UMass College of Information and Computer Sciences (CICS) professors Gerome Miklau and Daniel Sheldon have received a $75,000 gift from Oracle Labs to support their work on tunable, differentially private synthetic data. Synthetic data is created by generating artificial records in a way that preserves important properties of the original data. This allows analytics and machine learning models to be applied while protecting the original data. However, subtle privacy violations can still occur unless a formal privacy standard, such as differential privacy, is adopted. A dataset is considered differentially private if it describes group patterns but still protects information about individuals in the dataset.
For example, health researchers may wish to analyze the number of times a woman in her 40s with diabetes visited the hospital, but can not disclose the medical history or identity of the specific patient. Differential privacy guarantees the anonymity of individuals, while still allowing for meaningful analysis. Miklau and Sheldon’s research will develop novel methods for synthetic data generation that meet the standard of ε-differential privacy, that is, the ability to tune the parameter ε to balance privacy and accuracy.
Miklau and Sheldon are part of the EQUATE initiative from CICS. EQUATE (Equity, Accountability, Trust, and Explainability) comprises CICS faculty who are engaged in research and education related to equitable algorithms and systems. For more information, visit the EQUATE website (equate.cs.umass.edu).