Full Citation

Title: Leveraging Public Data in Practical Private Query Release: A Case Study with ACS Data

Citation Type: Miscellaneous

Publication Year: 2021

ISBN:

ISSN:

DOI:

NSFID:

PMCID:

PMID:

Abstract: It has been shown that for differentially private query release, the MWEM algorithm due to Hardt, Ligett, and McSherry (2012) achieves nearly optimal statistical guarantees. However, running MWEM on high-dimensional data is often infeasible, making the algorithm only applicable to lowdimensional data. In this paper, we study the setting in which the data curator has access to public data that is drawn from a similar—but related—distribution. Specifically, we present MW-Pub, which adapts MWEM to leverage prior knowledge from public samples and scale to high-dimensional data. Empirical evaluation on the American Community Survey (ACS) and the ADULT dataset shows that our method outperforms state-of-the-art methods under high privacy regimes.

Url: https://ppai21.github.io/files/26-paper.pdf

User Submitted?: No

Authors: Lui, Terrance; Vietri, Guiseppe; Steinke, Thomas; Ullman, Jonathan; Wu, Zhiwei Steven

Publisher:

Data Collections: IPUMS USA

Topics: Methodology and Data Collection

Countries: