Preserving Micro Data Release: Categorical and Numerical Data

Data mining techniques, in spite of their benefit in a wide range of applications have also raised threat to privacy and data security. All the attributes in a data base table can be classified into three categories as identifying attributes, sensitive attributes and quasi-identifier attributes. KAnonymity is the popular approach for privacy preserving data mining and the problems with Kanonymity were overcome by the techniques like l-diversity and tcloseness. Even though, all these techniques increase privacy, they account for too much of information loss. Also the computational complexity of these techniques is high. The privacy problem is addressed by fuzzy based approach for numerical attributes and taxonomy tree based mapping table approach for categorical attribute. The Privacy Level, Disclosure Level values given by the individual allows personalized privacy preservation. Also fuzzy treatment to quasi identifier attribute avoids linking attacks. The proposed Fuzzy and mapping table based, privacy preserving publication of data, requires less computational effort but preserve information, compared to perturbation methods and other KAnonymity methods.