Publication Type:Journal Article
Source:Journal of Defence & Security Technologies, Volume 1, Issue 1 (2017)
Keywords:concept detection, Dark Web, domain -specific, focused crawling, homemade explosives
This work presents a novel application for discovering Web resources containing recipes for manufacturing Home Made Explosives (HMEs), analyzing multimedia content for determining its relevance to the HME domain and detecting the HME-related objects it contains. The discovery of HME Web resources both on the Surface and the Dark Web is based on a hybrid infrastructure that combines two different approaches: (i) a Web crawler focused on the HME domain; (ii) the submission of HME domain-specific queries to general-purpose search engines. Both approaches are accompanied by a post-processing classification for reducing the potential noise by re-ranking the discovery results. The multimedia analysis detects HME-related semantic concepts in visual content and determines whether it is relevant to the HME domain. The proposed tools are developed in a user-driven manner based on the needs of law enforcement and security agency personnel, as well as HME domain experts. Overall, the ultimate goal of this application is to provide law enforcement agencies with additional operational means in their fight to keep the citizen safe.