military excels at collecting intelligence information.
Data streams in from satellites and airborne sensors, from groundbased signals interceptors, from troops and intelligence agents. What the military is not so good at yet is analyzing it all.
There are so many sensors collecting so much data that “they are overwhelming our ability to process” it, said Tod Hagan, a software scientist who is working to solve the problem.
“The goal is to produce actionable intelligence in a useful timeframe,” Hagan said. But data floods in unfiltered, unprioritized, unstructured and in disparate formats. Intelligence analysts simply cannot keep up. But “exploitation” software can, said Hagan, who is director of intelligence, surveillance and reconnaissance software solutions at Modus Operandi, a
software company that specializes in writing programs to turn the military’s tsunami of data into useful intelligence.
There are a number of daunting problems to solve.
One is the sheer volume of data collected, including intercepted phone calls, video feeds from unmanned aerial vehicles (UAVs), radar signatures, patrol reports, open source data from Web sites, newspapers and TV stations, satellite images and more.
Even a single intelligence source such as the Army’s Guardrail aircraft generates “an enormous amount of emitter geo-location and audio data,” Hagan said. Guardrail intercepts electronic signals to precisely locate enemy targets on a battlefield.
The Army also has a groundbased signals interceptor, the Humvee-mounted Prophet. It monitors the battlefield to provide commanders with a picture of the “electronic emitters” in the battlespace. If Guardrail’s data could be combined with Prophet’s, the result would be a more accurate picture of enemy and friendly locations and greater confidence in intelligence, Hagan said.
But as data from Guardrail, Prophet and other sensors pour in, “it’s challenging for an intelligence analyst to keep up with it,” Hagan said.
In addition, merging data from different collectors is a complicated process. Guardrail and Prophet do not speak the same digital language. Nor do most of the Army’s 11 other battlefield intelligence systems.
Just comparing different sensors’ calculation of enemy locations is complex. Among intelligence systems, “the Defense Department uses over 50 different formats for describing location,” Hagan said.
The solution is to somehow translate the information collected by each of the sensors into a common language that can then be analyzed by computers. Hagan’s solution is software called Wave Exploitation Framework, or Wave EF.
The software starts by performing textual analysis of the data collected by the sensors. Wave EF scans the data and attaches metadata tags to information for the analysts. A metadata tag is a bit of computer code that identifies or describes the item or “entity” being tagged. The tag helps other computer programs find items they are searching for.
Tagged data enables intelligence analysts to automatically retrieve data related to events, people, organizations, times, locations and other elements of interest from an otherwise overwhelming volume of multisource intelligence data, Hagan said. The tags can get quite specific. For an individual, they might include name, ethnicity, country of origin and location spotted.
The tags are written to be compatible with one of several Defense Department ontologies. An ontology is a system for classifying data and the relationships between entities within the data. It serves as a kind of higher-level language that makes it possible for data with metadata tags to be searched, correlated and analyzed even though the information resides in multiple, incompatible databases. Relevant information is extracted from different databases and then fused.
The result is a lot more intelligence on the battlefield, Hagan said. Computers “can discover facts in minutes that may take human analysts years to determine,” he said.
As the process is perfected and fielded, intelligence analysts “will be able to process an order of magnitude more intelligence” than they can today, Hagan said. And “machine agents acting on behalf of analysts” will be substantially cheaper than their human counterparts.
Besides producing better intelligence more quickly, Hagan said, computer-assisted intelligence analysis could lead to predicting an enemy’s next move. “That’s the holy grail,” he said.
The Army began working with Modus Operandi in January to produce such a system.
Using “advanced semantic reasoning,” the system would “search through vast amounts of data to identify critical patterns of behavior,” Hagan said. Once behavior patterns are spotted, intelligence analysts hope to be able to predict behavior on the battlefield, the company said.
In January, the Navy launched a project to use Wave EF aboard submarines to integrate and fuse intelligence gathered by multiple sensors.
And the Air Force’s 45th Space Wing uses Wave EF to keep track of “asset and resource availability” of hundreds of items and services on its rocket launching pads.