One-Shot Private Confidence Regions via Resampling
Abstract
We propose a simple framework for constructing differentially private confidence regions \textit{in one shot}, i.e., by adding noise only to the final resampling quantile instead of privatizing the estimator computed on each resample. The cost of privacy of our procedure is only logarithmic in the number of resamples B under with-replacement (m-out-of-n) sampling and independent of B under without replacement sampling (subsampling), avoiding the √B factor that arises in previous works. We provide nonasymptotic Gaussian Differential Privacy (GDP) and utility guarantees for both subsampling and m-out-of-n resampling, covering mean-like estimators with small global sensitivity as well as estimators admitting efficiently computable smooth sensitivity bounds, including quantiles and degenerate U-statistics. This allows us to also obtain private confidence regions for degenerate U-statistics where the private error is much smaller than the non-private error. In all, we provide a toolbox for widely applicable DP uncertainty quantification procedures under popular resampling strategies while avoiding the computational and privacy costs of privatizing many intermediate resample statistics.
Related
- Differentially Private Confidence Intervals for Relative Risk and Odds Ratio
- Non-Asymptotic Confidence Regions for Separable Nonlinear Models
- From matrix inversion to constraints: provably tighter confidence regions for importance weights in label shift
- Active learning strategy for excursion-set confidence regions of functional simulator outputs
- Confidence regions for SAFE AI compliance scores
BibTeX
@misc{pandey2026oneshot,
title = {{One-Shot Private Confidence Regions via Resampling}},
author = {Shourya Pandey and Purnamrita Sarkar and Po-Ling Loh and Debepsita Mukherjee},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.08460},
eprint = {2610.08460},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.08460},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - One-Shot Private Confidence Regions via Resampling
AU - Shourya Pandey
AU - Purnamrita Sarkar
AU - Po-Ling Loh
AU - Debepsita Mukherjee
PY - 2026
DA - 2026/10/08
PB - arXiv
DO - 10.48550/arxiv.2610.08460
UR - https://arxiv.org/abs/2610.08460
AB - We propose a simple framework for constructing differentially private confidence regions \textit{in one shot}, i.e., by adding noise only to the final resampling quantile instead of privatizing the estimator computed on each resample. The cost of privacy of our procedure is only logarithmic in the number of resamples B under with-replacement (m-out-of-n) sampling and independent of B under without replacement sampling (subsampling), avoiding the √B factor that arises in previous works. We provide nonasymptotic Gaussian Differential Privacy (GDP) and utility guarantees for both subsampling and m-out-of-n resampling, covering mean-like estimators with small global sensitivity as well as estimators admitting efficiently computable smooth sensitivity bounds, including quantiles and degenerate U-statistics. This allows us to also obtain private confidence regions for degenerate U-statistics where the private error is much smaller than the non-private error. In all, we provide a toolbox for widely applicable DP uncertainty quantification procedures under popular resampling strategies while avoiding the computational and privacy costs of privatizing many intermediate resample statistics.
ER -