Title:PRECISE: PRivacy-loss-Efficient and Consistent Inference based on poSterior quantilEs

Abstract:Differential privacy (DP) is a mathematical framework for releasing information with formal privacy guarantees. Despite the existence of various DP procedures for performing a wide range of statistical analysis and machine learning tasks, methods of good utility are still lacking in valid statistical inference with DP guarantees. We address this gap by introducing the notion of valid Privacy-Preserving Interval Estimation (PPIE) and proposing PRivacy-loss-Efficient and Consistent Inference based on poSterior quantilEs (PRECISE). PRECISE is a general-purpose Bayesian approach for constructing privacy-preserving posterior intervals. We establish the Mean-Squared-Error (MSE) consistency for our proposed private posterior quantiles converging to the population posterior quantile as sample size or privacy loss increases. We conduct extensive experiments to compare the utilities of PRECISE with common existing privacy-preserving inferential approaches in various inferential tasks, data types and sizes,and privacy loss levels. The results demonstrated a significant advantage of PRECISE with its nominal coverage and substantially narrower intervals than the existing methods, which are prone to either under-coverage or impractically wide intervals.