The Elusive Pursuit of Reproducing PATE-GAN: Benchmarking, Auditing, Debugging
This work addresses critical reproducibility and privacy issues in differentially private generative models for researchers and practitioners, but it is incremental as it focuses on auditing existing implementations rather than proposing new methods.
The paper tackled the reproducibility and privacy of PATE-GAN implementations by benchmarking six open-source versions, finding that none reproduced the original utility performance and all leaked more privacy than intended, uncovering 19 privacy violations and 5 other bugs.
Synthetic data created by differentially private (DP) generative models is increasingly used in real-world settings. In this context, PATE-GAN has emerged as one of the most popular algorithms, combining Generative Adversarial Networks (GANs) with the private training approach of PATE (Private Aggregation of Teacher Ensembles). In this paper, we set out to reproduce the utility evaluation from the original PATE-GAN paper, compare available implementations, and conduct a privacy audit. More precisely, we analyze and benchmark six open-source PATE-GAN implementations, including three by (a subset of) the original authors. First, we shed light on architecture deviations and empirically demonstrate that none reproduce the utility performance reported in the original paper. We then present an in-depth privacy evaluation, which includes DP auditing, and show that all implementations leak more privacy than intended. Furthermore, we uncover 19 privacy violations and 5 other bugs in these six open-source implementations. Lastly, our codebase is available from: https://github.com/spalabucr/pategan-audit.