Generative Adversarial Networks for Synthetic Data Generation: A Comparative Study

TL;DR

This paper compares various GAN-based methods for generating synthetic census microdata against traditional techniques, evaluating data utility and privacy risks to assess their effectiveness and safety.

Contribution

It provides a comprehensive comparison of GANs and traditional methods for synthetic tabular data generation, focusing on utility and privacy metrics.

Findings

GANs produce high-utility synthetic census data

GANs exhibit different privacy risk profiles compared to traditional methods

The study highlights strengths and limitations of GANs for data synthesis

Abstract

Generative Adversarial Networks (GANs) are gaining increasing attention as a means for synthesising data. So far much of this work has been applied to use cases outside of the data confidentiality domain with a common application being the production of artificial images. Here we consider the potential application of GANs for the purpose of generating synthetic census microdata. We employ a battery of utility metrics and a disclosure risk metric (the Targeted Correct Attribution Probability) to compare the data produced by tabular GANs with those produced using orthodox data synthesis methods.