Comparative study of distance metrics for t-closeness

Abstract

In our present technical world we all have to submit our personal information to various organisations driven by mutual benefits. The collected data has to be published. There was a need for exchange of the published data between different parties. Given data in its original form contains sensitive information of the person. If the given data is published directly sensitive information is revealed to others which violates the privacy of individual directly. In order to publish the data without violating one’s personal privacy we use a technique called t -closeness. In this method the metric used was Earth Mover’s Distance(EMD). But this metric does not satisfies probability scaling property which makes it not to reflect the difference between the probabilities. This make EMD to produce inaccurate results which may increase the anonymization. In order to have a metrics that satisfies all the distance metric properties including probability scaling property we make a study on different metrics like Squared Root Jensen-Shannon,Pearson and Divergence. We compare these metric by taking different parameters like Discenibility Metrics,propensity score and precision.