CRII: SaTC: Democratizing Differential Privacy via Algorithms for Hybrid Models

个人产生了大量的个人数据，这些数据随后被组织和政府收集和存储。这些数据为网络服务、医疗保健和交通运输等领域的许多创新应用提供了动力，但它们也增加了隐私风险。差分隐私是一种严格推理算法隐私属性的框架，它为实现隐私保护而又有用的数据分析提供了巨大的希望。然而，它的采用仅限于拥有大量用户基础的实体。该项目旨在通过使其适用于用户基数较小的实体，使部署差异隐私的能力民主化。该项目的研究活动包括制定一个新的混合隐私模型，该模型模拟个人和当前行业实践的异质隐私偏好。该项目还开发了新的算法，在利用混合模型改善效用结果和评估其性能的同时，保护差异隐私。通过将差异隐私的适用性扩展到更广泛的实体，这项工作在消除数据驱动创新的重大障碍方面取得了进展。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Individuals generate enormous amounts of personal data that are subsequently collected and stored by organizations and governments. The data powers many innovative applications in areas such as web services, health care, and transportation, but they also increase privacy risks. Differential privacy, a framework to rigorously reason about privacy properties of algorithms, holds tremendous promise for enabling privacy-preserving yet useful data analyses. However, its adoption has been limited to entities with massive user bases. This project aims to democratize the ability to deploy differential privacy by making it practical for entities with smaller user bases. Research activities in this project include formulation of a new, hybrid, privacy model that models heterogeneous privacy preferences of individuals and current industry practices. The project also develops novel algorithms that preserve differential privacy while taking advantage of the hybrid model to improve utility outcomes, and evaluation of their performance. The work makes progress towards eliminating one of the significant barriers to data-driven innovation by expanding the applicability of differential privacy to a wider range of entities.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.