A Constraint-Based Model for the Frequent Itemset Hiding Problem

Vassilios S. Verykios, Elias C. Stavropoulos, Vasilis Zorkadis, Ahmed K. Elmagarmid

Abstract: This paper introduces a novel constraint-based hiding model to drastically reduce the preprocessing overhead that is incurred by border-based techniques in the hiding of sensitive frequent itemsets. The proposed model is solved by an efficient constraint-based mining algorithm that pushes a conjunction of antimonotone constraints into an Apriori-like algorithm, for inducing the support theory of non-sensitive frequent itemsets along with its negative border. The patterns induced by the constraint-based mining algorithm can be used in border-based hiding algorithms to construct a sanitized version of the original database, where the sensitive knowledge is concealed. The efficiency of the constraint-based mining algorithm is evaluated on real and synthetic datasets.