Privacy Information Leakage Prevention in Cognitive Social Mining Applications

Abstract

Online social networks, such as Facebook are increasingly used by many users and these networks allow people to publish and share their data to their friends. The problem is user privacy information can be inferred via social relations. This chapter makes a study and performs research on managing those confidential information leakages which is a challenging issue in social networks. It is possible to use learning methods on user released data to predict private information. Since the main goal is to distribute social network data while preventing sensitive data disclosure, it can be achieved through sanitization techniques. Then the effectiveness of those techniques is explored, and the methods of collective inference are used to discover sensitive attributes of the user profile data set. Hence, sanitization methods can be used efficiently to decrease the accuracy of both local and relational classifiers and allow secure information sharing by maintaining user privacy.