Max-Min Algorithms

Abstract

Moustakides & Verykios in [50, 51] propose two border based methodologies that rely on the max–min criterion for the hiding of the sensitive itemsets. Both methodologies use the revised positive border of the frequent itemsets to track the impact of each tentative item modification that helps towards the hiding of a sensitive itemset. Then, they select to apply those item modifications to the original database that effectively conceal all the sensitive knowledge, while minimally affecting the itemsets of the revised positive border and, consequently, the nonsensitive frequent itemsets. For each item of a sensitive itemset, the Max–Min algorithms identify the set of itemsets from the revised positive border which depend on it, and select among them the ones that are supported the least. Then, from among all minimum supported border itemsets (coming from the previously computed sets for the different items of the sensitive itemset), the itemset with the highest support is selected as it is the one with the maximum distance from the borderline that separates the frequent from the infrequent itemsets.