New privacy method reduces noise by handling uncommon data points
Propose-Filter-Release: Scaling Noise to Typical Sensitivity via Private Filtering
Machine Learning
Summary
When protecting data privacy, systems often add noise based on the worst-case effect of any single data record, even if that record is unusual. This results in more noise than needed for most data. The authors introduce a method called propose-filter-release (PFR) that first filters out records that don't fit typical data patterns, doing this filtering privately. This lets the system add less noise overall while still ensuring strong privacy guarantees. Their approach works broadly and can be customized for various data types and tasks.
What this means in practice
- •For data analysis teams: Privately calculate statistics with lower noise by excluding atypical records before releasing results.
- •For graph algorithm engineers: Estimate effective resistances in graphs privately with improved accuracy by filtering out sensitive edges.
Authors
Moshe Shenfeld, Eliad Tsfadia, Katrina Ligett
Abstract
An estimator can be accurate and stable on typical data, yet sensitive to a few possible records. Differentially private mechanisms commonly scale their noise to the worst-case influence of a single record, which lets these records dictate the noise even when the remaining data would allow far less. Removing them first would restore the smaller scale, but the removal must itself be private. We introduce \emph{propose-filter-release} (PFR), which, by privately filtering its input, turns a mechanism that protects the addition or removal of a typical record, one that fits the remaining data, into one that is differentially private in the standard sense, on every input. Using our novel filtering step, PFR requires only a concave friendliness score, which measures how well a record fits a dataset, and a release mechanism that is friendly-DP, that is, private when the changed record fits the remaining data. For every such score and release mechanism, it guarantees differential privacy under public conditions on the number of records, so it is a general privacy primitive, comparable to propose-test-release, and FriendlyCore can be viewed as one of its implementations. We illustrate the generality of PFR by instantiating it for scalar statistics, principal subspace estimation, the KL divergence of user histograms, the effective resistances of a graph, and treatment-rule selection.