PD2S: A Privacy-Preserving Differentiated Data Sharing Scheme based on Blockchain and Federated Learning
Yuan Liu,Peng Liu,Weipeng Jing,Houbing Herbert Song
DOI: https://doi.org/10.1109/jiot.2023.3295763
IF: 10.6
2023-01-01
IEEE Internet of Things Journal
Abstract:With the rapid development of the Internet of things(IoT), massive terminals and emerging applications bring a large amount of data. As an essential prerequisite for mining potential value from these data, data sharing is crucial and getting more and more attention. However, security and privacy concerns, and the lack of effective incentive mechanisms in the decentralized environment hinder the data owners from sharing their data. In this paper, we propose a federated learning and blockchain based privacy-preserving data sharing system, which solve the privacy leakage problem of federated learning and the scalability problem of blockchain by their technical complementarity. Considering the resource shortage and the heterogeneity of terminals, we first design a cross-layer architecture by exploiting the cloud-edge-terminal collaboration, and innovatively propose differential data sharing by dividing the providers into origin data providers and model providers. Then, we design a targeted incentive mechanism to maximize the utility of requesters and two types of data providers by formulating it into a two-stage Stackelberg game. Finally, a gradient-based algorithm is designed to get the optimal solution. Theoretical analysis and numerical simulations show that the proposed data-sharing system could eliminate privacy concerns and motivate more participants to share data. With 500 nodes, DASC is 1.72 seconds faster than SMC and 2.59 seconds faster than SDC. In terms of utility of requster, DASC is 2.0×103 higher than SDC, and 9.27 higher than SMC in average utility of providers.
computer science, information systems,telecommunications,engineering, electrical & electronic