Deploying differential privacy for the 2020 census of population and housing

The privacy risks associated with the publication of official statistics have increased significantly over the last decade, fueled by the proliferation of detailed, third-party data sources and technological advances that make re-identification of individuals in public data releases increasingly easy. This presentation will discuss the U.S. Census Bureau’s research into these emerging privacy threats, and the agency’s efforts to modernize its disclosure avoidance methods using differential privacy to ensure the confidentiality of individuals’ data. Differential privacy offers significant advantages over traditional disclosure avoidance methods for safeguarding individuals’ privacy, but the implementation of differential privacy at scale for the 2020 Decennial Census has also posed a number of challenges. This presentation will explore these challenges and discuss the lessons learned from this initiative.

* Note: the presentation started with a vote to decide which slide-deck to cover: "Differential Privacy at the US Census Bureau: Status Report" (previously advertised title) or "Deploying differential privacy for the 2020 census of population and housing" (this was selected).