Rumour detection and monitoring in open source intelligence: understanding publishing behaviours as a prerequisite

In the context of information warfare, rumour detection has become a central issue. From classical media-related campaign, to propaganda and indoctrination that lie at the core of terrorism, rumour is a mean widely used and thus a threat that must be identified as soon as possible, and in the best-case scenario, anticipated and curbed. The emergence of a new informational environment due to the adoption of the Internet as a massive information diffusion medium has led to a situation suitable to the creation and propagation of rumours. Indeed, the Web gives everyone not only the possibility to observe information flows but also the opportunity to influence and create them. In order to tackle the issue of rumour detection, one has to understand the mechanisms underlying their propagation. In this perspective, we believe that it is essential to identify and understand the publishing behaviours of the sources. Therefore, we focus in this paper on the identification of groups of sources with similar publishing characteristics. We propose to tackle this problematic by using clustering methods on data extracted from Web sources. The resulting clusters obtained from the clustering are then interpreted as groups of Websites behaving similarly and used to characterize publishing behaviours.