Title: Some Intelligence Analysis Problems and Their Graph Formulations

Speaker: Rich Colbaugh, Sandia National Laboratories

Date/Time: Tuesday, February 12, 2008, 9:00 am - 11:00 am

Location: CSRI Building, Room 90 (Sandia NM)

Brief Abstract: The application of graph analysis methods to Intelligence Analysis (IA) problems is of great current interest, with some advocates claiming these methods have the potential to revolutionize IA. However, less has been said concerning the specific benefits of this approach to IA beyond generic claims of automation and scalability. This talk focuses on three important classes of IA problems – assessment, discovery, and prediction – and demonstrates that such questions can be naturally formulated as graph analysis problems and that this formulation enables the development and implementation of powerful new analytic techniques. Assessment questions of interest include: Are the objectives of facility X what is claimed? What would be useful indicators that these objectives have changed? Who are the collaborators of person Y? What are the plans and intentions of this group? We consider a specific, real world problem of this type – assessing the activities and capabilities of a suspected “dual use” bioweapons facility, and illustrate the efficacy of graph analysis methods for this problem. Our focus in the discovery domain is “unknown unknowns” sorts of questions: Are the Russians doing something “interesting” in an advanced technology area? Is there an A.Q. Khan of bioweapons? What new technical and/or proliferation threats are emerging in the WMD domain? Here again we concentrate on a specific problem – the threat posed by “nontraditional” nuclear weapons ideas and substate actors – in order to make explicit the power of graph analysis for addressing hard IA questions. Finally we consider predictive analysis questions: How is a given threat situation likely to evolve? What would be the associated consequences? What are the likely effects of various intervention options? What additional data collection might be useful for predictive analysis? How can we move from a reactive to a proactive posture with adversary Z? In this case we examine the question of anticipating Islamic radicalization and show that novel insights can be obtained for this challenging problem through graph dynamics-based analysis.

CSRI POC: Rich Colbaugh, (505) 284-4116; Laura McNamara, 844-1366