Originally created 01/13/02

Face-recognition machines coming to airports



WASHINGTON - Face-recognition technology, now being lined up for use in airports, might be capable of finding Osama bin Laden if he showed up at a U.S. airport to catch a flight.

But critics say there's no high-tech quick fix that can ferret out the ordinary terrorist from the ranks of millions of Americans on the move - and error rates on the machines are so great it's just as likely to be the innocent traveler who trips off the alarm.

Michael Thieme, a senior consultant with the International Biometric Group, a New York firm that analyzes the industry, said face-recognition technology can be effective under the best conditions, where the camera gets a good frontal picture and turns it into a complex mathematical formula that matches that of a picture stored in its data bank.

But the CIA doesn't have good pictures of most wanted al Qaeda terrorists to put in a data bank. Even on some of the most wanted, like the Taliban's one-eyed spiritual leader Mullah Omar, on whom there is a $5 million reward, the U.S. government has just a grainy picture of Omar in the back seat of his limousine taken from a distance.

Even if the government can find a passport picture in the files of a foreign government, computers don't understand how the aging process changes people, and the technology is easily fooled by weight gain and beards, or thrown off by sunglasses and hats, Thieme said.

What is even more difficult in airports, he said, are the different backgrounds and the angles at which cameras take pictures of people walking through. He declined to say what the normal error rates of the technology are.

Thieme said the problems are so great that the only value he sees of installing face-recognition machines at airports is as a deterrent. "They provide an incremental gain in security, but it's very small," he said.

That's not stopping airport authorities across the country.

Massachusetts Gov. Jane Swift said she embraced the technology as part of security improvements she ordered at Boston's Logan Airport to restore public confidence in flying. And the Federal Aviation Administration has launched a three-month test of a face-recognition system installed at Fresno Yosemite International Airport in California, T.F Green Airport in Providence, R.I., and Palm Beach International Airport in Florida are also looking at the systems.

Face recognition has some enthusiastic supporters, such as Lt. Bill Todd of the Tampa Police Department, who oversees an experiment using such technology to cut crime in Tampa's Ybor City entertainment district, and supports its use in counterterrorism at airports.

"It would be an absolutely great use in that role," Todd said.

Detractors such as Richard Smith, a Massachusetts privacy consultant who has analyzed the software used by such systems, disagree. Smith says the government would be wiser to pay more attention to solidifying cockpit doors and putting together in-flight security squads than looking for quick-fix high-tech solutions.

"The problem is that we really don't know who the bad guys are," Smith said. "You can put Osama bin Laden in the data base, but he's not going to hijack an airplane."

Smith noted the recent case of Briton Richard Reid, a known criminal who spent three years in jail, and was stopped and questioned before he got on an American airlines trans-Atlantic flight. It was only then that he allegedly attempted to ignite plastic bomb material in the soles of his shoes. Face-recognition technology wouldn't have detected Reid's intentions in advance, just as airport security failed to discover his plans, Smith said.

Smith said the other difficulty with the technology is that it often mistakenly identifies someone as matching a picture in its data base - known in the business as a "false positive."

"You are going to waste a lot of time talking to people fingered by the system," he said. The machines can be adjusted to lower the number of false positives, but that only results in a system less likely to identify a potential target.

The industry says face-recognition technology works well enough that it's being used in Las Vegas casinos to detect card counters and identify others the casinos want banned from gaming houses. Visionics, a New Jersey firm that makes face-recognition software, says "facial recognition is a valuable new technology that has a role to play in counterterrorism" and cites testimonials from police in the London suburb of Newham, who found a 37 percent decline in crime in the two years they have used the system.

The American Civil Liberties Union says information it has gathered shows less success. Documents released to the ACLU under the Freedom of Information Act show that during the first two months the machines operated in Ybor City, they didn't produce a single arrest. During the first four days they operated, the machines signaled 14 matches with pictures in a database, but all were "false positives" and two of those were of people of the opposite sex from those in the data bank.

Face recognition is "all hype and no action," said Barry Steinhardt, associate director of the ACLU. He said it gives people a false sense of security.

Todd said the Ybor City system is currently being upgraded with 45,000 criminal records and added camera systems to make it work better. "The system is still a work in progress," Todd said, adding the system was designed to be both race- and sex-neutral, and so the two cases of opposite-sex identifications should not be criticized.

He believes it has cut crime.

"I don't look at success equaling arrests," he said.

On the Net:

www.visionics.com

http://computerbytesman.com