PATE-GAN: Generating Synthetic Data with Differential Privacy Guarantees

By James Jordon, Jinsung Yoon, Mihaela van der Schaar. International Conference on Learning Representations, 2018

850 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/af1841e1db6579f1f1777a59c7e9e4658d2ac466

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

GAN-Leaks: A Taxonomy of Membership Inference Attacks against Generative Models (2019, background): “The generative model, on the other hand, has also gained increasing attention and has delivered appealing applications including photorealistic image synthesis [20, 44, 52, 70], text and sound generation [4, 49, 64, 65], sanitized dataset generation [2, 7, 35, 66, 73], etc.”
Machine Learning for Synthetic Data Generation: a Review (2023, methodology): “Jordon et al. [49] modifies the Private Aggregation of Teacher Ensembles (PATE) framework and applies it to GANs.”
Time-series Generative Adversarial Networks (2019, methodology): “Furthermore, by applying the "train on synthetic, test on real (TSTR)" framework [5, 9] to the sequence prediction task, we evaluate how well the generated data preserves the predictive characteristics of the original.”
Private FL-GAN: Differential Privacy Synthetic Data Generation Based on Federated Learning (2020, methodology): “Based on this, [13] proposed PATE-GAN which mod-iﬁed the PATE framework and applied it to GAN to generate synthetic data.”
Modeling Tabular data using Conditional GAN (2019, background): “The generator in GANs does not have access to real data during the entire training process; thus, we can make CTGAN achieve differential privacy [14] easier than TVAE.”
DP-CGAN: Differentially Private Synthetic Data and Label Generation (2019, methodology): “l privacy budget is accounted, the results do not look promising even on MNIST dataset and there is no methodology for creating labels for synthetic images. Similar to DPGAN method, PATE-GAN approach [30] enforces privacy by making the discriminator differentially private. In PATE-GAN, the discrimin”
Generating synthetic data in finance: opportunities, challenges and pitfalls (2020, methodology): “Several deep generative models based on GANs, conditional GANs and variational autoencoders (VAEs) have been proposed for tabular data, such as MedGAN [21], TableGAN [61], PATEGAN [45], CT-GAN and T-VAE [82].”
CTAB-GAN: Effective Table Data Synthesizing (2021, background): “PATE-GAN [27] is not specifically designed for tabular data generation, but it proposes a framework which generates synthetic data with differential privacy guarantees.”
Synthetic Data for Deep Learning (2019, background): “[675] circumvents the non-differentiability problem by training a “student-discriminator” on already differentially private synthetic data produced by the generator.”
Harnessing the power of synthetic data in healthcare: innovation, application, and privacy (2023, background): “Bridging this gap between the theoretical robustness of differential privacy and its practical implementation in the health sector, groundbreaking efforts such as the one by Jordon et al. 70 have started to emerge, utilizing differential privacy in the context of synthetic data generation.”
How to DP-fy ML: A Practical Guide to Machine Learning with Differential Privacy (2023, methodology): “, 2018) and PATE-GAN (Yoon et al., 2019), and the ensembling approach was recently proposed in (Liu et al.”
TabDDPM: Modelling Tabular Data with Diffusion Models (2022, background): “Generative models for tabular problems are currently an active research direction in the machine learning community (Xu et al., 2019; Engelmann & Lessmann, 2021; Jordon et al., 2018; Fan et al., 2020; Torfi et al., 2022; Zhao et al., 2021; Kim et al., 2021; Zhang et al., 2021; Nock & GuillameBert, 2”
Deep Neural Networks and Tabular Data: A Survey (2021, methodology): “With suitable data generators at hand, developers can use large, synthetic, and yet realistic datasets to develop better models, while not being subject to privacy concerns [145].”