NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:85
Title:Average-Case Averages: Private Algorithms for Smooth Sensitivity and Mean Estimation

This paper re-visits the smooth sensitivity framework from the point of view of concentrated DP and proposes three new perturbation methods that work under this framework. The results are applied to the problem of estimating the mean of distributions with unbounded support from iid observations, and provide finite-sample bounds on the accuracy. This is a nice fundamental innovation in the theory of DP. When preparing a final version of the manuscript the authors are strongly encouraged to include a discussion about potential further applications of the new mechanisms beyond mean estimation.