A Tutorial on Computing $t$-Closeness

Abstract

This paper presents a tutorial of the computation of $t$-closeness. An established model in the domain of privacy preserving data publishing, $t$-closeness is a measure of the earth mover's distance between two distributions of an anonymized database table. This tutorial includes three examples that showcase the full computation of $t$-closeness in terms of both numerical and categorical attributes. Calculations are carried out using the definition of the earth mover's distance and weighted order distance. This paper includes detailed explanations and calculations not found elsewhere in the literature. An efficient algorithm to calculate the $t$-closeness of a table is also presented.