PRIVACY PRESERVING CLUSTERING BY ADDING DIFFERENT NOISE COMPONENTS GENERATED FROM PROGRESSIONS TO DIFFERENT CONFIDENTIAL ATTRIBUTES

摘要

Privacy Preserving Data mining has been emerged as the one of the most prominent research areas in recent days. In this paper, we proposed a method for privacy preserving clustering of data. The proposed method considers a relational table and then identifies the confidential and non confidential attributes in the relational table so that non confidential attributes are removed and only the confidential attributes are retained in the table. The values of different confidential attributes are perturbed by adding different noise values generated from progressions. Then cluster analysis is performed on the original data and perturbed data using K-means algorithm with varying number of clusters. The results obtained show that the mean squared error obtained from the original data is same as the mean squared error obtained from the perturbed data but the order of values differ due to the random selection of cluster centers.