DMS Graduate Student Seminar

Speaker: Dr. Rob Molinari

Title: Perturbed M-Estimation: Ideas for Differential Privacy and Utility

Abstract: Differential Privacy (DP) provides a mathematical framework for defining a provable disclosure risk in the presence of arbitrary adversaries; it guarantees that whether an individual is in a database or not, the results of a DP procedure should be similar in terms of their probability distribution. While DP mechanisms are provably effective in protecting privacy, they often negatively impact  the utility of the query responses, statistics, and/or analyses that come as outputs from these mechanisms. To address this problem, we use ideas from the area of robust statistics which aims at reducing the influence of outlying observations on statistical inference. The resulting privacy mechanism, named "Perturbed M-Estimation," shows important potential in terms of improved statistical utility of its outputs as suggested by some preliminary results