Inferring Private Attributes in Online Social Networks

Abstract

Online social networks (OSNs) are playing an important role in current world and the way people communicate with each other. Despite the advantage of using online social networks, there are certain privacy risks that can affect users of such services. Since users provide a lot of personal information in OSNs, concerns about how data placed in online social networks may raise among the users. Social networking sites have responded to these concerns by introducing privacy filters to their site, allowing users to specify which aspects of their profile are visible to whom. Such privacy settings is not effectively used by half of the OSN users based on our analysis and we collect large number of public profile information from the well-known social network Hyves.nl in the Netherlands. We then show that public friendship links of a person can expose different attributes about him. Based on friendship links we are able to infer and predict some of the attributes of a user with good accuracy.