mlsanityData to 14 Oct 2026

Delving into Differentially Private Transformer

arXiv 2405.18194 ↗First posted 28 May 2024cs.LG cs.CRcited by 5

Not on the map. Why?

Abstract (arXiv)

Deep learning with differential privacy (DP) has garnered significant attention over the past years, leading to the development of numerous methods aimed at enhancing model accuracy and training efficiency. This paper delves into the problem of training Transformer models with differential privacy. Our treatment is modular: the logic is to `reduce' the problem of training DP Transformer to the more basic problem of training DP vanilla neural nets. The latter is better understood and amenable to many model-agnostic methods. Such `reduction' is done by first identifying the hardness unique to DP Transformer training: the attention distraction phenomenon and a lack of compatibility with existing techniques for efficient gradient clipping. To deal with these two issues, we propose the Re-Attention Mechanism and Phantom Clipping, respectively. We believe that our work not only casts new light on training DP Transformers but also promotes a modular treatment to advance research in the field of differentially private deep learning.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Builds on 50 of its 71 references that are in the map. First posted on arXiv 28 May 2024.

Builds on

50 of 71 references.

References · 50

  1. 2017Attention Is All You Needcited by 60,361
  2. 2014Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translationcited by 4,057
  3. 2016Deep Learning with Differential Privacycited by 2,728
  4. 2018Self-Attentive Sequential Recommendationcited by 1,184
  5. 2018The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networkscited by 864
  6. 2015Image-based Recommendations on Styles and Substitutescited by 706
  7. 2019Does Learning Require Memorization? A Short Tale about a Long Tailcited by 460
  8. 2021Large Language Models Can Be Strong Differentially Private Learnerscited by 412
  9. 2021Differentially Private Fine-tuning of Language Modelscited by 400
  10. 2021Opacus: User-Friendly Differential Privacy Library in PyTorchcited by 383
  11. 2018Privacy Amplification by Subsampling: Tight Analyses via Couplings and Divergencescited by 242
  12. 2020Differentially Private Learning Needs Better Features (or Much More Data)cited by 235
38 more
  1. 2022Unlocking High-Accuracy Differentially Private Image Classification through Scalecited by 224
  2. 2020Tempered Sigmoid Activations for Deep Learning with Differential Privacycited by 122
  3. 2019Privacy- and Utility-Preserving Textual Analysis via Calibrated Multivariate Perturbationscited by 119
  4. 2021Large-Scale Differentially Private BERTcited by 118
  5. 2021Hyperparameter Tuning with Renyi Differential Privacycited by 104
  6. 2022Automatic Clipping: Differentially Private Deep Learning Made Easier and Strongercited by 88
  7. 2022Position: Considerations for Differentially Private Learning with Large-Scale Public Pretrainingcited by 88
  8. 2018Lightweight Probabilistic Deep Networkscited by 77
  9. 2019Sampling-free Epistemic Uncertainty Estimation Using Approximated Variance Propagationcited by 74
  10. 2020Training Production Language Models without Memorizing User Datacited by 68
  11. 2023DP-Forward: Fine-tuning and Inference on Language Models with Differential Privacy in Forward Passcited by 67
  12. 2022Differentially Private Optimization on Large Model at Small Costcited by 65
  13. 2020Locally Private Graph Neural Networkscited by 64
  14. 2015Privacy for Free: Posterior Sampling and Stochastic Gradient Monte Carlocited by 63
  15. 2022The Limits of Word Level Differential Privacycited by 63
  16. 2023Locally Differentially Private Document Generation Using Zero Shot Promptingcited by 63
  17. 2019Improving Deep Transformer with Depth-Scaled Initialization and Merged Attentioncited by 59
  18. 2016Towards Bayesian Deep Learning: A Framework and Some Existing Methodscited by 58
  19. 2020Enabling Fast Differentially Private SGD via Just-in-Time Compilation and Vectorizationcited by 58
  20. 2022Exploring the Limits of Differentially Private Deep Learning with Group-wise Clippingcited by 52
  21. 2020On Differentially Private Stochastic Convex Optimization with Heavy-tailed Datacited by 51
  22. 2020Scaling up Differentially Private Deep Learning with Fast Per-Example Gradient Clippingcited by 47
  23. 2021Improved Rates for Differentially Private Stochastic Convex Optimization with Heavy-Tailed Datacited by 47
  24. 2021Private Adaptive Gradient Methods for Convex Optimizationcited by 46
  25. 2021TEM: High Utility Metric Differential Privacy on Textcited by 43
  26. 2015Efficient Per-Example Gradient Computationscited by 42
  27. 2020Optimizing Deeper Transformers on Small Datasetscited by 42
  28. 2021High Dimensional Differentially Private Stochastic Optimization with Heavy-tailed Datacited by 42
  29. 2022Private Adaptive Optimization with Side Informationcited by 38
  30. 2022Normalized/Clipped SGD with Perturbation for Differentially Private Non-Convex Optimizationcited by 36
  31. 2016Natural-Parameter Networks: A Class of Probabilistic Neural Networkscited by 34
  32. 2023Split-and-Denoise: Protect large language model inference with local differential privacycited by 32
  33. 2021The Role of Adaptive Optimizers for Honest Private Hyperparameter Selectioncited by 25
  34. 2023ViP: A Differentially Private Foundation Model for Computer Visioncited by 17
  35. 2023Differentially Private Sharpness-Aware Trainingcited by 14
  36. 2022DPIS: An Enhanced Mechanism for Differentially Private SGD with Importance Samplingcited by 13
  37. 2022LPGNet: Link Private Graph Networks for Node Classificationcited by 12
  38. 2021DPlis: Boosting Utility of Differentially Private Deep Learning via Randomized Smoothingcited by 10

Built on by

Cited by year

Citing papers per year: 2024: 3; 2025: 2.

5 papers.

  1. 2024PrE-Text: Training Language Models on Private Federated Data in the Age of LLMscited by 29
  2. 2024Provable Differentially Private Computation of the Cross-Attention Mechanismcited by 14
  3. 2024A Theoretical Survey on Foundation Modelscited by 1
  4. 2025Optimizer Dynamics at the Edge of Stability with Differential Privacycited by 1
  5. 2025ConvXformer: Differentially Private Hybrid ConvNeXt-Transformer for Inertial Navigationcited by 0