By Calvin Biesecker
The Department of Homeland Security (DHS) Science and Technology (S&T) in July 2011 plans to conduct a limited user evaluation of a multi-sensor screening solution that attempts to automatically determine if somebody intends to do something bad by measuring their physiological and behavioral indicators.
The upcoming limited user evaluation will be an important step for the Future Attribute Screening Technology (FAST) program in that it will begin to move the test protocol out of the laboratory environment and into an operational type setting.
“It basically let’s us understand what the system is really capable of in a more operational environment,” Bob Middleton, the program manager for FAST, told sister publication TR2.
Additionally it will be an opportunity to observe how a potential user of the screening technology might use it and obtain their feed back, Middleton said.
FAST is being managed by the Human Factors Division within DHS S&T. The program began about three years ago with the goal of using sensors to screen people for malintent, that is, in real-time to automatically determine if an individual may be planning a bad action, such as a terrorist attack. FAST is strictly a research and development program, at least for the time being.
FAST is meant to be a primary screening tool that would speed persons through a checkpoint while only diverting those alarming for malintent for additional screening.
For the limited user evaluation, the FAST technology suite will include test subjects who believe they are attending a high profile event. Some of the subjects will be trying to smuggle contraband into the event.
Making the test subjects believe they are participating in a real event is crucial to the validity of the program. Middleton believes that the “ecological validity” for the upcoming test is high.
S&T began testing its protocols for FAST two years ago. Originally, the testing involved contact sensors and has migrated to non-contact sensors, including a millimeter wave imager, thermal camera, an eye tracker, and a pad that a person stands on to measure gross body movement. These sensors are collecting a number of attributes such as cardiac, respiration, skin signature, eye movement, gaze, pupil dilation and fidgeting.
The data from the sensors is fused and algorithms are used to determine if malintent is present. As of now, the probability of correct classification is about 80 percent, Middleton said. That’s progress, “but probably not good enough to use in an operational environment,” he said.
A year ago, DHS put the ability to detect indicators of malintent at better than 50-50.
“We are still learning which of these particular indicators is diagnostic of malintent,” Middleton said.
The more testing that takes place then the more data is collected, which leads to improvements in the data fusion algorithms, Middleton said. The research is also leading to improvements in the quality of the sensor signals, he said.
In the past year, the FAST program has conducted several test protocols and demonstrations and is currently in the midst of another that is expected to conclude early in November. The testing is taking place at Draper Labs in Cambridge, Mass. Draper is responsible for the fusion algorithms.
In the current test, and another slated to occur in the first quarter of 2011, all of which are leading up to the July 2011 limited user evaluation, the protocols are becoming more complex in terms of what the subjects are conditioned to and also what is being required of the sensors, Middleton said. The wintertime test will more closely approximate the July test, he said.
Next spring the program will do a “tune-up” to make sure the sensor suite is operating as it should, he added.
The sensor suite is housed in a mobile trailer that individuals basically enter from one end and exit the other.