mlsanityData to 7 Oct 2026

Let's Ask Gauss: Improved One-Run Privacy Auditing

arXiv 2606.12733 ↗First posted 10 Jun 2026cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Privacy auditing provides an important safeguard by estimating the actual information leaked by a model, thus ensuring that theoretical privacy guarantees hold in practice. We study empirical privacy auditing for differentially private (DP) machine learning, focusing on efficient one-run methods for mechanisms such as DP-SGD. Prior one-run approaches threshold training examples or "canaries" into binary membership guesses, which discards useful information. We show that, in the white-box DP-SGD setting, canary-aligned signals naturally form a sequence of random variables whose normalized sum is asymptotically Gaussian. Leveraging this distributional perspective, we develop a DP-auditing framework that leads to tighter privacy lower bounds from a single training run.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 22 of its 35 references that are in the map. First posted on arXiv 10 Jun 2026.

Builds on

22 of 35 references.

References · 22

  1. 2016Deep Learning with Differential Privacycited by 2,682
  2. 2016Membership Inference Attacks against Machine Learning Modelscited by 2,221
  3. 2018The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networkscited by 847
  4. 2018Privacy Amplification by Subsampling: Tight Analyses via Couplings and Divergencescited by 237
  5. 2020Auditing Differentially Private Machine Learning: How Private is Private SGD?cited by 213
  6. 2021Adversary Instantiation: Lower Bounds for Differentially Private Machine Learningcited by 193
  7. 2021Practical and Private (Deep) Learning without Sampling or Shufflingcited by 169
  8. 2023Privacy Auditing with One (1) Training Runcited by 119
  9. 2020DP-CGAN: Differentially Private Synthetic Data and Label Generationcited by 118
  10. 2019Graphical-model based estimation and inference for differential privacycited by 73
  11. 2023Harnessing large-language models to generate private synthetic textcited by 65
  12. 2018Kymatio: Scattering Transforms in Pythoncited by 54
10 more
  1. 2023One-shot Empirical Privacy Estimation for Federated Learningcited by 35
  2. 2022Bayesian Estimation of Differential Privacycited by 31
  3. 2024Auditing f-Differential Privacy in One Runcited by 28
  4. 2022CANIFE: Crafting Canaries for Empirical Privacy Measurement in Federated Learningcited by 24
  5. 2024Nearly Tight Black-Box Auditing of Differentially Private Machine Learningcited by 19
  6. 2025Privacy Audit as Bits Transmission: (Im)possibilities for Audit by One Runcited by 10
  7. 2025How Well Can Differential Privacy Be Audited in One Run?cited by 8
  8. 2025General-Purpose f-DP Estimation and Auditing in a Black-Box Settingcited by 4
  9. 2025Enhancing One-run Privacy Auditing with Quantile Regression-Based Membership Inferencecited by 4
  10. 2025Tight Privacy Audit in One Runcited by 2

Built on by

Not cited yet.