Privacy Preserving for Tagging Recommender Systems

Abstract

Tagging recommender systems offer users the possibility to annotate resources with personalized tags so as to enable users to easily find suitable tags for a resource. They combine the advantages of automation in traditional recommender systems and flexibility of tagging systems. A large collection of data has been generated by those social network web sites with tagging recommender systems during the last few years, and the issue of privacy in the recommender process has generally been overlooked. An adversary with background information may re-identify a particular user in a tagging dataset and obtain the user’s historical tagging records. Compared to general recommender systems, the privacy problem in tagging recommendation systems is more complicated due to its unique structure and semantic content. In this chapter, we will focus on the dataset releasing for tagging recommender systems and utilize differential privacy to prevent the leaking of private information when releasing the dataset. A private tagging release algorithm is presented in this chapter to provide comprehensive privacy-preserving capability for individuals and maximizing the utility of the released dataset. The algorithm offers a tailored differential privacy mechanism that optimizes the performance of recommendation with a fixed level of privacy.