Privacy Preserving in Data Stream Mining Using Multi-iterative K-Anonymization

Abstract

Currently data mining performs an energetic part in information domain where it has been extensively functional in several sectors. Sharing of this data may contain sensitive information about personal or organization. Privacy preservation is a crucial and important concern in publication of own data. It has also elevated a potential stolen or revealing sensitive data of an individual or organizational when the data are publically available. Anonymization and perturbation type of privacy preserving methods have been proposed to tackle the protecting privacy problem with minimum information loss. The loss of information may affect the data quality and in great case the data may become totally unusable. We have to preserve the sensitive data in cost of minimum information loss with maximum data quality. Proposed work aims privacy of sensitive information before release while obtaining accuracy of mining results with minimum information loss using multi-iterative k-anonymization in data stream mining. An extension to k-anonymization has been proposed where privacy outcome has been computed based on Multiiterative for selective anonymization for set of un-anonymized instances. In proposed algorithm, generalisation has been applied for each quasi-identifier. Experimental results show that the proposed approach cannot only preserve data privacy, but also mine data set accurately.