Detecting and preventing inference attacks in online social networks: A data-driven and holistic framework

ABSTRACT With increasing user involvement, social networks nowadays serve as a repository of all kinds of information. While there have been various studies demonstrating that private information can be inferred from social networks, few have taken a holistic view on designing mechanisms to detect and alleviate the inference attacks. In this study, we present a framework that leverages the social network data and data mining techniques to proactively detect and prevent possible inference attacks against users. A novel method is proposed to minimize the modifications to user profiles in order to prevent inference attacks while preserving the utility.