Detecting anomalies in open source information diffusion

François Nel (1, 2) , Antoine Carré (3) , Philippe Capet (2) , Thomas Delavallade (2)

1 MALIRE - Machine Learning and Information Retrieval
2 Thales Communications [Colombes]
3 X - École polytechnique

The adoption of the Internet as a massive information diffusion medium has considerably modified information dynamics. An increasing amount of available information from uncontrolled, various and unreferenced sources makes this new mediatic environment suitable for various information diffusion phenomena like amplification phenomena that may affect significantly political, strategical or economical matters. Strategical or economical intelligence analysts using open sources have to adapt their methods to face significant and changing information flows. In this context, they need automatic tools to select interesting phenomena among large quantities of data. This paper focuses on substantial variations in open source information diffusion that we call anomalies. Firstly, we introduce a model of a website network considering relevant parameters taking part in the dynamics of information flows over the Web. Then we propose a methodology based on this network to detect anomalies.

Domaines: Recherche d'information [cs.IR]

François Nel, Antoine Carré, Philippe Capet, Thomas Delavallade. Detecting anomalies in open source information diffusion. IST087 NATO Symposium on Information management and Exploitation, Oct 2009, Stockholm, Sweden. pp.1-13. ⟨hal-00626817⟩