimental results show that the proposed approach can improve the data sharing efficiency. Furthermore, the reputation mechanism is able to deal with malicious behaviors effectively.KeywordsData sharingInternet of VehiclesReputation mechanismFederated learningBlockchain