mlsanityData to 14 Oct 2026

MCMC for Bayesian estimation of Differential Privacy from Membership Inference Attacks

arXiv 2504.16683 ↗First posted 23 Apr 2025cs.LG stat.MLcited by 1

On the map: LLM safety and security → Membership inference attacks

Abstract (arXiv)

We propose a new framework for Bayesian estimation of differential privacy, incorporating evidence from multiple membership inference attacks (MIA). Bayesian estimation is carried out via a Markov chain Monte Carlo (MCMC) algorithm, named MCMC-DP-Est, which provides an estimate of the full posterior distribution of the privacy parameter (e.g., instead of just credible intervals). Critically, the proposed method does not assume that privacy auditing is performed with the most powerful attack on the worst-case (dataset, challenge point) pair, which is typically unrealistic. Instead, MCMC-DP-Est jointly estimates the strengths of MIAs used and the privacy of the training algorithm, yielding a more cautious privacy analysis. We also present an economical way to generate measurements for the performance of an MIA that is to be used by the MCMC method to estimate privacy. We present the use of the methods with numerical examples with both artificial and real data.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Membership inference attacks, one of 12 directions in LLM safety and security. Builds on 18 of its 26 references that are in the map. First posted on arXiv 23 Apr 2025.

Builds on

18 of 26 references.

Closest ancestors · 3

Same direction, newest first

  1. 2023Gaussian Membership Inference Privacy12 Jun
  2. 2021Membership Inference Attacks From First Principles7 Dec
  3. 2021Enhanced Membership Inference Attacks against Machine Learning Models18 Nov

Other references · 15

  1. 2016Membership Inference Attacks against Machine Learning Modelscited by 2,281
  2. 2017Renyi Differential Privacycited by 650
  3. 2016Concentrated Differential Privacy: Simplifications, Extensions, and Lower Boundscited by 346
  4. 2013The Composition Theorem for Differential Privacycited by 297
  5. 2019White-box vs Black-box: Bayes Optimal Strategies for Membership Inferencecited by 271
  6. 2018Privacy Amplification by Subsampling: Tight Analyses via Couplings and Divergencescited by 242
  7. 2020Auditing Differentially Private Machine Learning: How Private is Private SGD?cited by 218
  8. 2021Adversary Instantiation: Lower Bounds for Differentially Private Machine Learningcited by 199
  9. 2023Privacy Auditing with One (1) Training Runcited by 123
  10. 2023One-shot Empirical Privacy Estimation for Federated Learningcited by 36
  11. 2022A General Framework for Auditing Differentially Private Machine Learningcited by 25
  12. 2022CANIFE: Crafting Canaries for Empirical Privacy Measurement in Federated Learningcited by 24
3 more
  1. 2023Unleashing the Power of Randomization in Auditing Differentially Private MLcited by 20
  2. 2024The Last Iterate Advantage: Empirical Auditing and Principled Heuristic Analysis of Differentially Private SGDcited by 11
  3. 2020Metropolis-Hastings with Averaged Acceptance Ratioscited by 6

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Bayesian Membership Privacy for Graph Neural Networkscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2022Bayesian Estimation of Differential Privacy
  2. 2025Membership Inference Attacks as Privacy Tools: Reliability, Disparity and Ensemble
  3. 2024Lost in the Averages: A New Specific Setup to Evaluate Membership Inference Attacks Against Machine Learning Models
  4. 2025Synth-MIA: A Testbed for Auditing Privacy Leakage in Tabular Data Synthesis
  5. 2026A Bayesian Approach to Membership Inference for Statistical Release
  6. 2025The DCR Delusion: Measuring the Privacy Risk of Synthetic Data
  7. 2026Membership Inference via Pairwise Likelihood Ratios
  8. 2025Bayesian Inference of Training Dataset Membership