Attribute Privacy: Framework and Mechanisms

Attribute Privacy: Framework and Mechanisms
复制标题

属性隐私:框架和机制

DOI:
10.1145/3531146.3533139
复制
发表时间:
2022
期刊:
and Transparency
影响因子:
--
通讯作者:
Cummings, Rachel
Cummings, Rachel
中科院分区:
--
文献类型:
--
作者:
Zhang, Wanrong;Ohrimenko, Olga;Cummings, Rachel

文献摘要

参考文献

被引文献

相似文献

确保训练数据的隐私性越来越受到关注,因为许多机器学习模型都是在机密和潜在敏感数据上训练的。在大型数据集的分析过程中,保护个人隐私的方法受到了很大的关注。然而,在许多设置中,数据集的全局属性也可能是敏感的(例如,医院中的死亡率,而不是数据集中特定患者的存在)。在这项工作中,我们从个人隐私出发,开始研究属性隐私,其中数据所有者担心在分析过程中泄露整个数据集的敏感属性。我们提出了在两种相关情况下捕获属性隐私的定义,其中可能需要保护全局属性:(1)特定数据集的属性和(2)数据集采样的底层分布的参数。我们还提供了两个有效的机制,为特定的数据分布和一个一般的,但效率低下的机制,满足这些设置的属性隐私。我们的研究结果基于河豚框架的一种新颖而非平凡的使用,以解释数据中属性之间的相关性,从而解决了Kifer和Machanavajjhala在2014年留下的“开发河豚实例和一般聚合秘密算法的挑战性问题”。
Ensuring the privacy of training data is a growing concern since many machine learning models are trained on confidential and potentially sensitive data. Much attention has been devoted to methods for protecting individual privacy during analyses of large datasets. However in many settings, global properties of the dataset may also be sensitive (e.g., mortality rate in a hospital rather than presence of a particular patient in the dataset). In this work, we depart from individual privacy to initiate the study of attribute privacy, where a data owner is concerned about revealing sensitive properties of a whole dataset during analysis. We propose definitions to capture attribute privacy in two relevant cases where global attributes may need to be protected: (1) properties of a specific dataset and (2) parameters of the underlying distribution from which dataset is sampled. We also provide two efficient mechanisms for specific data distributions and one general but inefficient mechanism that satisfy attribute privacy for these settings. We base our results on a novel and non-trivial use of the Pufferfish framework to account for correlations across attributes in the data, thus addressing “the challenging problem of developing Pufferfish instantiations and algorithms for general aggregate secrets” that was left open by Kifer and Machanavajjhala in 2014 [15].
DOI: 10.1007/978-3-030-29959-0_7
发表时间: 2019
期刊: Proc. of the 24th European Symposium on Research in Computer Security (ESORICS 2019), Part I, Lecture Notes in Computer Science
影响因子: --
作者:
Kohei Nishino;Takashi Kitamura;Tomoji Kishi;Cyrille Artho;Yutaka Matsuno;Yusuke Kawamoto and Takao Murakami
通讯作者: Yusuke Kawamoto and Takao Murakami
关于无界密度函数 $infty $-Wasserstein 距离的经验测度的收敛速度
DOI: 10.1090/qam/1541
发表时间: 2019
影响因子: 0.8
作者:
Liu, Anning;Liu, Jian-Guo;Lu, Yulong
通讯作者: Lu, Yulong
DOI: 10.1145/2514689
发表时间: 2014-01-01
影响因子: 1.8
作者:
Kifer, Daniel;Machanavajjhala, Ashwin
通讯作者: Machanavajjhala, Ashwin
DOI: 10.1007/978-3-642-25385-0_12
发表时间: 2011-12
期刊: IACR Cryptol. ePrint Arch.
影响因子: --
作者:
Raghav Bhaskar;Abhishek Bhowmick;Vipul Goyal;S. Laxman;Abhradeep Thakurta
通讯作者: Raghav Bhaskar;Abhishek Bhowmick;Vipul Goyal;S. Laxman;Abhradeep Thakurta
DOI: --
发表时间: 2006
期刊: ACM SIGACT-SIGMOD-SIGART Symposium on Principles of Database Systems
影响因子: --
作者:
Ashwin Machanavajjhala;J. Gehrke
通讯作者: J. Gehrke