mlsanityData to 7 Oct 2026

On the Utility Recovery Incapability of Neural Net-based Differential Private Tabular Training Data Synthesizer under Privacy Deregulation

arXiv 2211.15809 ↗First posted 28 Nov 2022cs.LGcited by 8

Not on the map. Why?

Abstract (arXiv)

Devising procedures for auditing generative model privacy-utility tradeoff is an important yet unresolved problem in practice. Existing works concentrates on investigating the privacy constraint side effect in terms of utility degradation of the train on synthetic, test on real paradigm of synthetic data training. We push such understanding on privacy-utility tradeoff to next level by observing the privacy deregulation side effect on synthetic training data utility. Surprisingly, we discover the Utility Recovery Incapability of DP-CTGAN and PATE-CTGAN under privacy deregulation, raising concerns on their practical applications. The main message is Privacy Deregulation does NOT always imply Utility Recovery.

Key facts. Cited by 8 papers in the map (data through 7 Oct 2026). Builds on 30 of its 51 references that are in the map. First posted on arXiv 28 Nov 2022.

Builds on

30 of 51 references.

References · 30

  1. 2016Deep Learning with Differential Privacycited by 2,682
  2. 2019Modeling Tabular data using Conditional GANcited by 695
  3. 2016Semi-supervised Knowledge Transfer for Deep Learning from Private Training Datacited by 347
  4. 2018Differentially Private Generative Adversarial Networkcited by 235
  5. 2021How Faithful is your Synthetic Data? Sample-level Metrics for Evaluating and Auditing Generative Modelscited by 193
  6. 2018Data Synthesis based on Generative Adversarial Networkscited by 184
  7. 2021CTAB-GAN: Effective Table Data Synthesizingcited by 156
  8. 2020DP-CGAN: Differentially Private Synthetic Data and Label Generationcited by 118
  9. 2020GS-WGAN: A Gradient-Sanitized Approach for Learning Differentially Private Generatorscited by 117
  10. 2020Synthetic Data -- Anonymisation Groundhog Daycited by 105
  11. 2021Winning the NIST Contest: A scalable and general approach to differentially private synthetic datacited by 101
  12. 2018Synthesizing Tabular Data using Generative Adversarial Networkscited by 92
18 more
  1. 2021Benchmarking Differentially Private Synthetic Data Generation Algorithmscited by 80
  2. 2020DP-MERF: Differentially Private Mean Embeddings with Random Features for Practical Privacy-Preserving Data Generationcited by 78
  3. 2019Graphical-model based estimation and inference for differential privacycited by 73
  4. 2022Synthetic Data -- what, why and how?cited by 66
  5. 2019G-PATE: Scalable Differentially Private Data Generator via Private Aggregation of Teacher Discriminatorscited by 53
  6. 2020Differentially Private Synthetic Data: Applied Evaluations and Enhancementscited by 36
  7. 2018Differentially Private Data Generative Modelscited by 33
  8. 2019Differentially Private Generative Adversarial Networks for Time Series, Continuous, and Discrete Open Datacited by 31
  9. 2021PEARL: Data Synthesis via Private Embeddings and Adversarial Reconstruction Learningcited by 23
  10. 2016PrivTree: A Differentially Private Algorithm for Hierarchical Decompositionscited by 22
  11. 2018Generating Synthetic but Plausible Healthcare Record Datasetscited by 12
  12. 2022dpart: Differentially Private Autoregressive Tabular, a General Framework for Synthetic Data Generationcited by 12
  13. 2022Faking feature importance: A cautionary tale on the use of differentially-private synthetic datacited by 8
  14. 2022Visualizing Privacy-Utility Trade-Offs in Differentially Private Data Releasescited by 7
  15. 2021DP-SGD vs PATE: Which Has Less Disparate Impact on GANs?cited by 6
  16. 2021Relaxed Marginal Consistency for Differentially Private Query Answeringcited by 5
  17. 2022Enabling Synthetic Data adoption in regulated domainscited by 4
  18. 2021The Privacy-Utility Tradeoff of Robust Local Differential Privacycited by 1

Built on by

Cited by year

Citing papers per year: 2023: 2; 2024: 3; 2025: 3.

8 papers.

  1. 2023Graphical vs. Deep Generative Models: Measuring the Impact of Differentially Private Mechanisms and Budgets on Utilitycited by 13
  2. 2024Data Plagiarism Index: Characterizing the Privacy Risk of Data-Copying in Tabular Generative Modelscited by 12
  3. 2025Synthetic Tabular Data: Methods, Attacks and Defensescited by 8
  4. 2023FLAIM: AIM-based Synthetic Data Generation in the Federated Settingcited by 6
  5. 2024Advancing Retail Data Science: Comprehensive Evaluation of Synthetic Datacited by 5
  6. 2025GReaTER: Generate Realistic Tabular data after data Enhancement and Reductioncited by 3
  7. 2024DEREC-SIMPRO: unlock Language Model benefits to advance Synthesis in Data Clean Roomcited by 2
  8. 2025Leveraging Vertical Public-Private Split for Improved Synthetic Data Generationcited by 2