Secure and Efficient Data Sharing for Internet of Vehicles: A Hierarchical Blockchain Enabled Asynchronous Federated Learning Approach
Hongbo Yin, Xiaoge Huang, Chengchao Liang, Bin Qian Cao, Mu Zhou · IEEE Transactions on Vehicular Technology · 2025
With the rapid development of technologies such as artificial intelligence (AI), vehicles could share their local data through the Internet of Vehicles (IoV) for better intelligent driving services. By sharing model knowledge rather than raw data, federated learning (FL) could effectively protect user data privacy. The combination of blockchain technology and FL provides a more secure sharing solution. However, the current blockchain enabled FL systems still face security threats, and have limited efficiency and scalability. In this paper, we propose a hierarchical blockchain enabled multi-region asynchronous swarm learning (HB-MASL) framework to achieve secure and efficient data sharing for IoVs. The HB-MASL divides the edge network into multiple regions to collaborate and parallelize FL-based knowledge sharing tasks to improve scalability while ensuring user privacy and data security. Empowered by the two-layer blockchain, the HB-MASL could achieve safe intra-regional and cross-regional knowledge sharing. Moreover, model compression technology and blockchain pruning technology are introduced to alleviate the communication and storage pressure of the system. In addition, we analyze the instantaneous communication rate based on vehicle mobility and derive the consensus confirmation latency of the blockchain. Based on this, a mathematical optimization problem for the total shared latency of the system is formulated, and a locally perceptual based two-step asynchronous sharing (LPTAS) algorithm is proposed to minimize the system latency. Finally, comprehensive simulation results demonstrate that the proposed algorithms could achieve better performance in terms of model accuracy, convergence rates, security and efficiency compared with existing algorithms.