Optimisation-based study of data privacy by using pram

Abstract

Dissemination of data with sensitive information has an implicit risk of unauthorised disclosure. Several masking methods have been developed in order to protect the data without losing too much information. One of the methods is called the Post Randomisation Method (PRAM) which is based on perturbations according to a Markov probability transition matrix. However, the method has the drawback that it is difficult to find an optimal transition matrix to perform perturbations which maximise data utility. In this paper we present an study of data privacy from the point of view of optimisation using evolutionary algorithms to generate optimal probability transition matrices. Optimality is with respect to a pre-defined fitness function which aims to preserve several data protection properties such as data utility and disclosure risk. We also provide experimental results using real datasets in order to illustrate and empirically evaluate the application of this technique.