A reasoning approach for modelling and predicting terroristic attacks in urban environments

During the last decade the world has witnessed a number of terroristic attacks and security incidents exposing numerous vulnerabilities of the urban environment. Hence the ability to proactively detect and even predict potential threats related to terrorist attacks is crucial for supporting government agencies in order to timely react (pro-act) to potentially alarming terrorist attacks. The work described in this paper is part of an overall larger effort in developing a framework for early identification and prediction of terrorist actions (the PROACTIVE project-http://www.fp7-proactive.eu/). The paper focuses on a near real-time reasoning layer designed around a number of reasoning capabilities for transforming raw and symbolic events into meaningful alerts. The reasoning layer was designed to process information sources at different abstraction levels (e.g. sensor information, police patrol inputs, external semantic crafted data sources) and simulates various expert user roles indicated as crucial in the intelligence analyst work flow (i.e. operational, tactical and strategic user roles). Additionally a special focus was given to support functional requirements of the overall terrorist attack prediction system, as producing near real-time detection of threat events by relying on reliable models regarding terrorist actions and predicting sensitive threat events. Hence the overall designed builds on top of approaches as event driven architecture, complex event processing systems, and machine learning techniques. A prototype implementation of layer is presented in a simulated validation scenario. The prototype allows an expert user to monitor threat probabilities for different physical environments, and influence the sensitivity of these environments in real-time as well as and provide feedback for adapting the machine learning models.