Privacy Protection Association Rules Based On Critical Band

With data mining technology becoming popular today, other problems increaseprominent. Especially in2011, after the disclosure of China’s Internet leaks their user’sinformation, people began to think more about the data privacy issues which have beenignored before.As most favored applications, association rule mining has been used in practical alot. Correspondingly, the research of pricacy protection is carried out earliest andachieved a lot.In this thesis, it is mainly based on the processing algorithm of FP-TREE and thehandling criteria of Privacy critical interval and the association factor, and protects thosesensitive association rules. The processing framework is divided into three stages: datapreparation stage, to hide the sensitive information stage and reconstruction of data setsstage.In the data preparation stage, the source data converted into FP-TREE.In the stageof hide sensitive information, algorithms are proposed on the basis of the FP-TREE tooperate, and the specific operation is divided into two ways--add items sets and reduceitems sets. Meanwhile, this paper takes the reconstruction of the transaction data set onFP-TREE as the overall algrithm framework. During the reconstruction we randomlyadd the non-sensitive items and pure sensitive items into the published data.The biggest difference between the algorithm in this paper and the previousalgorithms is that rather than focus on the frequent itemset, we mainly consider therelatinships among the different association rules. In this paper we analysis in detail thatin an association rule, different sets change will lead to potential change and influnce onitself. At the same time, the introduction of Privacy critical interval worked as thequantitative criteria of the degree of change. Through this approach, we can control thelost of non-sensitive information and reduce the side effects of hidden action.Finally, experimental comparative analysis shows that the algorithm proposed inthis paper complete the privacy protection for sensitive association rules very well.