Fundamental Technologies for Machine Learning Centric Data Trading

2021 Fiscal Year Annual Research Report

• Project/Area Number: 21J23090
• Allocation Type: Single-year Grants
• Research Institution: Kyoto University
• Principal Investigator: 鄭 舒元 京都大学, 情報学研究科, 特別研究員(DC1)
• Project Period (FY): 2021-04-28 – 2024-03-31
• Keywords: personal data market / privacy protection / data valuation

Outline of Annual Research Achievements

We have made several technological advances in promoting privacy-preserving, trustworthy, and user-friendly data trading. First, we extended our previous work and proposed a more general pricing method that can prevent data buyers' arbitrage behaviors. Second, we established a novel marketplace for selling training data for machine learning (ML) tasks. In this marketplace, data buyers can purchase utility-optimal ML models with ease according to their task-specific data demands, and data sellers can flexibly control the degree of their privacy leakage. Third, for this marketplace, we further initiated a study on how to fairly, efficiently, and securely evaluate the sellers' contribution to training the models and have obtained some insightful results.

Current Status of Research Progress

2: Research has progressed on the whole more than it was originally planned.

Reason

The planned project has been proceeding smoothly. We have submitted several papers on our research achievements to prestigious international conferences and journals.

Strategy for Future Research Activity

To incentivize data sellers to contribute valuable data and thus promote data circulation, we will continue investigating novel methods for fair, efficient, and privacy-preserving contribution evaluation in various scenarios of data trading.