An Un-realized Algorithm for Effective Privacy Preservation Using Classification and Regression Trees

Abstract

In this digital age, privacy preservation has attracted much attention, as a huge amount of data are generated from multiple sources and transmitted across the Internet. Several perturbation algorithms have emerged to keep sensitive data hidden behind additive noises. In this paper, a novel un-realization algorithm is developed based on a classification and regression tree (CART). First, the sample dataset was distorted, and the duplicate elements were removed, creating a perturbed dataset and an un-realized dataset. Then, a decision tree was set up by the modified CART algorithm and another by the traditional CART based on the un-realized dataset. Finally, the Gini values of the two trees were compared. If the results are the same, then the privacy of the data is preserved. The proposed algorithm was compared with several traditional un-realization algorithms through experiments. The results show that our algorithm achieved excellent results in Gini value, time complexity and output accuracy.