mlsanityData to 14 Oct 2026

DP-SGD vs PATE: Which Has Less Disparate Impact on Model Accuracy?

arXiv 2106.12576 ↗First posted 22 Jun 2021cs.LG cs.AI cs.CRcited by 33

Not on the map. Why?

Abstract (arXiv)

Recent advances in differentially private deep learning have demonstrated that application of differential privacy, specifically the DP-SGD algorithm, has a disparate impact on different sub-groups in the population, which leads to a significantly high drop-in model utility for sub-populations that are under-represented (minorities), compared to well-represented ones. In this work, we aim to compare PATE, another mechanism for training deep learning models using differential privacy, with DP-SGD in terms of fairness. We show that PATE does have a disparate impact too, however, it is much less severe than DP-SGD. We draw insights from this observation on what might be promising directions in achieving better fairness-privacy trade-offs.

Key facts. Cited by 33 papers in the map (data through 14 Oct 2026). Builds on 19 of its 26 references that are in the map. First posted on arXiv 22 Jun 2021.

Builds on

19 of 26 references.

References · 19

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2016Deep Learning with Differential Privacycited by 2,728
  3. 2016Membership Inference Attacks against Machine Learning Modelscited by 2,281
  4. 2019A Survey on Bias and Fairness in Machine Learningcited by 1,917
  5. 2016Semi-supervised Knowledge Transfer for Deep Learning from Private Training Datacited by 348
  6. 2019Differential Privacy Has Disparate Impact on Model Accuracycited by 316
  7. 2021Practical and Private (Deep) Learning without Sampling or Shufflingcited by 170
  8. 2019Deep Learning with Gaussian Differential Privacycited by 114
  9. 2020Revisiting Membership Inference Under Realistic Assumptionscited by 112
  10. 2018Differentially Private Fair Learningcited by 76
  11. 2020Privacy in Deep Learning: A Surveycited by 71
  12. 2020Neither Private Nor Fair: Impact of Data Imbalance on Utility and Fairness in Differential Privacycited by 61
7 more
  1. 2019Fair Decision Making using Privacy-Protected Datacited by 58
  2. 2020Differentially Private and Fair Deep Learning: A Lagrangian Dual Approachcited by 53
  3. 2020Chasing Your Long Tails: Differentially Private Prediction in Health Care Settingscited by 37
  4. 2021Decision Making with Differential Privacy under a Fairness Lenscited by 32
  5. 2017On Fairness, Diversity and Randomness in Algorithmic Decision Makingcited by 17
  6. 2020Differential Privacy of Hierarchical Census Data: An Optimization Approachcited by 9
  7. 2020Benchmarking Differentially Private Residual Networks for Medical Imagerycited by 3

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2021: 3; 2022: 9; 2023: 9; 2024: 5; 2025: 4; 2026: 3.

33 papers.

Sort:

By area:

  1. 2022Mixed Differential Privacy in Computer Visioncited by 54
  2. 2021Robin Hood and Matthew Effects: Differential Privacy Has Disparate Impact on Synthetic Datacited by 48
  3. 2022Differential Privacy and Fairness in Decisions and Learning Tasks: A Surveycited by 47
  4. 2022Pruning has a disparate impact on model accuracycited by 44
  5. 2023Privacy and Fairness in Federated Learning: on the Perspective of Trade-offcited by 26
  6. 2022Differential Privacy has Bounded Impact on Fairness in Classificationcited by 18
  7. 2023Unlocking Accuracy and Fairness in Differentially Private Image Classificationcited by 16
  8. 2022Differentially Private Model Compressioncited by 13
  9. 2024The Elusive Pursuit of Reproducing PATE-GAN: Benchmarking, Auditing, Debuggingcited by 13
  10. 2021A Fairness Analysis on Private Aggregation of Teacher Ensemblescited by 11
  11. 2023Learning with Impartiality to Walk on the Pareto Frontier of Fairness, Privacy, and Utilitycited by 10
  12. 2022SF-PATE: Scalable, Fair, and Private Aggregation of Teacher Ensemblescited by 8
21 more
  1. 2021DP-SGD vs PATE: Which Has Less Disparate Impact on GANs?cited by 7
  2. 2025Trustworthy Machine Learning via Memorization and the Granular Long-Tail: A Survey on Interactions, Tradeoffs, and Beyondcited by 6
  3. 2023On the Fairness Impacts of Private Ensembles Modelscited by 4
  4. 2023Survey on AI Ethics: A Socio-technical Perspectivecited by 4
  5. 2023A Simple and Practical Method for Reducing the Disparate Impact of Differential Privacycited by 4
  6. 2023SoK: Taming the Triangle -- On the Interplays between Fairness, Interpretability and Privacy in Machine Learningcited by 4
  7. 2022Exploring the Unfairness of DP-SGD Across Settingscited by 3
  8. 2022Privacy-Preserving Deep Learning Model for Covid-19 Disease Detectioncited by 3
  9. 2025SoK: What Makes Private Learning Unfair?cited by 3
  10. 2023Personalized DP-SGD using Sampling Mechanismscited by 2
  11. 2024Does Differential Privacy Impact Bias in Pretrained NLP Models?cited by 2
  12. 2023A Cautionary Tale: On the Role of Reference Data in Empirical Privacy Defensescited by 1
  13. 2024Privacy at a Price: Exploring its Dual Impact on AI Fairnesscited by 1
  14. 2024Assessing the Impact of Image Dataset Features on Privacy-Preserving Machine Learningcited by 1
  15. 2024The Impact of Generalization Techniques on the Interplay Among Privacy, Utility, and Fairness in Image Classificationcited by 1
  16. 2025Fairness Meets Privacy: Integrating Differential Privacy and Demographic Parity in Multi-class Classificationcited by 1
  17. 2026Where to Intervene? Benchmarking Fairness-Aware Learning on Differentially Private Synthetic Tabular Datacited by 1
  18. 2022Contextualize differential privacy in image database: a lightweight image differential privacy approach based on principle component analysis inversecited by 0
  19. 2025Towards Reliable and Generalizable Differentially Private Machine Learning (Extended Version)cited by 0
  20. 2026INO-SGD: Addressing Utility Imbalance under Individualized Differential Privacycited by 0
  21. 2026Privacy Cost as Equity Input: A Group Fairness Criterion for Differentially Private Machine Learningcited by 0