Signatures of Interacting Groups via Network Attributes Learning (SIGNAL)

The analysis of interactions within social media has received significant attention in recent years with respect to cybercrime prevention, online marketing, counter-espionage, political opinion trending, and intelligence analysis. However, while detailed study of these interactions might lead to powerful insights, the sheer quantity of data generated via social media makes manual analysis infeasible. Current automated methods for profiling actors in on-line environments rely too heavily on the behaviors of those actors alone. Given the function of social networks to foster communities of practice around all types of activitiesincluding anti-social activitiesthe behaviors of groups and dynamics of those behaviors should be leveraged to increase the accuracy of identifying hostile actors. Aptima proposes to develop an automated tool for detecting Signatures of Interacting Groups via Network Attributes Learning (SIGNAL). Our solution combines strong theoretical foundation in social group and role theories with statistical network inference algorithms. When fully developed, SIGNAL will provide intelligence analysts with a powerful analysis tool that (1) contains a theory-grounded library of online behavior patterns; (2) performs learning of group behavior patterns from data; (3) executes efficient queries over large social media datasets to find hidden groups; and (4) provides easy-to-use interactive network inference visualizations.