Disclosing users' data in an environment that preserves privacy

The conflict between Web service personalization and privacy is a challenge in the information society. In this paper we address this challenge by introducing MASKS, an architecture that provides data on the users' interests to Web services, without violating their privacy. The proposed approach hides the actual identity of users by classifying them into groups, according to their interests exhibited during the interaction with a Web service. By making requests on behalf of a group, instead of an individual user, MASKS provides relevant information to the Web services, without disclosing the identity of the users. We have implemented and tested a grouping algorithm, based on categories defined by the semantic tree of DMOZ. We used access logs from actual e-commerce sites to evaluate the grouping algorithm. Our tests show that 64% of the requests made to the e-commerce service could be grouped into meaningful categories. This indicates that the e-commerce sites could use the information provided by MASKS to do personalization of services, without having access to the individual users in the groups.