SFedChain: blockchain-based federated learning scheme for secure data sharing in distributed energy storage networks

• Published in: PeerJ. Computer science Vol. 8; p. e1027
• Main Authors: Meng, Mingming, Li, Yuancheng
• Format: Journal Article
• Language: English
• Published: San Diego PeerJ. Ltd 29.06.2022; PeerJ, Inc; PeerJ Inc
• Subjects: Access control; Accuracy; Algorithms; Artificial Intelligence; Big Data; Blockchain; Clustering; Collaboration; Computer storage devices; Cryptography; Data integrity; Data retrieval; Data security; Data sharing; Datasets; Deep learning; Distributed and Parallel Computing; Economic incentives; Electronic devices; Energy management systems; Energy storage; Federated learning; Information sharing; Information storage and retrieval; Intelligence; Internet of Things; Investment analysis; Machine learning; Mathematical problems; Natural Language and Speech; Network security; Networks; Performance evaluation; Privacy; Safety and security measures; Security; Smart contract; Storage area networks; Training
• ISSN: 2376-5992; 2376-5992
• DOI: 10.7717/peerj-cs.1027

The intelligence of energy storage devices has led to a sharp increase in the amount of detection data generated. Data sharing among distributed energy storage networks can realize collaborative control and comprehensive analysis, which effectively improves the clustering and intelligence. However, data security problems have become the main obstacle for energy storage devices to share data for joint modeling and analysis. The security issues caused by information leakage far outweigh property losses. In this article, we first proposed a blockchain-based machine learning scheme for secure data sharing in distributed energy storage networks. Then, we formulated the data sharing problem into a machine-learning problem by incorporating secure federated learning. Innovative verification methods and consensus mechanisms were used to encourage participants to act honestly, and to use well-designed incentive mechanisms to ensure the sustainable and stable operation of the system. We implemented the scheme of SFedChain and experimented on real datasets with different settings. The numerical results show that SFedChain is promising.