Modeling projections in microaggregation

Abstract

Microaggregation is a method used by statistical agencies to limit the disclosure of sensitive microdata. It has been proven that microaggregation is an NP-hard problem when more than one variable is microaggregated at the same time. To solve this problem in a heuristic way, a few methods based on projections have been introduced in the literature. The main drawback of such methods is that the projected axis is computed maximizing a statistical property (e.g., the global variance of the data), disregarding the fact that the aim of microaggregation is to keep the disclosure risk as low as possible for all records. In this paper we present some preliminary results on the application of aggregation functions for computing the projected axis. We show that, using the Sugeno integral to calculate the projected axis, we can reduce in some cases the disclosure risk of the protected data (when projected microaggregation is used).