• PC World
• » Security

Your Face Looks Familiar

Increasingly, digital surveillance cameras are being used with face-recognition software that links what a camera sees with a database of pictures and facial measurements, which in turn are linked to criminal records or other information.

When a face appears that matches a suspicious person's to a set degree of sensitivity, an alarm goes off in a control center and a human operator looks at other factors--height, sex, hair color, and so on--to see if they match, according to Joseph Atick, president and chief executive officer of Indentix, in Minnetonka, Minnesota, a maker of face-recognition systems.

Set to high sensitivity, a system can identify 90 out of 100 people sought, with 2 percent to 3 percent false positives, Atick said. A lower setting cuts the number of false positives, but also the detection rate. Meanwhile, the systems are getting better: They can now identify 40 characteristics of a face in real time, up from 20 a few years ago, he said.

The technology poses little danger to most people who walk through a public place, Atick said. False positives can be cleared up easily by a human operator monitoring the video stream or visiting the site in person, he said.

Public Safety

Face recognition has been deployed with surveillance cameras in public areas of several cities in the U.S. and the U.K., as well as in casinos, where files are sometimes kept on suspected cheaters, according to Gartner's Hunter. It's also beginning to be used at checkpoints, such as for airport security.

Atick and some other experts say face recognition can be used only to detect certain people and not to identify everyone. That may be all it can ever be used for, because lighting conditions change and pedestrians don't always face the camera.

"Over the next ten years or so, you're not going to be able to build a system that would be able to identify every person who walks by a camera in a natural outdoor environment," said Larry Davis, a professor of computer science at the University of Maryland in College Park.

At the Office

Researchers at the University of California at San Diego are developing "intelligent rooms" where hidden cameras and microphones are linked to software for analyzing someone's face, voice, and walk. The system is intended to compare the combination of those characteristics against a database of personal characteristics to identify people, said Mohan Trivedi, professor of electrical and computer engineering. It could even identify a person's mood from facial expressions.

Trivedi sees the technology as making it easier to hold a videoconference. Cameras could focus in on the person talking at any given time, and the session could be recorded and later searched by subject, speaker, and other factors. His group is now experimenting with a new, smaller generation of gear. The team has outfitted a laboratory with 50 cameras built into the walls and furniture.

"We would like to make all the sensors invisible and absolutely unobtrusive," he said.

Watching Over You

A critical hurdle for such systems is the capability to analyze images and spoken conversations, Hunter said. Winston Smith, Orwell's protagonist in 1984, never knew whether government agents were watching him through the telescreen. He thought there weren't enough of them to watch all the time.

But this kind of data analysis could let software, not humans, filter the incoming data. That could mean a lot more monitoring, according to Gartner's Hunter. By 2010, large-scale analysis of images and spoken words will be possible, but probably only in specialized domains with their own key words, such as health care or finance, Hunter predicted.

Trivedi's team has seen progress in this area.

"We are further along than what I used to think," he said. In UCSD's intelligent room, a computer now can identify two people shaking hands in real time.

Tracking Device

The growing power of microprocessors and software also is making it easier for others to know where you are. For that matter, the lowly cell phone has been a fairly effective tracking device for years.

"It's basically a vast surveillance network that half of us are now tied into, that follows us around," said Richard M. Smith, an independent security consultant in Cambridge, Massachusetts.

The network needs to keep track of where the handset is so it can be handed off to the next cell. That means it's possible to pinpoint a person's location to within less than a mile, according to Jim Southworth, chief technical officer of East By North, a broadband consulting company in Reston, Virginia. Government agencies can tap into that information in real time if necessary, or check carrier records, he said.

Location detection takes a big step forward with Global Positioning System, in which devices on the ground or in the air determine their own locations using signals received from a network of satellites. This system can pinpoint the location of any GPS device to within feet, Southworth said. The devices are quickly growing smaller and are being pitched as add-ons for mobile phones and cars.

• See more like this:
• digital surveillance,
• surveillance,
• privacy,
• security,
• camera,
• wireless,
• microphone,
• biometrics,
• facial,
• recognition,
• Carnivore,
• Echelon,
• keystroke,
• capture,
• Trojan horse,
• virus,
• McAfee,
• Norton