LFL-COBC:Lightweight Federated Learning On Blockchain-based Device Contribution Allocation

The volume of data generated by interconnected devices in industrial distributed cyber-physical systems is rapidly increasing, presenting new opportunities for enhancing the quality of service for emerging applications through data sharing. There is an urgent need for intelligent automatic detection of device anomalies to alleviate the workload. However, privacy concerns pose a significant barrier for data providers in wireless networks. In this paper, we propose a scheme called LFL-COBC: Fusion of Blockchain and Federated Learning to address these challenges. Additionally, we analyze multidimensional factors affecting computing device performance to further enhance customer motivation and derive a cooperation strategy between clients and servers that optimizes committee selection and node device reward payments. Furthermore, we employ a CNN model for device anomaly detection and design a near-optimal model pruning algorithm to improve node device efficiency amidst increasing client participation costs and potential data explosion scenarios. Our theoretical analysis is validated through numerical evaluation using HDFS and BGL public datasets, confirming the outstanding performance of our proposed scheme.