Scraping global threats in Facebook through movement patterns generated by random walks

Abstract

Predicting the occurrence of emerging global threats through Open Social Networks has become a paramount task for national security agencies. We study a set of four foraging strategies to provide an automated searching across the Facebook social network. Our work shows how movement patterns generated by random walks can be developed and applied as novel choices for facing a complex environment, e.g. the Facebook social graph. We develop four algorithms based on optimal foraging theory for crawling the social network and gathering publicly accessible data. We also use our random walks with the aim of sampling and collecting open information through millions of Facebook messages. Finally, this approach allows us to glean insights into the collective moods of regions or groups abroad.