支持多属性泛化的个性化(<span>α</span>,l,k)匿名模型-《计算机技术与发展》

文章信息/Info

Title:: A Personalized ( α,l,k) Anonymous Model of Supporting Multi-attribute Generalization

Author(s):: SU Lin-ping; DONG Zi-xian; LI Wei; WU Ke-he; CUI Wen-chao; School of Control and Computer Engineering,North China Electric Power University,Beijing 102200,China

Keywords:: k -anonymity; personalized privacy protection; generalization; sensitivity score; sensitive level

摘要:: 传统的个性化数据匿名模型一般可以分为两种机制:一种是面向个人的,一种是面向敏感值的。这两种方法一般都会因为追求敏感数据的个性化保护而过度泛化,造成大量的信息损失,使数据的可用性下降。为此,该文提出了一种个性化( α,l,k ) 匿名隐私保护模型。该模型有效结合了这两种传统的数据匿名机制,在最大程度地保证个性化匿名的需求下,根据敏感属性值敏感等级的不同,对各个等价组中的敏感属性值分别采取不同的匿名方式,优先泛化高敏感度等级的属性值,使等价组中的每个敏感属性满足对出现频率琢以及多样性 l 的约束条件,从而有效降低数据集中高敏感等级信息的泄露风险,并可以提高数据的可用性。实验结果表明,该模型能够在有限的运行时间内,相较其他个性化匿名模型有更低的信息损失量和更好的隐私数据保护能力。

Abstract:: The traditional personalized data anonymity model can be divided into two mechanisms:one is personal-oriented,and the other is sensitive value oriented. In general,these two methods tend to over-generalize due to the pursuit of personalized protection of sensitive data,resulting in a large number of information losses and declining availability of data. To this end,we propose a personalized (α,l,k )anonymous privacy protection model,which effectively is a combination of these two kinds of traditional data anonymous mechanism.Under the need of ensuring personalized anonymity in maximum extent,according to different sensitive attribute value level,the sensitive attribute values of each equivalent group respectively take different anonymously, priority generalization properties of high level of sensitivity,which makes each sensitive attribute in equivalent group meet the constraint conditions of occurrence frequency?α and diver sityl ,so as to effectively reduce the leakage risk of high-sensitive information in the data set,and improve the availability of data. The experiment shows that this model has lower information loss and better privacy data protection capability than other personalized anonymous models in limited running time.