Agent-based Terrorist Attack Prediction (ATAP)

We propose to develop a prototype for Agent-based Terrorist Attack Prediction (ATAP) for Homeland Defense. ATAP will employ mobile agents to retrieve on demand relevant data distributed across various services, multi-national forces, and intelligence agencies, while maintaining the autonomy of individual data sources. ATAP will use ontology-based data modeling for integrating heterogeneous, distributed data sources. ATAP will incorporate Bayesian Belief Network (BN) technology for building models representing interrelationships among various domain concepts such as terrorist activities and observable events, thus allowing models to predict potential attacks based on those observable events. Finally, ATAP will incorporate a simulation environment supporting speculative analysis and exploration of model-generated hypotheses, via realistic simulated terrorist attack events that are based on input such as the attack location, environmental factors, social networks, etc. The Phase I effort will focus primarily on bioterrorism, and ATAP will help predict both bio-attacks and early detection of disease outbreaks due to such attacks. We intend to rapidly develop a prototype based on: 1) our in-house agent-based information retrieval engine ACQUIRE; 2) a domain ontology encoded with Protégé 2000 from Stanford; 3) our in-house BN engine BNET 2000; and 4) Carnegie Mellon's simulation environment BioWar.