Public Credit Data Circulation and Sharing based on Privacy Computing

In the digital age, public credit data is crucial for financial stability and business integrity but is hindered by privacy concerns and data silos. This study proposes an innovative privacy computing model integrating differential privacy, homomorphic encryption, and federated learning to enhance data sharing while protecting privacy. Differential privacy masks data by adding noise, homomorphic encryption allows computations on encrypted data, and federated learning enables collaborative model improvement without data exchange. Experiments show this model effectively balances privacy protection and data usability. For example, varying the differential privacy parameter ε from 0.1 to 10 adjusted the trade-off between data privacy and accuracy. The findings suggest that while the model significantly improves data flow efficiency and prediction accuracy, it also maintains robust data privacy. This research advances the application of privacy-preserving technologies in public credit data management, offering a foundation for further exploration in larger and more diverse applications.