Enhancing the Performance of 3D Rotation Perturbation in Privacy Preserving Data Mining Using Correlation Based Feature Selection

Abstract

A large amount of valuable data is being produced every day with the development of technologies. To retrieve knowledge and information from these data, mining and analysis are mandatory. But, the data may contain sensitive information of the individuals like medical diagnostic reports which they do not want to expose. Privacy preserving data mining, i.e., PPDM can help in this issue keeping the sensitive information private as well as preserving the data utility. Rotation-based perturbation technique contributes to satisfying both aspects of PPDM, i.e., individuals’ privacy and data utility besides other PPDM techniques. In this work, we proposed a way for generating the triplet (set of three features) for 3D rotation perturbation technique using correlation among the features. This triplet generation is a fundamental step in 3D rotation perturbation technique. The analysis of information entropy, privacy protection and utility analysis elucidates that correlation-based triplet generation provides better data privacy and utility than existing triplet generation for 3D rotation perturbation technique.