Certain investigations on data perturbation techniques for privacy preserving data mining

Abstract

Data collection and analysis are increasing consistently, depending newlineon the generality and the overall availability of computers. The study of these newlineuseful patterns supports many companies and benefits businesses in different newlineareas. Storing such confidential data, however, raises some serious questions newlineabout privacy. The procedures that enable information to be extracted while newlinesafeguarding privacy are referred to as Privacy Preserving Data Mining newline(PPDM) techniques which are used for the protection of data privacy. newlineMethods of information reorientation help to achieve data protection without newlinesacrificing the power of resource extraction. Conventional methods of newlineinformation encoding can remove a series of valid structures, masking newlineconfidential trends. However, different types of information encoding newlineschemes should be developed to conceal vulnerable sequences from data sets. newlineIn the first part of the current research, the fuzzy membership newlinefunction based hybrid data distortion method is proposed for privacy newlineprotection. Numerous classification methods have been developed in the newlineexisting studies to shield statistical values using fuzzy functions. The aim of newlinethese methods is to optimise the balance between both the data protection and newlinethe data consumption. The problems including transition assuring secrecy and newlineusefulness between both original and transformed data sets are considered for newlineevaluation. The proposed hybrid method combines S-fuzzy, multiplicative newlinenoise and zero mean normalization to ensure confidentiality. Classification newlineexcellence is measured using misclassification error rates that return the newlineusefulness level of data. Extensive experiments reveal that the hybrid newlinealgorithm is more efficient, reliable and accurate than fuzzy based disruption newlinemethodology. newline newline