IncEFL: a sharing incentive mechanism for edge-assisted federated learning in industrial IoT

As the information sensing and processing capabilities of IoT devices increase, a large amount of data is being generated at the edge of Industrial IoT (IIoT), which has become a strong foundation for distributed Artificial Intelligence (AI) applications. However, most users are reluctant to disclose their data due to network bandwidth limitations, device energy consumption, and privacy requirements. To address this issue, this paper introduces an Edge-assisted Federated Learning (EFL) framework, along with an incentive mechanism for lightweight industrial data sharing. In order to reduce the information asymmetry between data owners and users, an EFL model-sharing incentive mechanism based on contract theory is designed. In addition, a weight dispersion evaluation scheme based on Wasserstein distance is proposed. This study models an optimization problem of node selection and sharing incentives to maximize the EFL model consumers' profit and ensure the quality of training services. An incentive-based EFL algorithm with individual rationality and incentive compatibility constraints is proposed. Finally, the experimental results verify the effectiveness of the proposed scheme in terms of positive incentives for contract design and performance analysis of EFL systems.