Efficient, Differentially Private Point Estimators

TL;DR

This paper demonstrates that for many parametric models, it is possible to construct differentially private estimators that are both efficient and asymptotically unbiased, aligning privacy with statistical validity.

Contribution

The paper introduces a method to create differentially private estimators that converge to maximum likelihood estimators, ensuring efficiency and unbiasedness.

Findings

Private estimators converge to MLE

Estimators are efficient and asymptotically unbiased

Privacy guarantees do not compromise statistical validity

Abstract

Differential privacy is a recent notion of privacy for statistical databases that provides rigorous, meaningful confidentiality guarantees, even in the presence of an attacker with access to arbitrary side information. We show that for a large class of parametric probability models, one can construct a differentially private estimator whose distribution converges to that of the maximum likelihood estimator. In particular, it is efficient and asymptotically unbiased. This result provides (further) compelling evidence that rigorous notions of privacy in statistical databases can be consistent with statistically valid inference.