Composable and Versatile Privacy via Truncated CDP

被引：87

作者：

Bun, Mark ^{[1
]}

Dwork, Cynthia ^{[2
]}

Rothblum, Guy N. ^{[3
]}

Steinke, Thomas ^{[4
]}

机构：

[1] Princeton Univ, Princeton, NJ 08544 USA

[2] Harvard Univ, Cambridge, MA 02138 USA

[3] Weizmann Inst Sci, Rehovot, Israel

[4] IBM Res Almaden, San Jose, CA USA

来源：

STOC'18: PROCEEDINGS OF THE 50TH ANNUAL ACM SIGACT SYMPOSIUM ON THEORY OF COMPUTING | 2018年

基金：

以色列科学基金会; 美国国家科学基金会; 英国工程与自然科学研究理事会;

关键词：

differential privacy; algorithmic stability; subsampling;

D O I：

10.1145/3188745.3188946

中图分类号：

TP301 [理论、方法];

学科分类号：

081202 ;

摘要：

We propose truncated concentrated differential privacy (tCDP), a refinement of differential privacy and of concentrated differential privacy. This new definition provides robust and efficient composition guarantees, supports powerful algorithmic techniques such as privacy amplification via sub-sampling, and enables more accurate statistical analyses. In particular, we show a central task for which the new definition enables exponential accuracy improvement.

引用

页码：74 / 86

页数：13