Privacy Protection and Quantile Estimation from Noise Multiplied Data

Abstract:

Most survey organizations and statistical agencies while releasing microdata  to the public use some statistical disclosure limitation (SDL) methods to protect disclosure risk. Commonly used methods are data swapping, multiple imputation, top/bottom  code, and noise perturbation. In this talk I will discuss a method of data analysis under noise multiplication from two points of view: privacy protection from an intruder (agency) and quantile estimation (users).