Detecting Users Who Share Extremist Content on Twitter

Identifying extremist-associated conversations on social media sites and blog forums is still an open problem. Extremist groups leverage social media to (1) spread their message and (2) gain recruits. In this chapter, we look at different work in this arena, focusing on metrics and features that researchers have proposed as proxies for misbehavior on Twitter. We begin this chapter by analyzing potential features a small amount of manually labeled data about ISIS supporters on Twitter. We then group these features into categories related to tweet content, viewpoints, and dynamics. After discussing different state of the art methods for extremism detection and similar problems, we present a case study looking at the ISIS extremist group. Finally, we discuss how one collects these data for a surveillance system and conclude by discussing some current challenges and future directions for effective surveillance of extremism.