A Lightweight Homomorphic Encryption Federated Learning Based on Blockchain in IoV
Dongcai Du, Wei Zhao, Linna Wei, Siyang Lu, Xuangou Wu · 2022
Artificial intelligence combined with the Internet of Vehicles (IoV) can improve the performance of automatic driving and service quality of vehicles. However, data privacy protection of IoV has become a challenging problem. In order to effectively protect user data security and reduce resource consumption, we propose a lightweight homomorphic encryption federated learning framework based on blockchain. Firstly, we combine federated learning with blockchain to train a global model collaboratively without sharing their raw data. Meanwhile, the aggregation of shared models is conducted in the on-chain nodes of the blockchain instead of a single server with traditional federated learning. Secondly, considering the further security of the sharing model on the chain, we design a lightweight homomorphic encryption approach because of the high computational cost with the existing homomorphic encryption federated learning. Finally, we conducted comparison experiments with existing homomorphic encryption federated learning schemes, and the experimental results show that our scheme can effectively protect data privacy and reduce computational costs and storage space.