Towards Auditing the Sensitive Information Leakage During Data Trading and Data Computing

Shuhao Zheng,Yanxi Lin,Yang Yu,Ye Yuan,Yongzheng Jia,Xue Liu

DOI: https://doi.org/10.48550/arxiv.2209.11451

2022-01-01

Abstract:Data markets enable users to own their data and trade their data for profits. Existing data market designs allow these data owners to directly sell or provide computation services on their private data in a secure and privacy-preserving way. However, an important problem that is neglected in previous work is that the transferred data or computation results over the data can leak information about other sensitive features that are not involved in the data market. To tackle this problem, we propose an auditing system based on smart contracts and zero-knowledge proof to measure and manage the sensitive information leakage risk in Web3 data markets. Primary results on the information leakage estimation show that even a small percent of information leakage can lead to significant predictability by simple machine learning models.