Privacy Preserving in Big Data Analysis for Sensitive Data

Abstract

Protecting data privacy has become a versatile task with the introduction newlineof sophisticated data analysis tools and advanced data mining techniques. To this newlineend, many data privacy protection methods have been developed in order to newlineprevent sensitive data disclosure. Being a powerful privacy enhancing newlinemechanism, anonymization dictates the datasets need to be anonymized before newlinebeing shared with third parties. Several privacy preservation methods have been newlinedesigned by numerous research persons to safeguard sensitive data disclosure. A newlinemachine learning privacy preservation method has been designed for privacy newlinepreservation of big healthcare data in case of a high level of anonymization. But, newlinethe accuracy of privacy preservation and information loss rate involved during newlinethe privacy preservation was not considerably focused. With this, four novel newlinemethods are proposed in this research for enhancing the privacy preservation of newlinebig healthcare data with better accuracy and lesser information loss rate. newlineIn the first phase of research work, Distinctive Context Sensitive and newlineHellinger Convolutional Learning (DCS-HCL) method is developed to ensure newlinethe privacy preservation of big healthcare datasets. Distinctive Impact Context newlineSensitive Hashing model is employed to find the distinctive and impact values. newlineThen, the analogous QI-classes are mapped to develop the efficient anonymized newlinedata. After that, Hellinger Convolutional Neural Privacy Preservation is applied newlineto safeguard the privacy of healthcare data. With this, the accuracy is improved newlineand information loss is minimized. newlineIn the second phase of the research work, Evolutionary tree-based quasiidentifier and federated gradient (ETQI-FD) method is introduced to preserve newlinethe privacy of big data. ETQI-FD method is designed with the novelty of newlineevolutionary tree-based indexed quasi identification model and federated newlineadaptive Lorentz privacy preservation algorithm. By using evolutionary treebased indexed quasi identification model, the quasi identifiers are determined for newline