A Decentralized Approach using Blockchain Enabled Federated Learning for Interconnected Vehicular Network

Abstract

Autonomous vehicles have made it important for vehicles to share information quickly and easily to make driving safer and make smart transportation possible. However, centralized data sharing approaches pose security and privacy risks. To address this issue, we present a decentralized approach that utilizes a hybrid model of blockchain and federated learning for interconnected vehicular networks. Our model ensures data integrity and secures the data sharing process through blockchain technology, while federated learning is employed to train machine learning models on the shared data without transferring it to a central location. We use the FedAvg algorithm for federated learning, which aggregates locally trained models from participating vehicles to produce a high-accuracy global model. We evaluate our approach on MNIST dataset and achieve an accuracy of 91.916%. Our results demonstrate the effectiveness of our proposed approach in ensuring secure data sharing and achieving high accuracy in trained models, making it a promising solution for decentralized data sharing in vehicular networks.