A MIXED MICRO-MACRO APPROACH TO STATISTICAL DISCLOSURE CONTROL FOR MACRODATA

National Statistics Offices, Central Banks, and any other organisms and agencies producing statistical information, disseminate data so that the individual information is sufficiently protected. At the same time, those entities aim at providing society with as much information as possible under this restriction. There is some contradiction between these two purposes, since high utility information is not always possible if one has to ensure data security against unauthorized accesses. Post-tabular techniques generate safe tables through non perturbative methods (such as cell suppression) or perturbative methods (such as rounding). Despite its effectiveness, these techniques prevent users from making a more detailed statistical analysis since the published data doesn’t have the desired similarity with real values. For instance, cell suppression hides non-sensitive cells, leading to higher losses of information while perturba-tive methods may conceal the reality. In this paper we propose a new post-tabular perturbative method which applies mathematical restrictions directly on the respondents within each sensitive cell and computes safe values. Since this method focuses on respondents, it is possible to identify sensitive cells that don’t represent disclosure risk. The comparative study between this technique and others commonly used, shows significant improvements in the data utility, keeping a low risk level.