mlsanityData to 7 Oct 2026

Tight Differential Privacy for Discrete-Valued Mechanisms and for the Subsampled Gaussian Mechanism Using FFT

arXiv 2006.07134 ↗First posted 12 Jun 2020stat.ML cs.CR cs.LGcited by 43

Not on the map. Why?

Abstract (arXiv)

We propose a numerical accountant for evaluating the tight (varepsilon,δ)-privacy loss for algorithms with discrete one dimensional output. The method is based on the privacy loss distribution formalism and it uses the recently introduced fast Fourier transform based accounting technique. We carry out an error analysis of the method in terms of moment bounds of the privacy loss distribution which leads to rigorous lower and upper bounds for the true (varepsilon,δ)-values. As an application, we present a novel approach to accurate privacy accounting of the subsampled Gaussian mechanism. This completes the previously proposed analysis by giving strict lower and upper bounds for the privacy parameters. We demonstrate the performance of the accountant on the binomial mechanism and show that our approach allows decreasing noise variance up to 75 percent at equal privacy compared to existing bounds in the literature. We also illustrate how to compute tight bounds for the exponential mechanism applied to counting queries.

Key facts. Cited by 43 papers in the map (data through 7 Oct 2026). First posted on arXiv 12 Jun 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 7; 2022: 12; 2023: 6; 2024: 11; 2025: 3; 2026: 4.

43 papers.

Sort:

By area:

  1. 2021Differentially Private Fine-tuning of Language Modelscited by 392
  2. 2021Numerical Composition of Differential Privacycited by 174
  3. 2021Optimal Accounting of Differential Privacy via Characteristic Functioncited by 83
  4. 2022Improved Differential Privacy for SGD via Optimal Private Linear Operators on Adaptive Streamscited by 59
  5. 2022Connect the Dots: Tighter Discrete Approximations of Privacy Loss Distributionscited by 50
  6. 2022Composition of Differential Privacy & Privacy Amplification by Subsamplingcited by 42
  7. 2024Private Fine-tuning of Large Language Models with Zeroth-order Optimizationcited by 41
  8. 2022Bayesian Estimation of Differential Privacycited by 31
  9. 2021Fast and Memory Efficient Differentially Private-SGD via JL Projectionscited by 29
  10. 2024Fine-Tuning Large Language Models with User-Level Differential Privacycited by 28
  11. 2024Mind the Privacy Unit! User-Level Differential Privacy for Language Model Fine-Tuningcited by 21
  12. 2021Computing Differential Privacy Guarantees for Heterogeneous Compositions Using FFTcited by 20
31 more
  1. 2022Individual Privacy Accounting with Gaussian Differential Privacycited by 19
  2. 2023Gradient Descent with Linearly Correlated Noise: Theory and Applications to Differential Privacycited by 18
  3. 2023Privacy-Aware Document Visual Question Answeringcited by 17
  4. 2022Faster Privacy Accounting via Evolving Discretizationcited by 14
  5. 2024Avoiding Pitfalls for Privacy Accounting of Subsampled Mechanisms under Compositioncited by 12
  6. 2023A Randomized Approach for Tight Privacy Accountingcited by 11
  7. 2023Improving the Privacy and Practicality of Objective Perturbation for Differentially Private Linear Learnerscited by 11
  8. 2024Subsampling is not Magic: Why Large Batch Sizes Work for Differentially Private Stochastic Optimisationcited by 11
  9. 2022Skellam Mixture Mechanism: a Novel Approach to Federated Learning with Differential Privacycited by 7
  10. 2024Auditing Differential Privacy Guarantees Using Density Estimationcited by 6
  11. 2025Gaussian DP for Reporting Differential Privacy Guarantees in Machine Learningcited by 6
  12. 2021Tight Accounting in the Shuffle Model of Differential Privacycited by 5
  13. 2022Group privacy for personalized federated learningcited by 5
  14. 2022The Saddle-Point Accountant for Differential Privacycited by 5
  15. 2022Cactus Mechanisms: Optimal Differential Privacy Mechanisms in the Large-Composition Regimecited by 3
  16. 2023Breaking the Communication-Privacy-Accuracy Tradeoff with f-Differential Privacycited by 3
  17. 2026Efficient privacy loss accounting for subsampling and random allocationcited by 3
  18. 2021D3p -- A Python Package for Differentially-Private Probabilistic Programmingcited by 2
  19. 2022DPVIm: Differentially Private Variational Inference Improvedcited by 2
  20. 2024Tight Group-Level DP Guarantees for DP-SGD with Sampling via Mixture of Gaussians Mechanismscited by 2
  21. 2024Privacy Profiles for Private Selectioncited by 2
  22. 2024Closed-Form Bounds for DP-SGD against Record-level Inferencecited by 2
  23. 2024Convex Approximation of Two-Layer ReLU Networks for Hidden State Differential Privacycited by 2
  24. 2025Learning Rate Scheduling with Matrix Factorization for Private Trainingcited by 2
  25. 2022Protecting Data from all Parties: Combining FHE and DP in Federated Learningcited by 0
  26. 2023Enhancing Trade-offs in Privacy, Utility, and Computational Efficiency through MUltistage Sampling Technique (MUST)cited by 0
  27. 2024Weights Shuffling for Improving DPSGD in Transformer-based Modelscited by 0
  28. 2025Towards Reliable and Generalizable Differentially Private Machine Learning (Extended Version)cited by 0
  29. 2026Differentially Private Model Mergingcited by 0
  30. 2026Provable Robustness against Backdoor Attacks via the Primal-Dual Perspective on Differential Privacycited by 0
  31. 2026A Sieve-Accelerated Quadrature Method for Exact Privacy Accounting in the 2020 U.S. Decennial Censuscited by 0