On Analysis of Mixed Data Classification with Privacy Preservation

Abstract

Privacy-preserving data classification is a pervasive task in privacy-preserving data mining (PPDM). The main goal is to secure the identification of individuals from the released data to prevent privacy breach. However, the goal of classification involves accurate data classification. Thus, the problem is, how to accurately mine large amount of data for extracting relevant knowledge while protecting at the same time sensitive information existing in the database. One of the ways is to anonymize the data set that contains the sensitive information of individuals before getting it released for data analysis. In this paper, we have mainly analyzed the proposed method Microaggregation based Classification Tree (MiCT) which use the properties of decision tree for privacy-preserving classification of mixed data. The evaluations are done based on various privacy models developed keeping in mind the various situations which may arise during data analysis.

Keywords:Microaggregation, decision tree, mixed data, data perturbation, classification accuracy, anonymous data.