Probabilistic dimension reduction method for privacy preserving data clustering

Abstract

The frequent use of data mining techniques in business organizations is useful to sustain competition in real world but it leads to violation of privacy of individual customers while publishing original customer’s data into real world. This paper proposes a distance preserving perturbation method for Privacy Preserving Data Mining (PPDM) using t-Stochastic Neighbor Embedding (t-SNE). The t-SNE is mainly used for dimension reduction technique; it reduces higher dimensional data sets into required lower dimensional data sets and maintains same distance in lower dimensional data. We choose K-means algorithm as knowledge-based technique; it works based on distance between data records. Setting perplexity parameter in the proposed method creates complexity to unauthorized persons to convert from low-dimensional data to original data. The proposed method is evaluated using Variation Information (VI) between original and modified data clusters. In this work, the proposed method is applied on various data sets and compared original and modified data clusters through VI.