Home: emilianodc.com
www.linkedin.com/posts/activi...
www.linkedin.com/posts/activi...
The Importance of Being Discrete: Measuring the Impact of Discretization in End-to-End Differentially Private Synthetic Data
We investigate the critical yet overlooked role of discretization in end-to-end differentially private (DP) tabular synthetic data generation
The Importance of Being Discrete: Measuring the Impact of Discretization in End-to-End Differentially Private Synthetic Data
We investigate the critical yet overlooked role of discretization in end-to-end differentially private (DP) tabular synthetic data generation
Paper: arxiv.org/pdf/2312.051...
Paper: arxiv.org/pdf/2312.051...
A "tiny" paper studying the impact of how you extract the data domain from tabular datasets while training generative models for synthetic data.
A "tiny" paper studying the impact of how you extract the data domain from tabular datasets while training generative models for synthetic data.
- We analyze 6 implementations of PATE-GAN, including 3 by the original authors.
- None reproduce the utility reported in the original paper
- All implementations leak more privacy than intended
- We uncover 19 privacy violations and 5 other bugs
- We analyze 6 implementations of PATE-GAN, including 3 by the original authors.
- None reproduce the utility reported in the original paper
- All implementations leak more privacy than intended
- We uncover 19 privacy violations and 5 other bugs
arxiv.org/abs/2304.08847
arxiv.org/abs/2304.08847