Variable Distinct l-diversity Algorithm Applied on Highly Sensitive Correlated Attributes

摘要

In this information age, large amount of data is available online. These data are used by both internal and external sources for analysis and research purposes. The collected data is stored into huge data sets containing sensitive and Non—Sensitive Attributes. For the reason that attributes are generally separated, the correlation between these various attributes is lost. Thus, it will be necessary to prevent attributes from losing the correlation between them or at least reduce the correlation loss. As a solution, correlated attributes are grouped together. Although, the data utility is preserved by reducing the correlation loss between Sensitive Attributes, privacy protection remains a serious concern. The main problem here is publishing data sets without revealing the sensitive information of individuals and in the same time preserving data utility. Most of the current researches on ensuring privacy in big data are centered on data anonymization. L-diversity is an anonymization technique that can be applied on a data set with one or multiple Sensitive Attributes. This paper proposes an algorithm that deals with sensitive numerical and non—numerical attributes. The algorithm applies the principle of l-diversity technique after grouping highly correlated attributes together through a vertical partitioning. Our proposed algorithm makes a balance between privacy and data utility.