Blockchain-Based Federated Learning for IoT Sharing: Incentive Scheme with Reputation Mechanism

Huge amounts of data produced by millions of IoT devices are expected to be shared and leveraged as the cornerstone of real-world IoT applications, such as industrial IoT, smart grid, and intelligent transportation system. However, there still exist bottlenecks when implementing IoT sharing, such as the privacy leakage of user data, IoT data quality, and incentives for sharing these data. In this paper, we propose an online incentive framework for model sharing based on blockchain and federated learning to improve the privacy protection of IoT data. To ensure the quality of submitted data, a reputation mechanism is further designed to punish the users who do not complete the model-sharing task. Based on these settings, the model sharing problem based on federated learning is formulated as an online incentive mechanism, then we use deep reinforcement learning to obtain optimal sets of sharing users with the goal of maximizing long-term social welfare. Numerical results indicate the effectiveness of the proposed framework and incentive mechanism in IoT sharing.