Low Dimensional Data Privacy Preservation Using Multi Layer Artificial Neural Network

Abstract

Government agencies, business enterprises and non-profit organizations are searching for innovative methods to collect and analyze data about individuals or businesses to support their decision making processes. Data mining techniques are able to derive sensitive knowledge from unclassified data, causing a severe threat to privacy. The authors provide a promising solution to address the demand for privacy preservation in clustering analysis. They propose a novel dimensionality expansion based data privacy preservation technique using multi-layer artificial neural network. By applying this idea, the authors can project a low dimensional data into a high dimensional space to enhance the privacy level. Clustering was done using K-means and the results show that privacy level and the nature of data were very much preserved even after this transformation. The results arrived at were significant and the proposed method transformed the data better than the classical Geometric data transformation based methods.