Sofya Raskhodnikova
Local Model for Differentially Private Data Analysis

The goal in private data analysis is to compute accurate global
information about a dataset while preserving privacy of individuals
whose information it contains. Differential privacy places a
restriction on the algorithms that perform computational tasks on
sensitive datasets that, roughly, requires that the distribution on
outputs are close for any two datasets that differ in data of only one
person. In this talk, we will discuss the local model of diffrential
privacy, where each participant in the computation holds their own
data and only releases it via a differentially private algorithm.