A Blockchain-based Privacy Protection Scheme for V2G Interaction

V2G interaction plays a crucial role in power peak and valley scheduling. Federated learning, as a privacy protection technology, is widely used in the V2G interaction process. However, issues such as the risk of local data leakage during local training, vulnerability of model gradients to reconstruction attacks during aggregation, and instability of central servers remain prevalent. To address these concerns, this paper proposes a blockchain-based privacy protection federated learning model for V2G interaction. By deploying the model on a blockchain, decentralization is achieved. Homomorphic encryption is employed during local training to protect users' temporal privacy without compromising the accuracy of local model training. During the model aggregation phase, a lightweight secure aggregation protocol is used to obscure the true gradient values of local models, ensuring model robustness and fault tolerance.