VFLChain: Blockchain-enabled Vertical Federated Learning for Edge Network Data Sharing

Abstract

With the widespread use of Internet of things(IoT), a large amount of data will be generated in the edge of network, which can facilitate a significant transformation in edge intelligent services by integrating with edge computing, 5G and artificial intelligence. However, since the intelligent edge services seriously rely on big data and computing resource, it challenges the traditional centralized data processing model. Data sharing is a promising way to tackle this problem, but some critical technical challenges still remain, such as fragile data privacy protection, inefficient data exchange and low quality of data fusion. To address these problems, a privacy-enhanced and intelligence-preserved data sharing system, name VFLChain, is proposed in this article. The proposed VFLChain is designed based on consortium blockchain and vertical federated learning, which can ensure trustworthy and secure data sharing without relying on any center platforms or third parties. Furthermore, a blockchain-assisted decentralized vertical federated learning is presented to adapt to the decentralized system and support privacy-preserved, intelligent and efficient edge data sharing, while improving quality of data through learning with different characteristic data samples. Then, a data sharing processing workflow in VFLChain is also described to demonstrated details of data sharing. The simulation experiments confirm that the proposed system and mechanism have good accuracy and stability, and guarantee an effective data sharing.