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.
