Protection of micro-data subject to edit constraints against statistical disclosure

Abstract

Before releasing statistical outputs, data suppliers have to assess if the privacy of the statistical units is endangered and apply Statistical Disclosure Control (SDC) methods if necessary. SDC methods perturb, modify or summarize the data, depending on the format for releasing the data, whether as micro-data or tabular data. The goal is to choose an optimal method that manages disclosure risk below a tolerable risk threshold while ensuring high utility and high quality statistical data. In this article we first overview several SDC methods for continuous and categorical micro-data. All the methods perturb the data in some way. Changing values, however, will cause fully edited records in micro-data to fail edit constraints (i.e., logical rules or edits), resulting in low utility data. Moreover, an inconsistent record will signal it as having been perturbed for disclosure control and attempts can be made to unmask the data. In order to deal with these problems, we develop new implementation methods for the perturbation and minimize record level edit failures as well as overall measures which assess information loss and utility. This is done by perturbing within control strata and imputing for failed edits, ensuring additivity constraints, and preserving totals, means and covariance matrices.

More information

Identifiers

Catalogue record

Export record