CIF21 DIBBs: An Integrated System for Public/Private Access to Large-Scale, Confidential Social Science Data

这个研究项目将开发一个综合系统的试点，用于传播关于人的大规模数据。该项目将解决阻碍大规模数据库广泛传播的关键挑战，这些数据库可以推进基本的社会、行为和经济科学研究，并为社会提供巨大的潜在利益。这些数据的传播带来的挑战之一是意外泄露数据主体的身份和敏感属性，从而违反了旨在保护数据主体隐私和保密性的承诺，有时还违反了法律。该项目的产品将促进安全和有用的大规模数据集的发展和传播。该项目将产生可扩展的和开源的产品，这些产品构成了概念的证明，并将为将来更大规模的系统实现提供有价值的信息。因此，该项目将为数据传播的潜在转型奠定基础，为数据管理员提供他们所需的基础设施，以发布促进社会科学、政策制定和培训的数据产品。该项目还将为博士后研究员以及研究生和本科生提供教育和培训机会。研究人员将创造新的方法和广泛适用的工具，以满足数据传播的挑战。从技术角度来看，他们将推进通过能够处理高维数据的非参数方法生成合成数据集的方法。他们将推进对从大量编辑的数据中得出的推论的质量提供反馈的方法，并将制定方法，深入评估和描述在有或没有核查服务器的情况下发布大规模合成数据所固有的披露风险。从基础设施的角度来看，研究人员将开发用于集成三个核心工具（合成数据、验证服务器和远程访问）的系统和体系结构，从而实现对数据的安全、可扩展访问。该试点系统的目标是传播一个关于联邦政府雇员工作历史的数据集版本。

This research project will develop a pilot of an integrated system for disseminating large-scale data about people. This project will address critical challenges that have inhibited the wide-spread dissemination of large-scale databases that can advance basic social, behavioral, and economic science research and that offer enormous potential benefits to society. Among the challenges the dissemination of these data have posed are the unintended disclosures of data subjects' identities and sensitive attributes, thereby violating promises and sometimes laws designed to protect data subjects' privacy and confidentiality. The products of this project will facilitate the development and dissemination of safe and useful large-scale datasets. The project will result in extensible and open-source products that constitute a proof of concept and that will provide valuable information for future larger-scale implementations of the system. The project therefore will lay the groundwork for a potential transformation in data dissemination, providing data stewards with the infrastructure they need to release data products that advance social science, policy making, and training. The project also will provide education and training opportunities for a post-doctoral researcher as well as graduate and undergraduate students.The investigators will create new methodology and broadly applicable tools for meeting data dissemination challenges. From a technical perspective, they will advance methodology for generating synthetic datasets via nonparametric methods capable of handling highly dimensional data. They will advance methodology for providing feedback on the quality of inferences from heavily redacted data, and they will develop methods for in depth assessment and characterization of disclosure risks inherent in releasing large-scale synthetic data with and without verification servers. From an infrastructure perspective, the investigators will develop systems and architecture for integrating the three core tools (synthetic data, verification servers, and remote access) in ways that result in secure, scalable access to data. The pilot system will be built with the goal of disseminating a version of a dataset on the work histories of federal government employees.