Privacy Preserving Data Publishing for Heterogeneous Multiple Sensitive Attributes with Personalized Privacy and Enhanced Utility

Abstract

In recent years, personal data availability has become vast, which leads to the concept of Privacy-preserving. Privacy-Preserving is an essential issue in all research fields. Many privacy methods are available for privacy-preserving data publishing (PPDP); however, it suffers from a few drawbacks i) couldn't use on heterogeneous multiple sensitive attributes; ii) Customized sensitivity requirements are ignored. To make the model satisfy both criteria, we have proposed a Quasi-Identifier-Multiple heterogeneous sensitive attribute (QI-MHSA) generalization algorithm. Our first work in this paper is to apply vertical partitioning in the microdata and partitioning it into i) Quasi-identifier bucket (QIB) ii) Multiple heterogeneous sensitive attribute bucket (MHSAB). Second, we have applied k-anonymity in QIB to anonymize the quasi-identifiers and l-diversity in MHSAB to anonymize the different sensitive attributes (categorical and numerical). A Top-down generalization method is adopted to generalize the categorical and numerical attributes. Finally, a new approach has been implemented in the personalized privacy of sensitive attributes. A flag is set for both categorical and numerical sensitive attributes based on their sensitivity requirements in MHSAB. The generalization approaches differ according to the level of sensitivity requirement. Extensive implementation is done on two datasets to compare the algorithm's efficiency and prove that our model has a better balance between privacy and utility.