Blockchain and distributed digital watermarking effort on federated learning: innovating intellectual property protection

Federated Learning with Digital Watermarks (FLDW) have been recognized as a promising solution for property protection. However, the existing FLDW-related technologies neglect the requirements of decentralized settings, leading to recurrent issues such as discrepancies in distributed client data. This paper introduces a Blockchain Federated Learning Intellectual Property Protection Framework (BFLIPR), to address the data security and model validation challenges in decentralized federated learning environments. BFLIPR merges blockchain, digital watermarking, and federated learning technologies. By harnessing the blockchain’s tamper-proof properties, digital watermarking’s concealment capabilities, and federated learning’s distributed feature, the framework offers a solution that aligns with intellectual property protection mechanism, to bolster data security and property safeguarding. Experimental findings demonstrate its high feasibility and robust for data privacy and model security in the federated learning.