Published May 2008 | Version Submitted
Book Section - Chapter Open

A learning theory approach to non-interactive database privacy

  • 1. ROR icon Carnegie Mellon University

Abstract

We demonstrate that, ignoring computational constraints, it is possible to release privacy-preserving databases that are useful for all queries over a discretized domain from any given concept class with polynomial VC-dimension. We show a new lower bound for releasing databases that are useful for halfspace queries over a continuous domain. Despite this, we give a privacy-preserving polynomial time algorithm that releases information useful for all halfspace queries, for a slightly relaxed definition of usefulness. Inspired by learning theory, we introduce a new notion of data privacy, which we call distributional privacy, and show that it is strictly stronger than the prevailing privacy notion, differential privacy.

Additional Information

© 2008 ACM. Supported in part by the National Science Foundation under grant CCF-0514922. Supported in part by an AT&T Labs Graduate Research Fellowship and an NSF Graduate Research Fellowship. We thank David Abraham, Cynthia Dwork, Shiva Kasiviswanathan, Adam Meyerson, Sofya Raskhodnikova, Amit Sahai, and Adam Smith for many useful discussions. We thank Ryan O'Donnell for the insight that led to the proof of Theorem A.6.

Attached Files

Submitted - 1109.2229.pdf

Files

1109.2229.pdf

Files (219.0 kB)

Name Size
md5:765c94f4e0c1bab0251519c2c9fd7990
219.0 kB Preview Download

Additional details

Identifiers

Eprint ID
92263
DOI
10.1145/1374376.1374464
Resolver ID
CaltechAUTHORS:20190114-152232213

Funding

NSF
CCF-0514922
AT&T
NSF Graduate Research Fellowship

Dates

Created
2019-01-22
Created from EPrint's datestamp field
Updated
2021-11-16
Created from EPrint's last_modified field