Calibrating Data to Sensitivity in Private Data Analysis

TL;DR

This paper introduces a novel approach to differential privacy that reduces the contribution of challenging data records rather than increasing noise, improving accuracy in private data analysis.

Contribution

It presents wPINQ, a generalized platform that enables non-uniform data scaling for enhanced privacy-utility trade-offs and introduces new techniques for complex data measurements.

Findings

Non-uniform data scaling improves accuracy over uniform noise addition.

wPINQ can replicate and enhance recent graph analysis results.

Probabilistic inference enables synthesis of datasets with complex measurements.

Abstract

We present an approach to differentially private computation in which one does not scale up the magnitude of noise for challenging queries, but rather scales down the contributions of challenging records. While scaling down all records uniformly is equivalent to scaling up the noise magnitude, we show that scaling records non-uniformly can result in substantially higher accuracy by bypassing the worst-case requirements of differential privacy for the noise magnitudes. This paper details the data analysis platform wPINQ, which generalizes the Privacy Integrated Query (PINQ) to weighted datasets. Using a few simple operators (including a non-uniformly scaling Join operator) wPINQ can reproduce (and improve) several recent results on graph analysis and introduce new generalizations (e.g., counting triangles with given degrees). We also show how to integrate probabilistic inference techniques to synthesize datasets respecting more complicated (and less easily interpreted) measurements.