Collaborative Research: IMR: MM-1B: Foundations for Differentially Private Internet Measurement

合作研究：IMR：MM-1B：差分隐私互联网测量的基础

• 批准号: 2220433
• 负责人: Tianhao Wang
• 金额: $ 29.96万
• 依托单位: University of Virginia Main Campus
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-10-01 至 2025-09-30
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2220433&HistoricalAwards=false
• 关键词: Collaborative; Research; IMR; MM; 1B

Internet measurement research has been providing critical insights and evidence support in guiding the design of Internet infrastructures. However, how to share and use Internet measurement data while respecting users' privacy is still challenging. To protect privacy, most works choose to anonymize sensitive fields or only publish the aggregated statistics, but such methods are vulnerable to privacy attacks. Differential privacy, which provides guaranteed privacy protection for data release, has gained prominent traction from companies and government agencies, and is a natural choice for sharing Internet measurement data. However, a critical gap remains to identify how differential privacy can be applied to networking problems. This project aims to understand privacy issues and then lay the foundations for deploying differential privacy in the processing pipeline of Internet measurement data. The project's broader significance and importance include transferring the technologies to industry, involving members from under-represented groups, and disseminating outcomes through K-12 outreach and community services.The project is structured around three main aims. Thrust 1 conducts a comprehensive study of the privacy issues in the existing practices in collecting and sharing Internet measurement data, and develops an inference attack benchmark to assess the privacy risks and the benefits of different protection mechanisms. Thrust 2 investigates how differential privacy can be adjusted and integrated into the processing pipeline of Internet measurement data, by developing new configurable and relaxed differential privacy notions. Thrust 3 develops methods to publish synthetic Internet measurement data while satisfying differential privacy.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.

互联网测量研究为指导互联网基础设施的设计提供了重要的见解和证据支持。然而，如何在尊重用户隐私的前提下共享和使用互联网测量数据仍然是一个挑战。为了保护隐私，大多数作品选择将敏感字段匿名化或只发布汇总统计数据，但这种方法容易受到隐私攻击。差分隐私为数据发布提供有保障的隐私保护，受到企业和政府机构的广泛关注，是共享互联网测量数据的自然选择。然而，一个关键的差距仍然是确定如何将差异隐私应用于网络问题。本项目旨在了解隐私问题，为在互联网测量数据处理管道中部署差分隐私奠定基础。该项目的更广泛意义和重要性包括将技术转让给工业界，让代表性不足的群体成员参与进来，并通过K-12外展和社区服务传播成果。该项目围绕三个主要目标构建。Thrust 1对现有互联网测量数据收集和共享实践中的隐私问题进行了全面研究，并开发了一个推理攻击基准，以评估不同保护机制的隐私风险和收益。推力2通过开发新的可配置和宽松的差分隐私概念，研究如何调整差分隐私并将其集成到互联网测量数据的处理管道中。推力3开发了在满足差分隐私的同时发布综合互联网测量数据的方法。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。

项目成果

Differentially Private Vertical Federated Clustering

• DOI: 10.14778/3583140.3583146
• 发表时间: 2022-08
• 期刊: Proc. VLDB Endow.
• 作者: Zitao Li;Tianhao Wang;Ninghui Li

PrivTrace: Differentially Private Trajectory Synthesis by Adaptive Markov Models

PrivTrace：自适应马尔可夫模型的差分隐私轨迹合成

• 发表时间: 2023
• 期刊: USENIX
• 作者: Wang, Haiming;Zhang. Zhikun;Wang, Tianhao;He, Shibo;Backes, Michael;Chen, Jiming;Zhang, Yang
• 通讯作者: Zhang, Yang

Practical Differentially Private and Byzantine-resilient Federated Learning

• DOI: 10.1145/3589264
• 发表时间: 2023-04
• 期刊: Proceedings of the ACM on Management of Data
• 作者: Zihang Xiang;Tianhao Wang;Wanyu Lin;Di Wang

Federated Boosted Decision Trees with Differential Privacy

• DOI: 10.1145/3548606.3560687
• 发表时间: 2022-10
• 期刊: Proceedings of the 2022 ACM SIGSAC Conference on Computer and Communications Security
• 作者: Samuel Maddock;Graham Cormode;Tianhao Wang;C. Maple;S. Jha