An effective privacy preserving mechanism for 1: M microdata with high utility

Abstract

Privacy-preserving data publishing (PPDP) aims at providing an anonymized view of a private microdata to the recipients, e.g., researchers, pharmaceutical companies etc. This private data contains sensitive information about individuals that needs to be protected. In the literature, it is generally assumed that there exists a single record for one individual in any given microdata (1:1 dataset). However, more practically, there are many instances in which an individual can have multiple records in microdata (termed as 1: M datasets). Several techniques have been proposed for the 1:1 microdata but, a few researchers paid attention towards 1:M microdata problems, that perhaps led to new privacy disclosures. A novel privacy model (k, l)-diversity was proposed to cater such disclosure risks and based on this model, an algorithm named 1: M generalization was proposed. Although it was efficient than several other techniques; still has a drawback of huge information loss. In this paper, we propose a hybrid approach named as l-anatomy for 1: M microdata and prove that l-anatomy ensures the privacy of given individuals. Also, experiments performed on two real-world datasets (namely INFORMS and YOUTUBE) reveal that the proposed scheme exhibits higher efficiency and effectiveness as compared to its counterpart.