Differential Privacy addresses the problem of releasing aggregate statistics of some dataset (e.g. mean) while preserving the privacy of users who contributed the data (usually involving adding noise). DP is used for example when releasing census data. There might be situations where the users do not trust the authority collecting the data. For such cases, Locally Differentially Private protocols have been developed, wherein users only ever send noisy versions of their data to the central server, instead of the server collecting the data, processing it, and then adding noise. In this talk, I will present a protocol for estimating the mean of a Gaussian variable in an LDP manner. Based on "Distributed Private Mean Estimation" by Girgis, Data, and Diggavi.
