Agent-based Terrorist Attack Prediction (ATAP II)

To support the Homeland Defense Initiative, we have developed a general, component-based architecture for Agent-based Terrorist Attack Prediction (ATAP). We have applied it to develop a prototype for the early detection of bioterrorism. We demonstrated overall feasibility using a high-fidelity simulation (CMU's BioWar facility), successfully detecting simulated Anthrax attack events in an urban setting. ATAP is composed of four components: 1) mobile agents to retrieve data distributed across various services; 2) ontology-based data modeling for integrating heterogeneous, distributed data sources; 3) Bayesian belief network (BN) technology for generating alerts based on models representing spatio-temporal interrelationships among domain concepts; and, 4) scenario-based simulation to validate the model in the absence of real data. For Phase II, we propose full-scope development of ATAP. In the bioterrorism domain, we will enhance ATAP's data retrieval capability to include more data sources, and extend our ontology to include additional bio-agents. We will also widen the application scope to include aviation terrorism. In this respect, we have established a collaborative partnership with the Transport Security Administration (TSA) of the Department of Homeland Security (DHS). ATAP will analyze TSA's incident reports that are routinely collected from airports nation-wide and will identify spatio-temporal patterns that indicate possible threats.