Utilization of Imprecise Rules for Privacy Protection

摘要

In this paper, we utilize imprecise rules for privacy protection in the publication of data sets. We assume that data sets show the classification results with more than two classes. First k-anonymous imprecise rules are induced. Using several k-anonymous imprecise rules explaining an object, the object is replaced with several imprecise patterns corresponding to the k-anonymous imprecise rules which are common in at least k objects. In this way, privacy protected data tables are composed. The proposed data table is investigated by numerical experiments from its usefulness in rule induction as well as from its privacy protection ability. The results show that the proposed method will be satisfactorily useful.