Algorithmically Effective Differentially Private Synthetic Data
Data Structures and Algorithms
2023-05-22 v3 Cryptography and Security
Probability
Statistics Theory
Statistics Theory
Abstract
We present a highly effective algorithmic approach for generating -differentially private synthetic data in a bounded metric space with near-optimal utility guarantees under the 1-Wasserstein distance. In particular, for a dataset in the hypercube , our algorithm generates synthetic dataset such that the expected 1-Wasserstein distance between the empirical measure of and is for , and is for . The accuracy guarantee is optimal up to a constant factor for , and up to a logarithmic factor for . Our algorithm has a fast running time of for all and demonstrates improved accuracy compared to the method in (Boedihardjo et al., 2022) for .
Cite
@article{arxiv.2302.05552,
title = {Algorithmically Effective Differentially Private Synthetic Data},
author = {Yiyun He and Roman Vershynin and Yizhe Zhu},
journal= {arXiv preprint arXiv:2302.05552},
year = {2023}
}
Comments
23 pages. to appear in COLT 2023