PrivateSQL: Reimagining and designing a new differentially private SQL query engine
While differential privacy is considered the dominant standard for private data analysis, much of the research literature applies only to flat tables, ignoring the complexity of multi-relational schemas and constraints.
Research
Summary:
While differential privacy is considered the dominant standard for private data analysis, much of the research literature applies only to flat tables, ignoring the complexity of multi-relational schemas and constraints. This work presents PrivateSQL, a first-of-its-kind end-to-end differentially private relational database system. We demonstrate that our system can offer error rates 1,000 times lower than competing approaches to answering SQL queries under differential privacy.