A hybrid-security model for privacy-enhanced distributed data mining

Abstract

This study encompasses the proposal of a novel hybrid-security model that incorporates the benefits of both the centralised data mining system as well the distributed data mining system. The hybrid model provides two levels of security. The first security level perturbs the individual datasets by transforming it into a non-understandable form using the four-dimensional rotation transformation, and the second security level helps in performing secure distributed data mining using the ratio of secure summation protocol. The trivial data mining techniques, such as k-means clustering technique and naïve Bayes classification technique, verifies the efficiency and accuracy of the hybrid security model. The hybrid security model provides security to sensitive data without compromising the quality of the data. The accuracy obtained in classification task and clustering task using naïve Bayes and k-means technique is high when different datasets in a privacy-enhanced distributed data mining environment verify the working of the hybrid security model.