mlsanityData to 7 Oct 2026

Certification for Differentially Private Prediction in Gradient-Based Training

arXiv 2406.13433 ↗First posted 19 Jun 2024cs.LG cs.AIcited by 6

Not on the map. Why?

Abstract (arXiv)

We study private prediction where differential privacy is achieved by adding noise to the outputs of a non-private model. Existing methods rely on noise proportional to the global sensitivity of the model, often resulting in sub-optimal privacy-utility trade-offs compared to private training. We introduce a novel approach for computing dataset-specific upper bounds on prediction sensitivity by leveraging convex relaxation and bound propagation techniques. By combining these bounds with the smooth sensitivity mechanism, we significantly improve the privacy analysis of private prediction compared to global sensitivity-based approaches. Experimental results across real-world datasets in medical image classification and natural language processing demonstrate that our sensitivity bounds are can be orders of magnitude tighter than global sensitivity. Our approach provides a strong basis for the development of novel privacy preserving technologies.

Key facts. Cited by 6 papers in the map (data through 7 Oct 2026). Builds on 39 of its 55 references that are in the map. First posted on arXiv 19 Jun 2024.

Builds on

39 of 55 references.

References · 39

  1. 2021On the Opportunities and Risks of Foundation Modelscited by 3,961
  2. 2016Deep Learning with Differential Privacycited by 2,682
  3. 2019Machine Unlearningcited by 1,020
  4. 2021Membership Inference Attacks From First Principlescited by 839
  5. 2022Quantifying Memorization Across Neural Language Modelscited by 755
  6. 2017Reluplex: An Efficient SMT Solver for Verifying Deep Neural Networkscited by 739
  7. 2019Certified Data Removal from Machine Learning Modelscited by 512
  8. 2021Large Language Models Can Be Strong Differentially Private Learnerscited by 405
  9. 2021Opacus: User-Friendly Differential Privacy Library in PyTorchcited by 375
  10. 2021Remember What You Want to Forget: Algorithms for Machine Unlearningcited by 360
  11. 2016Semi-supervised Knowledge Transfer for Deep Learning from Private Training Datacited by 347
  12. 2018On the Effectiveness of Interval Bound Propagation for Training Verifiably Robust Modelscited by 322
27 more
  1. 2020Descent-to-Delete: Gradient-Based Methods for Machine Unlearningcited by 303
  2. 2022A Survey of Machine Unlearningcited by 298
  3. 2018Scalable Private Learning with PATEcited by 266
  4. 2017Machine Learning Models that Remember Too Muchcited by 165
  5. 2023Flocks of Stochastic Parrots: Differentially Private Prompt Learning for Large Language Modelscited by 80
  6. 2014Differential Privacy and Machine Learning: a Survey and Reviewcited by 66
  7. 2021Differential privacy and robust statistics in high dimensionscited by 64
  8. 2022TAN Without a Burn: Scaling Laws of DP-SGDcited by 51
  9. 2022Certified Training: Small Boxes are All You Needcited by 51
  10. 2021Antipodes of Label Differential Privacy: PATE and ALIBIcited by 50
  11. 2017Feature-Guided Black-Box Safety Testing of Deep Neural Networkscited by 48
  12. 2018Privacy-preserving Predictioncited by 48
  13. 2022Differential Privacy and Fairness in Decisions and Learning Tasks: A Surveycited by 47
  14. 2020Probabilistic Safety for Bayesian Neural Networkscited by 41
  15. 2021Partition-based formulations for mixed-integer optimization of trained ReLU neural networkscited by 39
  16. 2023When Deep Learning Meets Polyhedral Theory: A Surveycited by 39
  17. 2021CaPC Learning: Confidential and Private Collaborative Learningcited by 33
  18. 2022Robust Explanation Constraints for Neural Networkscited by 23
  19. 2020The Trade-Offs of Private Predictioncited by 22
  20. 2017Accuracy First: Selecting a Differential Privacy Level for Accuracy-Constrained ERMcited by 20
  21. 2022Individual Privacy Accounting for Differentially Private Stochastic Gradient Descentcited by 20
  22. 2014Characterizing the Sample Complexity of Private Learnerscited by 19
  23. 2019Privately Answering Classification Queries in the Agnostic PAC Modelcited by 18
  24. 2023Tight Bounds for Machine Unlearning via Differential Privacycited by 18
  25. 2024Certified Robustness to Data Poisoning in Gradient-Based Trainingcited by 17
  26. 2024Auditing Private Predictioncited by 7
  27. 2024Scaling Mixed-Integer Programming for Certification of Neural Network Controllers Using Bounds Tighteningcited by 6

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 2; 2026: 4.

6 papers.

  1. 2025Atlas: A Framework for ML Lifecycle Provenance & Transparencycited by 8
  2. 2026Exact Certification of Data-Poisoning Attacks Using Mixed-Integer Programmingcited by 3
  3. 2025Provably Safe Model Updatescited by 2
  4. 2026SafeAdapt: Provably Safe Policy Updates in Deep Reinforcement Learningcited by 0
  5. 2026Hybrid Robustness Verification for Spatio-Temporal Neural Networkscited by 0
  6. 2026Certification-Based Differentially Private Learningcited by 0