||November 5th, 2019
||Kelly Engineering Center, Room 1003
||Event is free and open to the public
Topic: Differentially private hypothesis testing
Imagine a new paradigm where statisticians must perform analyses without access to the actual data. This is the paradigm established by differential privacy, a criterion that emerged from the field of cryptography to bound the amount of information released by a query on a data set. This technique has already seen adoption within industry and is beginning to spread to the sciences, where it is running into statistical methods that assume the analyst has full access to the data set. This talk will provide an introduction to the foundations of differential privacy and discuss the implications for statistics. It will also share current research to adapt classical inferential techniques to the framework of differential privacy.