The Blockchain-Based Edge Computing Framework for Privacy-Preserving Federated Learning

Abstract

Nowadays, privacy-preserving artificial intelligence is gaining traction, with the goal of learning multiple models based on private data without leaking any personal information. Since the existing multi-party computation methods and other encryption-based methods have their flaws, we developed our own blockchain-based edge computing framework to achieve the decentralization and enhance the efficiency. Our framework enables a trustful, simplified and asynchronous federated learning in IoT and provides a convenient and secret classification service. Extensive evaluations on efficiency are provided, confirming the performance of our solutions.