Everyone is welcome to attend the on-line N&O seminar with Marek Elias. Title of his lecture: Differentially Private Release of Synthetic Graphs.
Abstract:
Existence or non-existence of an edge in a graph sometimes represents a sensitive information (e.g. a friendship in some social network). To allow differentially private analysis of a graph whose edges represent such sensitive information, we would like to release it in a form of so called synthetic graph which does not reveal the private information but still approximates some features of the original private graph.
In this talk, I will present an algorithm based on optimization techniques for release of a synthetic graph which approximates cut sizes of the original private graph. In the beginning, I will also provide a short introduction to Differential Privacy. This is a joint work with Michael Kapralov, Janardhan Kulkarni, and Yin Tat Lee.
Please contact Daniel Dadush for the zoom link.