Graphical-model based estimation and inference for differential privacy

TL;DR

This paper introduces a graphical-model based approach for efficiently estimating data distribution answers from noisy privacy measurements, significantly improving accuracy and scalability over existing methods.

Contribution

It presents a novel, efficient estimation technique using graphical models tailored for high-dimensional data with low-dimensional privacy measurements.

Findings

Outperforms existing privacy estimation techniques in efficiency

Enhances accuracy of privacy mechanism outputs

Scales better with high-dimensional data

Abstract

Many privacy mechanisms reveal high-level information about a data distribution through noisy measurements. It is common to use this information to estimate the answers to new queries. In this work, we provide an approach to solve this estimation problem efficiently using graphical models, which is particularly effective when the distribution is high-dimensional but the measurements are over low-dimensional marginals. We show that our approach is far more efficient than existing estimation techniques from the privacy literature and that it can improve the accuracy and scalability of many state-of-the-art mechanisms.