Abstract
Privacy-preserving data analysis has a large literature spanning more than five decades and multiple disciplines. “Differential privacy” has provided a theoretically sound and powerful framework, and given rise to an explosion of research. Differential privacy composes, allowing the construction of complex differentially private analyses from simple private building blocks. Beginning with basic definitions and a handful of commonly used primitives, we will spend some time on each of three areas: a description of a prototype, recent theoretical advances, and fundamental limitations on accuracy consistent with (any kind of) privacy – that is, the cards we have been dealt. We will end with a glimpse of applications of differential privacy in the service of goals other than private data analysis.
The second session of this talk will take place on Friday, May 22 from 3:30 pm – 4:30 pm.