LEBFL: Lightweight authentication and efficient consensus for blockchained federated learning in vehicle–road cooperation systems with AIoT

Artificial Intelligence Internet of Things (AIoT) technology is gradually overcoming challenges related to traffic data transmission and processing in vehicle-road cooperative systems. However, the dynamism and openness of the vehicle-road cooperative networks make it susceptible to potential attacks, where attackers might intercept or tamper with transmitted local model parameters, thereby compromising the integrity of the models and leaking user privacy. Although existing solutions such as differential privacy and encryption can address these issues, they may reduce data availability or increase computational complexity. To tackle these challenges, we propose a lightweight authentication and efficient consensus for blockchained federated learning in vehicle–road cooperation systems(LEBFL), which provides lightweight privacy-enhanced authentication and efficient consensus while ensuring the privacy of local models and datasets. Specifically, we first introduce a blockchain-based federated learning architecture that enhances privacy and efficient consensus, utilizing the consortium blockchain to replace the centralized server. Subsequently, we design a lightweight anonymous authentication and key agreement protocol using efficient cryptographic primitives to establish secure session keys for the transmission of local models. Furthermore, we propose a utility-based Raft consensus algorithm, which selects the optimal fog server as the leader node using a resource matrix and weight vector, and enhances the performance of the blockchain network by leveraging the idle computing resources of fog servers. Security analysis and experimental results confirm that the proposed scheme shows superior performance without sacrificing security.