Private Post-GAN Boosting
Replication code for ICLR 2021 paper on differentially private synthetic data generation.
Synthetic Data
Method or Implementation? A Plea for More Rigor in Synthetic Data Benchmarking
Benchmarks that find GANs uncompetitive for tabular synthetic data often compare a default implementation, not the method.
Deceptively Real – Yet Privacy-Compliant? Synthetic Data with Formal Privacy Guarantees
Invited Talk at the Center for Clinical Data Science on synthetic data and privacy guarantees.
Using Prior Knowledge to Improve GANs for Tabular Data Without Compromising Privacy
Three ways to build prior knowledge about the data into GANs for private tabular synthetic data, with differential privacy.
An Introduction to Generative Adversarial Nets in R: The RGAN Package
Introduces GANs to R users and presents the RGAN package, which runs on torch and makes GAN design choices easy to vary.
On the Formal Privacy Guarantees of Synthetic Data (Generated Without Formal Privacy Guarantees)
Examining what privacy guarantees synthetic data can satisfy even without formal guarantees during synthesizer training.
Generating Synthetic Data is Complicated: Know Your Data and Know Your Generator
Ready-made synthetic data generators are not plug-and-play: preprocessing matters as much as tuning, and so does knowing the generator.
Continual Release of Differentially Private Synthetic Data from Longitudinal Data Collections
Algorithms for continually releasing differentially private synthetic data from longitudinal data collections.
Bootstrap-based, General-purpose Statistical Inference from Differential Private Releases
A general-purpose method combining bootstrap with differentially private non-parametric distribution estimation.
Private Post-GAN Boosting
Differentially private method combining samples from GAN training to create high-quality synthetic data.