Decomposition+: Improving ℓ-diversity for Multiple Sensitive Attributes

Abstract

In this paper, we analyse existing privacy-transformation techniques in the field of PPDP that anonymize datasets with Multiple Sensitive Attributes (MSA). Of these, we present an analysis of Decomposition, an algorithm which generates a dataset with distinct ℓ-diversity over MSA using a partitioning approach. We discuss some improvements which can be made over Decomposition: in the realms of its running time, its data utility, and its applicability in the case of Multiple Release Publishing. To this effect, we describe Decomposition+ an algorithm that implements some of these improvements and is thus more suited for use in real-life scenarios. © Institute for Computer Sciences, Social Informatics and Telecommunications Engineering 2012.