SaTC: CORE: Medium: Developing for Differential Privacy with Formal Methods and Counterexamples

SaTC：核心：媒介：使用正式方法和反例开发差异化隐私

• 批准号: 1702760
• 负责人: Daniel Kifer
• 金额: $ 120万
• 依托单位: Pennsylvania State Univ University Park
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-07-01 至 2023-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1702760&HistoricalAwards=false
• 关键词: SaTC; CORE; Medium; Developing; Differential

Differential privacy is an elegant formulation of conditions an algorithm must meet in order to avoid leaking personal information contained in its inputs. It is becoming mainstream in many research communities and has been deployed in practice in the private sector and some government agencies. Accompanying this growth is an unfortunate, but expected, difficulty: many algorithms that are claimed to be differentially private are incorrect. This phenomenon affects both new-comers and seasoned veterans of the differential privacy field because of the difficulty and subtlety of developing new differentially private algorithms.This proposal outlines a research plan for the development of a system called DevDP, whose purpose is to enable novice and expert users to develop prototypes and explore differentially private algorithms. In particular, the project will develop program analysis tools and theory that leverage both programming language and machine learning technology to aid the development of correct differentially private programs by automating much of the verification and reasoning about errors. As part of broader impacts, DevDP also has the potential to help educate students and less-technical members of the scientific community by providing interactive software tools. A solid understanding of differential privacy will become crucial as it makes its way into public policy.

差分隐私是算法必须满足的条件的一个优雅的公式，以避免泄露其输入中包含的个人信息。它正在成为许多研究界的主流，并已在私营部门和一些政府机构中实际部署。伴随着这种增长的是一个不幸的，但预期的困难：许多声称是差异隐私的算法是不正确的。这种现象影响新来者和经验丰富的老兵的差分隐私领域，因为开发新的差分隐私algorithms.This建议的困难和微妙的概述了一个研究计划的发展称为DevDP系统，其目的是使新手和专家用户开发原型和探索差分隐私算法。特别是，该项目将开发程序分析工具和理论，利用编程语言和机器学习技术，通过自动化大部分验证和错误推理来帮助开发正确的差异私有程序。作为更广泛影响的一部分，DevDP也有可能通过提供交互式软件工具来帮助教育学生和科学界技术含量较低的成员。对差异隐私的深入理解将成为至关重要的，因为它将进入公共政策。

项目成果

Differentially Private Confidence Intervals for Empirical Risk Minimization

经验风险最小化的差分私人置信区间

• DOI: 10.29012/jpc.660
• 发表时间: 2019
• 期刊: Journal of Privacy and Confidentiality
• 作者: Wang, Yue;Kifer, Daniel;Lee, Jaewoo
• 通讯作者: Lee, Jaewoo

A derivation framework for dependent security label inference

• DOI: 10.1145/3276485
• 发表时间: 2018-10
• 期刊: Proceedings of the ACM on Programming Languages
• 作者: Peixuan Li;Danfeng Zhang

Free gap estimates from the exponential mechanism, sparse vector, noisy max and related algorithms

来自指数机制、稀疏向量、噪声最大值和相关算法的自由间隙估计

• DOI: 10.1007/s00778-022-00728-2
• 发表时间: 2022
• 期刊: The VLDB Journal
• 作者: Ding, Zeyu;Wang, Yuxin;Xiao, Yingtai;Wang, Guanhong;Zhang, Danfeng;Kifer, Daniel
• 通讯作者: Kifer, Daniel

Renyi Differentially Private ERM for Smooth Objectives

• 发表时间: 2019-04
• 作者: Chen-Chen-Chen;Jaewoo Lee;Daniel Kifer

Detecting Outliers in Data with Correlated Measures

• DOI: 10.1145/3269206.3271798
• 发表时间: 2018-08
• 期刊: Proceedings of the 27th ACM International Conference on Information and Knowledge Management
• 作者: Yu-Hsuan Kuo;Z. Li;Daniel Kifer