A Blockchain-Driven Framework for Strengthening Model Ownership and Data Privacy in Federated Learning

This study presents an innovative federated learning framework that incorporates blockchain technology to enhance privacy protection and ensure model ownership. The framework addresses the issues of data privacy leakage and intellectual property threats in federated environments by embedding black-box watermarking and employing dynamic Differentially-Private Stochastic Gradient Descent (DP-SGD) for differential privacy. Additionally, a literature review approach is employed to analyze the security issues of blockchain-based distributed learning platforms, which informs the design of the framework in question and allows it to be made more scalable.