Integrating Discrimination Prevention and Privacy Preservation into Data Mining

Abstract: In information society, massive and automated data collection is required for different purposes in our daily life. There are mainly two threats for individuals whose information is published: privacy and discrimination. In data mining, decision models are mainly derived on the basis of records stored by means of various data mining methods. But there may be a risk that the extracted knowledge imposes discrimination. Many organizations collect a lot of data for decision making. The sensitive information of the individual whom the published data relate to, may be revealed, if the data owner publishes the data directly. Discrimination prevention and privacy preservation need to be ensured simultaneously in decision making process. In this paper, Discrimination Prevention Data Mining (DPDM) and Privacy Preservation Data Mining (PPDM) have been studied and their relationships have been explored. Different privacy models and its impact on the data have also been analysed. Keywords: Anti-discrimination, discriminatory attribute, classification rule, rule generalization, rule protection, k-anonymity.