Detecting Social Cliques for Automated Privacy Control in Online Social Networks

Abstract—As a result of the increasing popularity of online social networking sites, millions of people spend a considerable portion of their social life on the Internet. The information exchanged in this context has obvious privacy risks. Interestingly, concerns of social network users about these risks are related not only to adversarial activities but also to users they are directly connected to (friends). In particular, many users want to occasionally hide portions of their information from certain groups of their friends. To satisfy their users ’ needs, social networking sites have introduced privacy mechanisms (such as Facebook’s friend lists) that enable users to expose a particular piece of their information only to a subset of their friends. Unfortunately, friend lists need to be specified manually. As a result, users frequently do not use these mechanisms, either due to a lack of concern about privacy, but more often due to the large amount of time required for the necessary setup and management. In this paper, we propose a privacy control approach that addresses this problem by automatically detecting social cliques among the friends of a user. In our context, a social clique is a group of people whose members share a significant level of social connections, possibly due to common interests (hobbies) or a common location. To find cliques, we present an algorithm that, given a small number of friends (seed), uses the structure of the social graph to generate an approximate clique that contains this seed. The cliques found by the algorithm can be transformed directly into friend lists, making sure that a piece of sensitive data is exposed only to the members of a particular clique. Our evaluation on the Facebook platform shows that our method delivers good results, and the cliques that our algorithm identifies typically cover a large fraction of the actual social cliques. I