תאריך:
ב', 21/11/201615:30-16:30
מיקום:
חדר 4412 , חברה (חדר סמינרים)
Title: Differential Privacy of standard data dissemination schemes
Abstract: When an individual participates in a survey, and its results are published by the agency conducting the survey in some form, there is risk that the individual’s private information will be disclosed. I will describe some scenarios that demonstrate the problem. Statisticians and computer scientists came up with various measures of disclosure risk that depend on the data and the way it is published. I will describe the notion of Differential Privacy, a measure of risk proposed by computer scientists, and discuss its merits and shortcomings, and its application to standard data release schemes.
Recently, ideas related to Differential Privacy were proposed as an approach to dealing with problems arising from data snooping in data mining and multiple hypotheses testing. I will mention this very briefly, but perhaps it adds importance to the notion of Differential Privacy.
This is joint ongoing work with C. O’Keefe, C. Skinner, and N. Shlomo.