Transformation approach for boolean attributes in privacy preserving data mining

Abstract

Data Mining can be seen as the process of extracting hidden patterns from large databases. In many situations, the extracted patterns can reveal private information that should not be disclosed. There is a need to develop techniques that can extract hidden information without disclosing private data. Need of such techniques give rise to the new research direction in data mining that is privacy preserving data mining (PPDM). Many techniques for preserving privacy in data mining have been developed over the last decade such as cryptographic, randomization methods, k-anonymity, l-diversity etc. But these techniques can affect the accuracy of results and may result in the loss of information. Transformation based techniques were proposed in literature that can preserve the privacy by maintaining the information and accuracy. Transformation techniques were proposed for transforming numerical and categorical sensitive attributes. Many algorithms exist in the literature to transform sensitive attributes of numeric data type. But there is no technique for dealing with sensitive attributes of Boolean data type so that Boolean attributes do not disclose any private information without compromising data mining results. Our aim is to develop a technique for transforming sensitive Boolean attributes.