Blocking Schemes

An interesting line of research in the class of heuristic methodologies for association rule hiding regards approaches that block certain original values of a database by introducing

unknowns

(i.e., by replacing the real values of some items in selected transactions of the original database with question marks “?”). Unlike data distortion schemes, blocking methodologies do not add any false information to the original database and thus provide a much safer alternative for critical real life applications, where the distinction between “false” and “unknown” can be vital. Due to the introduction of unknowns, the support and the confidence of association rules that are mined from the sanitized database becomes fuzzified to an interval (rather than being an exact value) and can no longer be safely estimated. Blocking approaches, similarly to distortion approaches, can be partitioned into support-based and confidence-based, depending on whether they use the support or the confidence of the association rules to drive the hiding process. In this chapter, we review two methodologies that belong to this class of approaches.