Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!


Forgot your password?
Privacy The Military United States Your Rights Online

Homeland Security Department Testing "Pre-Crime" Detector 580

holy_calamity writes "New Scientist reports that the Department of Homeland Security recently tested something called Future Attribute Screening Technologies (FAST) — a battery of sensors that determine whether someone is a security threat from a distance. Sensors look at facial expressions, body heat and can measure pulse and breathing rate from a distance. In trials using 140 volunteers those told to act suspicious were detected with 'about 78% accuracy on mal-intent detection, and 80% on deception,' says a DHS spokesman."
This discussion has been archived. No new comments can be posted.

Homeland Security Department Testing "Pre-Crime" Detector

Comments Filter:
  • by Chris Burke ( 6130 ) on Tuesday September 23, 2008 @12:32PM (#25122475) Homepage

    Just an fyi, the accuracy number doesn't directly tell you the ratio of false negatives. It's a measure not just of how many true positives it gets (that's the sensitivity), but also of true negatives(that's the specificity), in that it should both identify the "suspicious" correctly and correctly identify the non-"suspicious".

    You can't go from the accuracy directly to the specificity and sensitivity, since it's a combination of several measurements. The result, though, will be highly dependent on the prevalence of "suspicious" people in their test, which is the ratio of how often what you're trying to detect actually occurs.

    I'm willing to bet that the prevalence they used in their testing is way, way higher than it would be in real life (like 1/4 to 1/2 of the test subjects were "suspicious", while in real life the odds of a random person in an airport being a terrorist is more like 1/1e6 on a bad day). So this would skew the accuracy measurement towards detecting the suspicious and understate the importance of figuring out correctly that someone is not suspicious. The problem is that when you're dealing with something very rare, even if your specificity is very high, the odds that someone you pull out of line because the machine flagged them is in fact innocent is extremely high (it's going to be over 99% chance unless this machine is -very- specific), and if your test methodology doesn't worry as much about specificity, then it's going to be even worse.

  • Re:sensors... (Score:3, Informative)

    by Aphoxema ( 1088507 ) * on Tuesday September 23, 2008 @12:50PM (#25122815) Homepage Journal

    Hey, you should register, you keep talking like that and you'll have karma out the ass!

  • Re:sensors... (Score:3, Informative)

    by Atrox666 ( 957601 ) on Tuesday September 23, 2008 @12:55PM (#25122917)
    As long as you act exactly like everyone else and never have a troubling thought I don't see what the problem is.

Make it myself? But I'm a physical organic chemist!