INSIGHT: Interpreting Network Structures to Obtain Intelligence on Groups of Hidden Terrorists

Detecting and disrupting a terrorist network requires the ability to: link disparate (noisy, deceptive) data and recognize patterns in the underlying networks, estimate where further intelligence would be useful, and predict the network's response topossible attacks.We propose to combine state-of-the-art social network analysis (meta-networks) and statistical inferencing (probabilistic relational models) techniques to develop mathematical models for characterizing terrorist organizations, and implement those models ina software tool that can be used to detect terrorist organizations and analyze their structural vulnerabilities. Meta-networks capture the connections between people, skills, events, and locations, and generate more effective representations oforganizations than do traditional person-to-person social networks. PRMs capture the variability and uncertainty in these connections in a compact representation and provide efficient inferencing mechanisms. By combining both, we can represent variedterrorist organization structures and facilitate updates and modifications to those structures, while maintaining the level of detail needed for accurate detection/exploitation.In Phase I we will develop and demonstrate models and software to characterize terrorist networks, and evaluate their sensitivity to varying levels of data quality and completeness. Phase II will extend the models and software, to detect terroristorganizations and perform exploitable structure analysis for disrupting terrorist networks. Benefits of this effort include increased utilization of intelligence data, earlier detection of terrorist organizations, and decision support tools to support efforts to disrupt terrorist networks. Tools from this effort will also support structuralcharacterization and exploitation for other domains, such as criminal investigation, corporate structure analysis, and diplomatic assessment.