基于条件函数依赖的隐私保护模型

Privacy-preserving model on data with conditional functional dependency

数据拥有者发布的数据中如果包含条件函数依赖会导致数据的隐私受到攻击,由条件函数依赖产生的属性间的关联会带来潜在的隐私泄露问题。针对现有的隐私保护方法均无法保护包含条件函数依赖的数据的隐私,形式化地定义了基于条件函数依赖的隐私攻击,提出了隐私保护模型l-deduction来对包含条件函数依赖的数据进行隐私保护;并设计了相应的匿名算法来实现l-deduction模型。理论分析和实验结果表明,该方法既能保护包含条件函数依赖的数据的隐私,又具有较小的信息损失度。

There exists privacy threat in the publishing data that contains CFDs.This paper showed that the correlations among attributes by CFDs could bring potential vulnerability to privacy.This paper formalized the CFD-based privacy attack,proposed the privacy model l-deduction and developed an algorithm that achieved l-deduction.Beyond a theoretical treatment of the problem,the experiments also demonstrate the effectiveness of the proposed technique.