|
美国人口普查局采用差别隐私
The U.S. Census Bureau Adopts Differential Privacy |
|
| |
| 课程网址: |
http://videolectures.net/kdd2018_abowd_differential_privacy/
|
| 主讲教师: |
John Abowd
|
| 开课单位: |
美国人口普查局
|
| 开课时间: |
2018-11-23
|
| 课程语种: |
英语
|
| 中文简介: |
美国人口普查局通过其科学咨询委员会宣布，将利用差别隐私保护2020年人口和住房普查中的出版物。十年一次的人口普查是宪法规定的人口统计，用于重新分配众议院和重新划分全国每个立法区。人口普查局进行了内部研究，确认用于2000年和2010年人口普查的统计披露限制系统存在严重漏洞，Dinur和Nissim（2003）数据库重建定理暴露了这些漏洞。我们设计了一个差异化的私人出版系统，直接解决了这些漏洞，同时保持了核心统计产品的适用性。这是有史以来规模最大的中央差别隐私实施。设计有效的算法属于计算机科学领域。选择要实现的实际隐私损失参数属于经济学领域。我们的算法针对任何隐私损失参数有效地分配差异隐私注入的噪声，以确保适合使用2020年人口普查统计数据。人口普查局的数据管理执行政策委员会在审查准确性与隐私损失权衡的图形摘要后选择隐私损失参数。这些决定是在对实际2020年人口普查数据运行算法之前做出的。
|
| 课程简介: |
The U.S. Census Bureau announced, via its Scientific Advisory Committee, that it would protect the publications from the 2020 Census of Population and Housing using differential privacy. The decennial census is the constitutionally mandated enumeration of the population used to reapportion the House of Representatives and redraw every legislative district in the country. The Census Bureau conducted internal research confirming that the statistical disclosure limitation systems used for the 2000 and 2010 Censuses had serious vulnerabilities that were exposed by the Dinur and Nissim (2003) database reconstruction theorem. We designed a differentially private publication system that directly addressed these vulnerabilities while preserving the fitness for use of the core statistical products. This is the largest-scale central differential privacy implementation ever undertaken. Designing efficient algorithms lies in the domain of computer science. Choosing the actual privacy-loss parameters to implement lies in the domain of economics. Our algorithms efficiently distribute the noise injected by differential privacy for any privacy-loss parameter in order to insure fitness-for-use of the 2020 Census statistics. The Census Bureau’s Data Stewardship Executive Policy Committee selects the privacy-loss parameters after reviewing graphical summaries of the accuracy versus privacy-loss tradeoff. These decisions are made before the algorithms are run on the actual 2020 Census data.
|
| 关 键 词: |
美国人口普查; 数据库重建定理; 有效地分配差异隐私; 隐私损失权衡的图形摘要
|
| 课程来源: |
视频讲座网
|
| 数据采集: |
2023-01-27：cyh
|
| 最后编审: |
2023-01-27：cyh
|
| 阅读次数: |
51
|
|
图 书 馆