{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,30]],"date-time":"2026-07-30T14:15:06Z","timestamp":1785420906805,"version":"3.56.0"},"reference-count":35,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2024,1,8]],"date-time":"2024-01-08T00:00:00Z","timestamp":1704672000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Big Data"],"abstract":"<jats:p>The usage of synthetic data is gaining momentum in part due to the unavailability of original data due to privacy and legal considerations and in part due to its utility as an augmentation to the authentic data. Generative adversarial networks (GANs), a paragon of generative models, initially for images and subsequently for tabular data, has contributed many of the state-of-the-art synthesizers. As GANs improve, the synthesized data increasingly resemble the real data risking to leak privacy. Differential privacy (DP) provides theoretical guarantees on privacy loss but degrades data utility. Striking the best trade-off remains yet a challenging research question. In this study, we propose CTAB-GAN+ a novel conditional tabular GAN. CTAB-GAN+ improves upon state-of-the-art by (i) adding downstream losses to conditional GAN for higher utility synthetic data in both classification and regression domains; (ii) using Wasserstein loss with gradient penalty for better training convergence; (iii) introducing novel encoders targeting mixed continuous-categorical variables and variables with unbalanced or skewed data; and (iv) training with DP stochastic gradient descent to impose strict privacy guarantees. We extensively evaluate CTAB-GAN+ on statistical similarity and machine learning utility against state-of-the-art tabular GANs. The results show that CTAB-GAN+ synthesizes privacy-preserving data with at least 21.9% higher machine learning utility (i.e., F1-Score) across multiple datasets and learning tasks under given privacy budget.<\/jats:p>","DOI":"10.3389\/fdata.2023.1296508","type":"journal-article","created":{"date-parts":[[2024,1,8]],"date-time":"2024-01-08T05:10:20Z","timestamp":1704690620000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":105,"title":["CTAB-GAN+: enhancing tabular data synthesis"],"prefix":"10.3389","volume":"6","author":[{"given":"Zilong","family":"Zhao","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Aditya","family":"Kunar","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Robert","family":"Birke","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hiek","family":"Van der Scheer","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lydia Y.","family":"Chen","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2024,1,8]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","DOI":"10.1145\/2976749.2978318","article-title":"\u201cDeep learning with differential privacy,\u201d","author":"Abadi","year":"2016","journal-title":"ACM SIGSAC Conference on Computer and Communications Security (CCS)"},{"key":"B2","first-page":"214","article-title":"\u201cWasserstein generative adversarial networks,\u201d","author":"Arjovsky","year":"2017","journal-title":"Proceedings of the 34th ICML"},{"key":"B3","article-title":"The cramer distance as a solution to biased wasserstein gradients","author":"Bellemare","year":"2017","journal-title":"arXiv preprint arXiv:1705.10743"},{"key":"B4","volume-title":"Pattern Recognition and Machine Learning (Information Science and Statistics)","author":"Bishop","year":"2006"},{"key":"B5","article-title":"GS-WGAN: a gradient-sanitized approach for learning differentially private generators","author":"Chen","year":"","journal-title":"arXiv preprint arXiv:2006.08265"},{"key":"B6","doi-asserted-by":"publisher","DOI":"10.1145\/3372297.3417238","article-title":"\u201cGAN-leaks: a taxonomy of membership inference attacks against generative models,\u201d","author":"Chen","year":"","journal-title":"ACM SIGSAC Conference on Computer and Communications Security (CCS)"},{"key":"B7","article-title":"\u201cNeural ordinary differential equations,\u201d","author":"Chen","year":"2018","journal-title":"Advances in Neural Information Processing Systems, Vol. 31"},{"key":"B8","first-page":"13773","article-title":"\u201cUnderstanding gradient clipping in private SGD: a geometric perspective,\u201d","author":"Chen","year":"","journal-title":"Advances in Neural Information Processing Systems 33"},{"key":"B9","article-title":"Generating multi-label discrete patient records using generative adversarial networks","author":"Choi","year":"2017","journal-title":"arXiv preprint arXiv:1703.06490"},{"key":"B10","doi-asserted-by":"crossref","DOI":"10.1007\/978-3-540-79228-4_1","article-title":"\u201cDifferential privacy: a survey of results,\u201d","volume-title":"International Conference on Theory and Applications of Models of Computation (TAMC)","author":"Dwork","year":"2008"},{"key":"B11","volume-title":"The Algorithmic Foundations of Differential Privacy","author":"Dwork","year":"2014"},{"key":"B12","doi-asserted-by":"publisher","DOI":"10.1016\/j.eswa.2021.114582","article-title":"Conditional Wasserstein GAN-based oversampling of tabular data for imbalanced learning","author":"Engelmann","year":"2020","journal-title":"arXiv preprint arXiv:2008.09202"},{"key":"B13","first-page":"2672","article-title":"\u201cGenerative adversarial nets,\u201d","volume-title":"Proceedings of the 27th NIPS - Volume 2","author":"Goodfellow","year":"2014"},{"key":"B14","first-page":"5769","article-title":"\u201cImproved training of Wasserstein GANs,\u201d","author":"Gulrajani","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"key":"B15","doi-asserted-by":"publisher","DOI":"10.1162\/99608f92.353c6f99","author":"Hawes","year":"2020","journal-title":"Implementing Differential Privacy: Seven Lessons from the 2020 United States Census"},{"key":"B16","article-title":"\u201cPate-GAN: generating synthetic data with differential privacy guarantees,\u201d","author":"Jordon","year":"2018","journal-title":"International Conference on Learning Representations (ICLR)"},{"key":"B17","first-page":"4263","article-title":"\u201cInvertible tabular GANs: killing two birds with one stone for tabular data synthesis,\u201d","author":"Lee","year":"2021","journal-title":"NeurIPS Conference"},{"key":"B18","article-title":"Scalable differentially private generative student model via pate","author":"Long","year":"2019","journal-title":"arXiv preprint arXiv:1906.09338"},{"key":"B19","doi-asserted-by":"publisher","DOI":"10.1109\/CSF.2017.11","article-title":"\u201cR\u00e9nyi differential privacy,\u201d","author":"Mironov","year":"2017","journal-title":"Computer Security Foundations Symposium (CSF)"},{"key":"B20","article-title":"\u201cAirline passenger name record generation using generative adversarial networks,\u201d","author":"Mottini","year":"2018","journal-title":"Workshop on Theoretical Foundations and Applications of Deep Generative Models"},{"key":"B21","doi-asserted-by":"publisher","first-page":"111","DOI":"10.1109\/SP.2008.33","article-title":"\u201cRobust de-anonymization of large sparse datasets,\u201d","author":"Narayanan","year":"2008","journal-title":"IEEE Symposium on Security and Privacy"},{"key":"B22","first-page":"2642","article-title":"\u201cConditional image synthesis with auxiliary classifier GANs,\u201d","author":"Odena","year":"2017","journal-title":"The 34th ICML - Volume 70"},{"key":"B23","article-title":"Semi-supervised knowledge transfer for deep learning from private training data","author":"Papernot","year":"2016","journal-title":"arXiv preprint arXiv:1610.05755"},{"key":"B24","doi-asserted-by":"publisher","first-page":"1071","DOI":"10.14778\/3231751.3231757","article-title":"Data synthesis based on generative adversarial networks","volume":"11","author":"Park","year":"2018","journal-title":"Proc. VLDB Endow"},{"key":"B25","article-title":"Synthetic data-a privacy mirage","author":"Stadler","year":"2020","journal-title":"arXiv preprint arXiv:2011.07018"},{"key":"B26","article-title":"Differentially private synthetic medical data generation using convolutional GANs","author":"Torfi","year":"2020","journal-title":"arXiv preprint arXiv:2012.11774"},{"key":"B27","doi-asserted-by":"publisher","DOI":"10.1109\/CVPRW.2019.00018","article-title":"\u201cDP-cGAN: differentially private synthetic data and label generation,\u201d","author":"Torkzadehmahani","year":"2019","journal-title":"Conference on Computer Vision and Pattern Recognition (CVPR) Workshops"},{"key":"B28","doi-asserted-by":"crossref","DOI":"10.1145\/3124749.3124754","article-title":"\u201cDeep & cross network for ad click predictions,\u201d","volume-title":"Proceedings of the ADKDD'17","author":"Wang","year":"2017"},{"key":"B29","doi-asserted-by":"publisher","DOI":"10.29012\/jpc.723","article-title":"\u201cSubsampled r\u00e9nyi differential privacy and analytical moments accountant,\u201d","author":"Wang","year":"2019","journal-title":"International Conference on Artificial Intelligence and Statistics (AISTATS)"},{"key":"B30","article-title":"From GAN to WGAN","author":"Weng","year":"2019","journal-title":"arXiv preprint arXiv:1904.08994"},{"key":"B31","article-title":"Differentially private generative adversarial network","author":"Xie","year":"2018","journal-title":"arXiv preprint arXiv:1802.06739"},{"key":"B32","article-title":"\u201cModeling tabular data using conditional GAN,\u201d","author":"Xu","year":"2019"},{"key":"B33","article-title":"\u201cGenerative adversarial networks for electronic health records: a framework for exploring and evaluating methods for predicting drug-induced laboratory test trajectories,\u201d","author":"Yahi","year":"2017","journal-title":"NIPS Workshop"},{"key":"B34","article-title":"Differentially private releasing via deep generative model (technical report)","author":"Zhang","year":"2018","journal-title":"arXiv preprint arXiv:1801.01594"},{"key":"B35","first-page":"97","article-title":"\u201cCTAB-GAN: effective table data synthesizing,\u201d","author":"Zhao","year":"2021","journal-title":"Proceedings of The 13th Asian Conference on Machine Learning"}],"container-title":["Frontiers in Big Data"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fdata.2023.1296508\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,1,8]],"date-time":"2024-01-08T05:10:35Z","timestamp":1704690635000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fdata.2023.1296508\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,1,8]]},"references-count":35,"alternative-id":["10.3389\/fdata.2023.1296508"],"URL":"https:\/\/doi.org\/10.3389\/fdata.2023.1296508","relation":{},"ISSN":["2624-909X"],"issn-type":[{"value":"2624-909X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,1,8]]},"article-number":"1296508"}}