Secure Aggregation Scheme for Federated Learning with Bilateral Verification in the Internet of Vehicles

In the Internet of Vehicles (IoV), Federated Learning (FL) allows vehicles to train machine learning models collaboratively, protecting data privacy and enhancing overall model performance. However, FL in IoV faces challenges such as model leakage, difficulty in verifying training results, and high computational costs. To address these challenges, this paper presents a secure aggregation scheme for FL that supports bilateral verification. First, clients mask their local gradients, ensuring that the aggregation server learns nothing other than the final aggregated result. Second, we introduce a bilateral verification mechanism to ensure that the aggregation server can perform batch authentication of client identities and that each client can independently verify the correctness of the aggregated output. In addition, we adopt a dynamic group management mechanism to tolerate users’ dropping out with no impact on their participation in future learning process. The simulation results show that, compared to the existing methods, the proposed scheme maintains high efficiency while supporting bilateral verification.