Eight months after makers of biometric face-recognition technology saw their shares skyrocket on panic about airport security, the shares are starting to reflect a realization that the technology may not be the security panacea once hoped for.
The concern is not primarily that facial recognition technology doesn't work. It does, say analysts and government studies, especially in routine tasks such as unlocking a personal computer or opening a garage door.
But in high-volume, high-precision applications such as airport security, some users have started to notice that even a seemingly small number of misfires by the machines can have problematic consequences.
Such worries lopped 6.9% off the shares of
on Tuesday, sending them down 79 cents to $10.70. Rival
shed 4.4%, or 28 cents, to $6.12.
The hope is that face recognition can help stop terror attacks in a variety of settings by spotting known terrorists and other dangerous people the government has identified. The shares trade at a premium based almost entirely on post-Sept. 11 zeitgeist; in the week prior to the terror attacks, Visionics shares cost a little more than $4 each, and Viisage's cost about $2. Since then, Visionics has gone above $16.50, while Viisage has been close to $14.
Visionics' FaceIt technology was tested earlier this year at airports in Dallas; Boston; Fresno, Calif.; and Palm Beach, Fla. Though FaceIt and a competing technology produced by Viisage were able to identify individuals from a preselected group passing through some of the airports at a better than 90% clip, at Palm Beach, Visionics' FaceIt had less than a 50% success rate.
According to a Visionics spokesman, its facial recognition technology can be set at different levels of sensitivity, according to a user's tolerance for false positives. The Palm Beach testing was set lower than at the other airports.
"Palm Beach has virtually a zero false alarm rate," said Meier Kahten, a spokesman for Visionics.
Brian Ruttenbur, an analyst with Morgan Keegan, said the real problem is that Visionics simply sells its technology, it doesn't implement it, so the training to use FaceIt and the parameters of the testing were not uniform at all of the airports.
Kahten, who did not know if the Transportation Security Administration, working with the Federal Aviation Administration, would buy or eventually use FaceIt, noted that one source of the Palm Beach data was a group that makes no bones about its opposition to the technology, the American Civil Liberties Union.
"The ACLU is trying to discredit technology even when it clearly works at an extremely high level of accuracy," Kahten said.
Data released Tuesday by the ACLU noted that in tests where the false positive rate was low -- specifically, 0.4% -- the total still came to two or three people an hour or 20 people at a checkpoint that handled 5,000 passengers in a day.
Visionics claims false positives are not a major problem. "The false alarms (one to two out of every 100 passengers) are easily cleared through visual inspection by a human operator, very similar to the way metal detectors work," the company claims.
The ACLU has publicly expressed concerns that such technology is invasive and less than effective.
Ruttenbur noted that most security stocks were down on Monday. He added that applications for facial recognition technology have uses beyond security.
"They're looking at other applications outside of airports," Ruttenbur said, noting that recent legislation regarding border crossings could create a big market. Facial biometrics could also be used to access anything from frequent-flier travel cards, speed passes for a variety of businesses, computer access, garages and more.
"The surveillance aspect is a much smaller market," Ruttenbur said.