Blockchain-based Model for Privacy-enhanced Data Sharing

In the era of the digital economy, data has emerged as the "new era's oil," possessing significant value that necessitates refinement. Secure and efficient data sharing is the most effective approach to enhance data utilization and unlock its value. The demand for collaboration and data sharing among various organizations, businesses, and individuals is steadily increasing. However, the challenges of privacy, data security, and laws and regulations around the world make it difficult for all parties to engage in open circulation and value mining of data, resulting in data silos. The article proposes a generic model for sharing data between parties based on blockchain and privacy computing techniques, namely BPDS, to address the above challenges. Firstly, the model leverages blockchain and federated learning techniques to ensure the sharing process's traceability, immutability, and security. Additionally, a federated differential privacy algorithm is implemented to provide stringent privacy guarantees for the training data. Secondly, a gradient pruning method is devised to compress the model parameters, reducing communication overhead. Finally, the proposed model's utility, scalability, and security are evaluated using a real dataset. Experimental results demonstrate that the model achieves privacy protection while ensuring the quality of shared data, has low communication overhead, and its communication and computation overhead increases linearly with the number of participants. Furthermore, a comparative analysis is conducted to compare and analyze the proposed model with other data sharing models, confirming its security and efficiency.