Using Network Flows to Identify Users Sharing Extremist Content on Social Media.

Social media has been leveraged by many groups to share their ideas, ideology, and other messages. Some of these posts promote extremist ideology. In this paper, we propose an approach for identifying users who engage in extremist discussions online. Our approach uses detailed feature selection to identify relevant posts and then uses a novel weighted network that models the information flow between the publishers of the relevant posts. An empirical evaluation of a post collection crawled from a web forum containing racially driven discussions and a tweet stream discussing the ISIS extremist group show that our proposed method for relevant post identification is significantly better than the state of the art and using a network flow graph for user identification leads to very accurate user identification.