Privacy preserving C4.5 using Gini index

Now-a-days privacy has become a major concern; the goals of security like confidentiality, integrity and availability do not ensure privacy. Data mining is a threat to privacy. Researchers today focus on how to ensure privacy while performing data mining task. As Data mining algorithms are typically complex and furthermore the input usually consists of massive data sets, the generic protocols in such a case are of no practical use and therefore more efficient protocols are required. This paper focus on the problem of decision tree learning with the popular C4.5 algorithm. C4.5, an extension of ID3 is a very popular decision tree building method in data mining. Entropy and Gini index are two different criteria used in ID3. While there is quite little work in privacy preserving ID3 using entropy and not much has been done for Gini index. This paper propose modified protocols based on secure multiparty computation for privacy preserving C4.5 using Gini index over distributed partitioned data, where the protocols do not require any third party server. However, some communication overhead is necessary so that the parties can carry out the secure protocols. The result like ROC(Receiver Operating characteristic) graph and detail accuracy through cost counting index is shown.