mlsanityData to 7 Oct 2026

Opacus: User-Friendly Differential Privacy Library in PyTorch

arXiv 2109.12298 ↗First posted 25 Sep 2021cs.LG cs.CRcited by 375

Not on the map. Why?

Cited by 22 in LLM safety and security, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

We introduce Opacus, a free, open-source PyTorch library for training deep learning models with differential privacy (hosted at opacus.ai). Opacus is designed for simplicity, flexibility, and speed. It provides a simple and user-friendly API, and enables machine learning practitioners to make a training pipeline private by adding as little as two lines to their code. It supports a wide variety of layers, including multi-head attention, convolution, LSTM, GRU (and generic RNN), and embedding, right out of the box and provides the means for supporting other user-defined layers. Opacus computes batched per-sample gradients, providing higher efficiency compared to the traditional "micro batch" approach. In this paper we present Opacus, detail the principles that drove its implementation and unique features, and benchmark it against other frameworks for training models with differential privacy as well as standard PyTorch.

Key facts. Cited by 375 papers in the map (data through 7 Oct 2026). Builds on 12 of its 21 references that are in the map. First posted on arXiv 25 Sep 2021.

Builds on

12 of 21 references.

References · 12

  1. 2016Deep Learning with Differential Privacycited by 2,682
  2. 2017Renyi Differential Privacycited by 643
  3. 2019R'enyi Differential Privacy of the Sampled Gaussian Mechanismcited by 199
  4. 2021CrypTen: Secure Multi-Party Computation Meets Machine Learningcited by 131
  5. 2014Differentially Private Empirical Risk Minimization: Efficient Algorithms and Tight Error Boundscited by 104
  6. 2019BackPACK: Packing more into backpropcited by 77
  7. 2020Enabling Fast Differentially Private SGD via Just-in-Time Compilation and Vectorizationcited by 58
  8. 2020Scaling up Differentially Private Deep Learning with Fast Per-Example Gradient Clippingcited by 46
  9. 2015Efficient Per-Example Gradient Computationscited by 42
  10. 2021Fast and Memory Efficient Differentially Private-SGD via JL Projectionscited by 29
  11. 2019Efficient Per-Example Gradient Computations in Convolutional Neural Networkscited by 17
  12. 2018Efficient Deep Learning on Multi-Source Private Datacited by 16

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 4; 2022: 63; 2023: 81; 2024: 93; 2025: 67; 2026: 67.

375 papers.

Sort:

By area:

  1. 2021Large Language Models Can Be Strong Differentially Private Learnerscited by 405
  2. 2023Analyzing Leakage of Personally Identifiable Information in Language Modelscited by 278
  3. 2022Transformers in Medical Imaging: A Surveycited by 245
  4. 2022Unlocking High-Accuracy Differentially Private Image Classification through Scalecited by 218
  5. 2024Improving LoRA in Privacy-preserving Federated Learningcited by 158
  6. 2023How to DP-fy ML: A Practical Guide to Machine Learning with Differential Privacycited by 150
  7. 2021On the Importance of Difficulty Calibration in Membership Inference Attackscited by 137
  8. 2022FLamby: Datasets and Benchmarks for Cross-Silo Federated Learning in Realistic Healthcare Settingscited by 121
  9. 2022Differentially Private Diffusion Modelscited by 118
  10. 2022Synthetic Text Generation with Differential Privacy: A Simple and Practical Recipecited by 108
  11. 2022Toward Training at ImageNet Scale with Differential Privacycited by 87
  12. 2022Automatic Clipping: Differentially Private Deep Learning Made Easier and Strongercited by 87
38 more
  1. 2022Differentially Private Optimization on Large Model at Small Costcited by 65
  2. 2023Differentially Private Synthetic Data via Foundation Model APIs 1: Imagescited by 63
  3. 2022Privacy of Noisy Stochastic Gradient Descent: More Iterations without More Privacy Losscited by 60
  4. 2022Exploring the Limits of Differentially Private Deep Learning with Group-wise Clippingcited by 52
  5. 2022TAN Without a Burn: Scaling Laws of DP-SGDcited by 51
  6. 2022Scalable and Efficient Training of Large Convolutional Neural Networks with Differential Privacycited by 47
  7. 2022Differentially Private Bias-Term Fine-tuning of Foundation Modelscited by 47
  8. 2022GAP: Differentially Private Graph Neural Networks with Aggregation Perturbationcited by 46
  9. 2023Membership Inference of Diffusion Modelscited by 46
  10. 2022Differentially Private Learning Needs Hidden State (Or Much Faster Convergence)cited by 45
  11. 2024Private Fine-tuning of Large Language Models with Zeroth-order Optimizationcited by 41
  12. 2023Practical Membership Inference Attacks Against Large-Scale Multi-Modal Models: A Pilot Studycited by 37
  13. 2023Exploring and Exploiting Decision Boundary Dynamics for Adversarial Robustnesscited by 35
  14. 2024Evaluations of Machine Learning Privacy Defenses are Misleadingcited by 35
  15. 2022Private Set Generation with Discriminative Informationcited by 34
  16. 2023Make Landscape Flatter in Differentially Private Federated Learningcited by 33
  17. 2022Federated Learning with Sparsified Model Perturbation: Improving Accuracy under Client-Level Differential Privacycited by 32
  18. 2022Defending against Reconstruction Attacks with R'enyi Differential Privacycited by 30
  19. 2022M^4I: Multi-modal Models Membership Inferencecited by 30
  20. 2022SNAP: Efficient Extraction of Private Properties with Poisoningcited by 29
  21. 2024PrE-Text: Training Language Models on Private Federated Data in the Age of LLMscited by 29
  22. 2023DP-LDMs: Differentially Private Latent Diffusion Modelscited by 28
  23. 2023DP-AdamBC: Your DP-Adam Is Actually DP-SGD (Unless You Apply Bias Correction)cited by 28
  24. 2022Differentially private training of residual networks with scale normalisationcited by 26
  25. 2022A General Framework for Auditing Differentially Private Machine Learningcited by 25
  26. 2022CANIFE: Crafting Canaries for Empirical Privacy Measurement in Federated Learningcited by 24
  27. 2023On the Privacy-Robustness-Utility Trilemma in Distributed Learningcited by 24
  28. 2022Global Convergence of MAML and Theory-Inspired Neural Architecture Search for Few-Shot Learningcited by 22
  29. 2024How Private are DP-SGD Implementations?cited by 22
  30. 2022Differentially Private Learning with Per-Sample Adaptive Clippingcited by 21
  31. 2023A Survey of What to Share in Federated Learning: Perspectives on Model Utility, Privacy Leakage, and Communication Efficiencycited by 21
  32. 2024Differentially Private Next-Token Prediction of Large Language Modelscited by 21
  33. 2022Individual Privacy Accounting for Differentially Private Stochastic Gradient Descentcited by 20
  34. 2023Privacy and Efficiency of Communications in Federated Split Learningcited by 20
  35. 2023Gaussian Membership Inference Privacycited by 20
  36. 2024Privacy Backdoors: Stealing Data with Corrupted Pretrained Modelscited by 20
  37. 2024The Utility and Complexity of in- and out-of-Distribution Machine Unlearningcited by 20
  38. 2022Beyond Uniform Lipschitz Condition in Differentially Private Optimizationcited by 19

50 most cited of 375 All 375 (JSON)