Semi Differential Privacy
Primary Investigator:
Jordan Awan
Young Hyun Cho, Jordan Awan
Abstract
Differential privacy (DP) is the state-of-the-art framework for formal privacy protection and has become the gold standard for rigorous privacy guarantees. However, it is not always implemented exactly especially when the exact statistics is published along with the DP mechanism. For example, US Census publishes a combination of DP and exact statistics. There is currently no satisfactory method of quantifying the privacy in such settings. In this regard, we propose a framework semi-DP, which properly accounts for the combination of private and non-private releases. Under semi-DP we derive optimal mechanisms under a variety of scenarios.