DIA preps for next-gen data environment
The Machine-Assisted Analytic Rapid-Repository System will leverage advances in cloud computing with artificial intelligence and machine learning.
The Defense Intelligence Agency plans to start building its next-generation intelligence environment in fiscal 2020.
The Machine-Assisted Analytic Rapid-Repository System, or MARS, aims to transform DIA's analysis database based on 1990s technology into an automated system that runs at "multiple orders-of-magnitude larger scale and transactional throughput" and that can meet 21st century information demands, DIA said in an 2018 industry day announcement.
Currently, DIA relies on physical servers spread over various networks and manual data entry, but the new cloud-based system will automate much of what intelligence analysts do now -- processing data at the network edge -- while inserting them at the most crucial part of the process, MARS functional requirements manager, Tom Dillaplain, said during an AFCEA NOVA event June 21.
DIA is at "the point where there's just simply too much information coming at our analysts," for them to manually assess granular objects, Dillaplain said.
"It would be impossible for us," he said, particularly for supporting cyber operations where granular understanding of relationships among networks and facilities is key. Even with quadruple the number of analysts, the amount of data would still be overwhelming, he said.
"We have access to a lot of information, but we can only display what the analyst puts in the database," he said. Without automation, analysts can capture only a small fraction of data points.
MARS will move from validating an analyst's work to validating an algorithm that can produce intelligence from raw data. As envisioned, the system will also enable the simulation of courses of action, allowing operators to quickly and fully grasp the likely effects of proposed activities or movements, DIA officials said.
Another MARS goal is to help re-engineer the data environment for several areas -- infrastructure, intelligence, mission data, space and cyberspace -- "instead of replacing one database with a more modern version of that database."
To get there, DIA will need to build analytic services to help with automation, enhance interoperability with mission partners and help push data out. Ultimately, he said, the goal is to create a multi-operational environment where analysts can seamlessly monitor objects as they move.
"The key to successful implementation of artificial intelligence is to place the human at the most useful point in the process. And so we're trying to figure out how to do that as well," he said.
MARS will begin putting out acquisition vehicles in fiscal 2020 to solicit proposals to build it incrementally, Dillaplain said.
This article was first posted to FCW, a sibling site of GCN.
NEXT STORY: Utah to roll out connected vehicle data platform