Software Engineering Institute

In today’s environment, the Department of Defense (DoD) has access to more data than ever. But how can it quickly turn these huge volumes of data from numerous and diverse sources into verified, actionable intelligence useful to warfighters in the field? And how can it ensure the systems responsible for processing this data remain operative under adverse conditions?

The SEI is engaging in two efforts to increase DoD situational awareness. The first is an artificial intelligence (AI) engineering collaboration with the Defense Threat Reduction Agency (DTRA). The team produced a system called Cornerstone.

“Cornerstone uses AI to extract information from public data sources and transfer it to a system called the Biological Materials Information Program,” explained Edwin Morris, a senior member of the SEI technical staff who leads part of the team supporting this effort. “That system contains information about facilities around the world that are using dangerous pathogens.” For example, Cornerstone is capable of providing historical information about pathogens in laboratories and also providing near-real-time surveillance of disease outbreaks at locations around the world.

Under the hood, Cornerstone’s natural-language processing (NLP) models are embedded in a flexible, SEI-developed architecture that makes it easy to change data sources, machine-learning (ML) algorithms, and data consumers. It also uses innovative self-monitoring that checks the ML algorithms and other system components for failures. The result is a situational awareness system that, by utilizing DevSecOps methodologies and containerization technologies, enables automated deployment and redeployment of the system.

The second SEI activity leverages Cornerstone’s self-monitoring capability to make situational awareness systems more resilient to failures. This resilience engineering project is creating a prototype that ingests data about the state of system components, including ML algorithms, infrastructure, and cybersecurity status, and then uses that data along with a system model to reason about the state of the executing system.

In ongoing work, the SEI team will use the system model and state to determine when the system is in danger of failing and automatically direct activities to restore the system, such as retraining ML algorithms or rebuilding and redeploying the system. To demonstrate the resilience capability, the SEI will enhance Cornerstone with additional self-monitoring capabilities, develop a system model that uses the resulting data to determine when intervention is necessary, and automate processing to restore the system to the necessary level of operation. If successful, the approach can be applied to ensure that situational awareness systems continue delivering critical intelligence to warfighters and other fielded DoD personnel, even in adverse conditions.