Anonymization in PPDM based on data distributions and attribute relations

Abstract

Objectives: Privacy Preserving Data Mining techniques deal with the secure data publication or communication without revealing the private and sensitive information about any individual. Anonymization technique has been considered as one of the most effective techniques since it can provide better tradeoffbetween data utility and privacy preservation. Methods/Statistical Analysis: Existing anonymization techniques works on individual attributes and their cardinalities and they do not consider the relations among different attributes of the data. In this paper we have considered auxiliary information and entropy and mutual information to calculate distribution of entities in an attribute and relations among different attributes respectively. Based on these calculations we shall be analyzing the best generalization level for data anonymization. Findings: An adverse user can analyze the data with numerous possible perspectives viz. auxiliary information, theoretical and manual data analysis and try to exploit the data vulnerability, so improved data privacy can be achieved if we could also see with the adversary eyes. Applications/Improvements: Different other techniques can be applied to find distribution and relations on the basis of data background and its area of application.