ProxyGuard: Direct Reliability Inference for Randomized Data Release Mechanisms with Shared Targets
Abstract
Researchers often choose a proxy dataset from many releases, transformations, or seeds. Search can make an invalid release appear adequate, while one adequate release does not establish that its generator is reliable. ProxyGuard controls both errors using prespecified bounded risks and a sealed target set. Named-release mode corrects for multiplicity and certifies specific releases. Direct shared-target mode evaluates independent mechanism draws on a common target, lower-bounds their favorable-score rate, and subtracts a bound on favorable scores contributed by invalid releases. Conditional on the target, release scores are independent, yielding a finite-sample mechanism-reliability guarantee without independent target batches or assumptions on release-level p-value dependence. We show that the mean-only penalty is sharp and derive a smooth-score certificate with additive target concentration. In a registered three-requirement study, direct mode raises power from 5.6% to 64.2% at reliability $0.95$, while named mode remains stronger under high-signal evidence. Prospective audits span full-pipeline Rice--TVAE, which retrains on every draw, and a non-tabular text mechanism.
Related
- Inference for Stress–Strength Reliability Under Unified Hybrid Censoring: A One-Parameter Model with Applications
- Reliability inference for semi-Markov models based on multiple trajectories
- Robust Inference for Stress–Strength Reliability of the Garhy Distribution Under Diverse Record Schemes with Engineering and Medical Applications
- Automated, Efficient, and Model-Free Inference for Randomized Clinical Trials via Data-Driven Covariate Adjustment
- Inference after selection on randomized p-values
BibTeX
@misc{mahato2026proxyguard,
title = {{ProxyGuard: Direct Reliability Inference for Randomized Data Release Mechanisms with Shared Targets}},
author = {Dipesh Tharu Mahato and Pramod Dhungana},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2608.18643},
eprint = {2608.18643},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.18643},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - ProxyGuard: Direct Reliability Inference for Randomized Data Release Mechanisms with Shared Targets
AU - Dipesh Tharu Mahato
AU - Pramod Dhungana
PY - 2026
DA - 2026/10/27
PB - arXiv
DO - 10.48550/arxiv.2608.18643
UR - https://arxiv.org/abs/2608.18643
AB - Researchers often choose a proxy dataset from many releases, transformations, or seeds. Search can make an invalid release appear adequate, while one adequate release does not establish that its generator is reliable. ProxyGuard controls both errors using prespecified bounded risks and a sealed target set. Named-release mode corrects for multiplicity and certifies specific releases. Direct shared-target mode evaluates independent mechanism draws on a common target, lower-bounds their favorable-score rate, and subtracts a bound on favorable scores contributed by invalid releases. Conditional on the target, release scores are independent, yielding a finite-sample mechanism-reliability guarantee without independent target batches or assumptions on release-level p-value dependence. We show that the mean-only penalty is sharp and derive a smooth-score certificate with additive target concentration. In a registered three-requirement study, direct mode raises power from 5.6% to 64.2% at reliability $0.95$, while named mode remains stronger under high-signal evidence. Prospective audits span full-pipeline Rice--TVAE, which retrains on every draw, and a non-tabular text mechanism.
ER -