Composable and versatile privacy via truncated CDP
Composable and versatile privacy via truncated CDP
复制标题
DOI：
10.1145/3188745.3188946
复制
发表时间：
2018-06
期刊：
影响因子：
--
通讯作者：
Mark Bun;C. Dwork;G. Rothblum;T. Steinke
中科院分区：
文献类型：
--
作者：
Mark Bun;C. Dwork;G. Rothblum;T. Steinke
我们提出了截短的集中差异隐私（TCDP），对差异隐私和集中差异隐私的改进。该新定义提供了强大而有效的构图保证，支持强大的算法技术，例如通过子采样来放大隐私放大，并实现了更准确的统计分析。特别是，我们展示了一项核心任务，新定义可以改善指数准确性。
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.