Motivated by the insufficiency of the existing quasi-identifier/sensitive-attribute (QI-SA) framework on modeling real-world privacy requirements for data publishing, we propose a novel versatile publishing scheme with which privacy requirements can be specified as an arbitrary set of privacy rules over attributes in the microdata table. To enable versatile publishing, we introduce the Guardian Normal Form (GNF), a novel method of publishing multiple sub-tables such that each sub-table is anonymized by an existing QI-SA publishing algorithm, while the combination of all published tables guarantees all privacy rules. We devise two algorithms, Guardian Decomposition (GD) and Utility-aware Decomposition (UAD), for decomposing a microdata table into GNF, and present extensive experiments over real-world datasets to demonstrate the effectiveness of both algorithms.
展开▼
机译：由于现有的准标识符/敏感属性（QI-SA）框架不足以对数据发布的真实世界的隐私要求进行建模，我们提出了一种新颖的通用发布方案，可以将隐私要求指定为任意一组微数据表中属性的隐私规则。为了实现广泛的发布，我们引入了Guardian Normal Form（GNF），这是一种发布多个子表的新颖方法，每个子表都可以通过现有的QI-SA发布算法来匿名化，而所有已发布表的组合可确保所有隐私权规则。我们设计了两种算法，用于将微数据表分解为GNF的Guardian分解（GD）和实用程序感知的分解（UAD），并在现实世界的数据集上进行了广泛的实验，以证明这两种算法的有效性。
展开▼
