Haystax Scientists to Present “Model-First” Framework Enhancements at International Workshops

By Haystax, July 5, 2016 | SHARE

Papers describe Bayesian probabilistic solutions to insider threat risk assessment and intelligence analysis argument modeling

McLean, VA – July 5, 2016 – Haystax Technology scientists will present three papers at upcoming international workshops in New York City on advances to the company’s Bayesian modeling framework, Fusion, which underpins Haystax’s “model-first” approach to insider threat detection solutions. Chief scientist and lead author of all three papers, Dr. Robert C. Schrag, will present two of them at the 16th Workshop on Computational Models of Natural Argument (CMNA) on July 9, during the 25th International Joint Conference on Artificial Intelligence. Senior scientist and co-author of the three papers, Dr. Ed Wright, will present another at the 13th Bayesian Modeling Applications Workshop (BMAW) on June 29, during the 2016 Conference on Uncertainty in Artificial Intelligence.

The CMNA papers, Probabilistic Argument Maps for Intelligence Analysis: Completed Capabilities and Probabilistic Argument Maps for Intelligence Analysis: Capabilities Underway, explain how Fusion models naturally generalize argument maps, often used to document arguments in fields ranging from law and public policy to healthcare.

In intelligence analysis applications, reasoning under uncertainty is a critical capability that existing practice usually addresses only informally. Standard maps structure arguments logically but lack built-in probabilistic reasoning. Therefore, Haystax has conducted research and development in Fusion to combine formal logical and probabilistic knowledge representation and reasoning capabilities to support intelligence analysis applications, making analytical products more versatile and powerful.

The BMAW paper, Target Beliefs for SME-oriented, Bayesian Network-based Modeling, describes computational techniques to facilitate reasoning in Fusion models when probabilities for some argument map statements are known from sources, either inside or outside the model, that should be considered authoritative. These known probabilities, or target beliefs, occur in intelligence analysis models as well as in Haystax’s primary insider threat risk assessment model, known as Carbon. Haystax Carbon now has more than two dozen target beliefs and more than 600 person attributes represented as Bayesian network nodes, making target belief processing for Carbon a larger problem than others discussed in preceding technical literature. Haystax’s techniques are also fast enough to make this processing practical in an interactive setting.

The new modeling capabilities described will be available across the full range of Fusion applications available to Haystax customers. The papers’ nine authors include five Haystax staff, two individuals from leading intelligence community providers and two from leading academic institutions.

For more information on Haystax Technology’s insider threat detection solutions, please visit haystax.com/technology/solutions/insider-threat.

About Haystax Technology

Haystax Technology’s mission is to give decision-makers the advanced analytical and risk-management tools they need to prevent, protect against, and respond to a wide array of threats to their most critical systems, facilities, and people. Every day, some 50 million people and 100,000 assets are protected by organizations using our security analytics platform, ranging from local and state agencies and major urban areas to large commercial enterprises and top federal government agencies. Haystax Technology is headquartered in McLean, Virginia. For more information about Haystax, visit www.haystax.com/technology or follow us on Twitter @haystaxtech.