Joint NSF-Census-IRS Workshop on synthetic data and confidentiality protection, July 2009 Washington, DC

这次讲习班汇集了大学的研究人员、国家统计界的成员和其他感兴趣的用户群体，讨论和批评新创建的公共使用的微型数据文件，这些文件基于合成数据和部分合成数据的概念。虽然这些想法已经有十多年的历史了，但统计机构现在才开始将它们作为其正在进行的统计业务的一部分。因此，重要的是激发来自用户社区与数据生产者互动的重要质量改进过程。在过去，这些机构隐瞒了保密的细节，以便最大限度地增加入侵者对其中一种产品进行攻击的难度。传统保护方法的细节具有保密性，没有留下多少通过与用户社区互动进行反复改进的余地。相比之下，合成数据产品是透明的，几乎所有重要的过程细节都可以公布。更重要的是，已经为基于合成数据的统计推断制定了准确的程序。最后，预计将根据用户社区的评估对发布产品进行交互式修订，作为机密性保护的一部分，因此不会损害它们。研讨会将关于这些新的合成数据产品的论文与用户社区的评估和讨论结合在一起。研究结果发表在《隐私与保密杂志》的特刊上，并在万维网上广泛传播，以启动教授用户社区如何适当分析此类数据的过程，并向数据提供者传授其当前方法需要改进的地方。研讨会计划于2009年7月31日在华盛顿特区人口普查局总部举行，为期一天。时间和地点允许与将于2009年8月2日至6日在华盛顿特区举行的2009年联合统计会议相吻合。运营研究数据中心网络的人口普查局经济研究中心、参与创建收入和方案参与调查(SIPP)综合Beta文件的美国国税局(IRS)以及OMB的首席统计师是共同发起人和支持者。简而言之，这是一个新的世界。统计机构比以往任何时候都更加依赖它们从其用户社区获得的关于发布数据质量的知情反馈，这些机构现在能够通过改进它们所依据的合成器来改进这些产品。这些观点在美国人口普查局S研究数据中心网络的大学合作伙伴SES-0427889资助的项目中得到了强调。在该项目中，人口普查局直接与人口普查局合作生产和评估各种合成微数据产品。通过演示基于合成数据方法的实际新创建和发布的公共使用产品的过程，该研讨会提供了一个具体的例子，说明用户社区和统计机构如何有效地合作开发更有用的保密

This workshop brings together university-based researchers, members of the national statistical community, and other interested user communities to discuss and critique newly created public-use micro-data files that are based on the concepts of synthetic data and partially synthetic data. Although these ideas are over a decade old, statistical agencies are only now beginning to use them as a part of their on-going statistical operations. Hence, it is important to incite the vital process of quality improvement that comes from the interaction of the user community with the data producers. In the past, the details of confidentiality protection have been withheld by the agencies in order to maximize the difficulty associated with an attack by an intruder on one of the products. The confidential nature of the details of the conventional protection methodologies left little room for iterative improvement by interaction with the user community. In contrast, synthetic data products are transparent, virtually all of the salient details of the process can be published. More important, exact procedures have already been developed for statistical inference based on synthetic data. Finally, interactive revision of the release product based on the assessments of the user community was anticipated as part of the confidentiality protections, and therefore does not compromise them.The workshop combines papers about these new synthetic data products with evaluations and discussions from the user community. The results are published in a special issue of the Journal of Privacy and Confidentiality and disseminated widely on the World Wide Web in order to jump-start the process of teaching the user community how to analyze such data appropriately and teaching the data providers where their current methods need improvement.The workshop is proposed for one full day, July 31, 2009, in the headquarters of the Census Bureau in Washington, DC. The timing and location permit dovetailing with the 2009 Joint Statistical Meetings, which will be held August 2-6, 2009 in Washington, DC. The Center for Economic Studies of the Census Bureau, which runs the Research Data Center network, the Internal Revenue Service (IRS), which has participated in the creation of the Survey of Income and Program Participation (SIPP) Synthetic Beta file, and the Chief Statistician at OMB are cosponsors and supporters.Broader Impacts. In short, it is a new world. The statistical agencies are more dependent than ever on the informed feedback they get from their user communities about the quality of the release data, and the agencies are now enabled to improve those products by improving the synthesizers upon which they are based. These points were emphasized in the project funded as NSF-ITR grant SES-0427889 in which the university-based partners in the Census Bureau?s Research Data Center network collaborated directly with the Census Bureau in producing and evaluating a variety of synthetic micro-data products. By demonstrating the process with real newly-created and release public-use products based on synthetic data methods, this workshop provides a concrete example of how the user community and the statistical agencies can effectively collaborate on the development of more useful confidentiality