Tools and Strategies to Protect Multiple Tables with the GHQUAR Cell Suppression Engine

Recent advances in optimization techniques for statistical tabular data protection

“…However, most of them have been included in the τ -Argus software (Hundepool et al, 2008), freely available from http://neon.vb.cbs.nl/casc/tau.htm, and being used by several European NSAs. In particular, this package includes the Benders decomposition of Fischetti and Salazar-González (2001), the shortest paths heuristic of Castro (2007a), and the two (infeasible) heuristics of deWolf (2002) and Giessing and Repsilber (2002). These four approaches can be tested with the same set of tables, generated within the τ -Argus common interface.…”

Cell suppression problem: A genetic-based approach

“…The method was designed to guarantee a sliding protection range for each confidential cell: a lower bound is calculated for the width of the cell's protection range, resulting from the suppression of the corner points of each hypercube, and that suppression is considered feasible if it turns out that the bound is sufficiently large. Suggestions on how to set the width of the protection ranges, to obtain a feasible set of suppressions, can be found in [17]. The adaptation of the hypercube method to the a priori fixed interval protection criterion has no theoretical difficulty.…”

“…The hypercube method, as described in [17], is a fast heuristic developed to find sets of complementary suppressions to protect confidential data in large complex tables. It decomposes complex tables into subtables, without substructure, and iteratively protects the subtables.…”

Statistical Disclosure Control in Tabular Data

“…The hypercube [25], initially developed for k-dimensional tables, is a simple and fast procedure. For two-dimensional tables it can be seen as a network flows approach that only considers a subset of the flows (thus providing less quality solutions than heuristics based on network optimization).…”

“…However, they can be run with the same table from the τ-Argus package, which implements four of them: the optimal approach of [24], the shortest paths heuristic of [9], and the two (infeasible) heuristics of [19] and [25]. To compare them, in [26] a toy table 1H2D was generated with τ-Argus.…”