Measure for measure
Know thy enemy, Chinese general Sun Tzu said in The Art of War. Knowing thy friend is just as important. In fact, one of the most important capabilities in any security system is recognition, thus the appeal of biometrics.
A biometric is a measurable physical characteristic that can be used to identify an individual. The process of biometrics involves automated methods to do exactly that, whether fingertip scanners to match prints or imagers for facial recognition or iris matching. Other examples of biometrics include hand geometry, voice, and signature.
Not surprisingly, photonics has a role to play in this area. Currently, the most common technology for scanning the friction skin on the pads of the fingertips is frustrated-total-internal reflection imaging. The subject places a finger on a transparent glass or polymer platen, and the system produces what is essentially a binary image--dark where the ridges contact the glass and light where the valleys in between do not.
All is not perfect, though. Oils and moisture on the skin, or even different skin textures, can present problems, as can variable environmental conditions. Optical coatings on the platen can improve imaging, but they can retain latent prints and require periodic cleaning, which may in turn compromise the coating.
As a result of these concerns, as well as form factor issues, developers are turning toward solid-state technologies.1 In one solid-state optical design, light guides transfer the image of an LED-illuminated finger to a CMOS image sensor. Non-photonic alternatives include arrays of capacitive sensors that image differently depending on whether they are in contact with a ridge or valley, arrays of piezoelectric pressure sensors, or even MEMS-based pressure or capacitive systems.
The technology is advancing on all fronts. The problem is that these systems need to operate in the real world. Indeed, the single most important challenge faced by biometrics system designers is the human factor. "These are technologies that interface with human beings, and we've ignored the human part," says Jim Wayman of the Office of Graduate Studies and Research at San Jose State University, and former head of the National Biometric Test Center (San Jose, CA).
The U.S. National Institute for Standards and Technology (Gaithersburg, MD) just released a study showing that the quality of prints captured by a fingerprint scanner varies considerably depending on its height from the ground.2 Part of the problem is getting people to use a biometric system properly. This would eventually be solved for, say, workers who had to figure out how to scan their fingers properly to get into work each day. In an airport screening scenario, though, the situation becomes more challenging. "The idea is you want somebody to come up to the device and inherently understand what to do, and we're not quite there yet," says Mary Theofanos, study co-author and project leader for Biometric Human Usability.
Building a fixture that would position a subject's hand properly for the scan might improve print quality but could ultimately backfire because some might be resistant to putting their hand inside an opening. If the platen surface is warm, even, Theofanos suggests, it can raise concerns of germs and contamination.
Iris mapping appears to be more effective, both in terms of mapping and its unlikelihood of returning a false positive. In the UK Passport Services eight-month biometrics enrollment trial of 10,016 subjects, the iris verification portion returned success rates of 91% to 96%, compared to 48% to 69% for facial recognition and 80% to 81% for fingerprint verification.2 Still, some subjects did encounter problems getting a successful scan. A new iris mapping system--Iris on the Move--is designed to eliminate those problems (see sidebar).
The events of September 11 put a focus on security. As a result, biometrics must be the next hot market, right? Sort of. Actually, according to Wayman, biometrics has been the next hot market since the late 1960s. "We've had 40 years of these implementations and they havent caught on yet, he says. This is not an emerging technology. Its a technology that hasn't emerged. "
"[Post 9/11] there was this big response in the biometrics industry," says Maxine Most, principal of Acuity Market Intelligence (Louisville, CO). "People thought, 'Now, this is a way we can actually protect ourselves in the future.' But there's not really a big correlation between biometrics and preventing terrorism." A biometric system could, perhaps, screen out known terrorists, but only if it contained their biometric templates. In reality, biometric systems are far more often used to recognize friendliestravelers enrolled in express screening programs, for example, or employees accessing a secured building.
And even for the government, the cost can be too high. Wayman points to the U.S. Immigration and Naturalization Services Passenger Accelerated Service System (INSPASS). With pilot systems deployed at six airports, processing as many as 146,000 admissions per year in aggregate, the project cost a cool $18 million, a fact not lost on federal auditors. "We found that the facilitation benefits provided by INSPASS in FY1998 were insignificant because INSPASS accounted for less than 1% of the total admissions," says a report from the U.S. Department of Justices Office of the Inspector General.3 The technology worked; the human factors didn't.
On the commercial side, biometrics is still in search of a value proposition. Acme Coffee might install a fingerprint reader so patrons can instantly charge their lattes to their account, but will it see a major upsurge in business that will offset the cost? Unlikely. If the recognition systems can reduce payment or check-cashing fraud, say, then theres a compelling argument for return on investment.
Acuity's Most cites a biometric system in place at an intermodal cargo transfer station (trucks to trains). As part of a larger driver authorization system that logs container number and vehicle license plate, the biometric system authenticates the drivers. "What they've done is increase throughput and decrease theft by controlling who comes in and out of these intermodal facilities," says Most.
This shows the technology is slowly finding niches where it is cost effective, but the killer app has yet to emerge.
If human factors create the problems, then take human participation out of the loop. That's the thinking behind the Iris on the Move project. James Matey and colleagues at Sarnoff Corp. (Princeton, NJ) developed the system, which requires the user only to look straight ahead and walk through a metal-detector-like portalthe system does the rest.
The capture volume for a typical system is approximately 3 cm x 3 cm x 10 cm. In contrast, the Sarnoff systems capture volume is 15 cm x 20 cm x 80 cm. An array of IR LEDs provides off-axis illumination, and a set of 2048 x 2048 electronic imagers captures the scene at 15 frames per second. By using strobed illumination and time-gating image capture, the system eliminates motion blur and much of the ambient light; narrowband filtering screens out the rest.
An algorithm searches the image for a pattern of specularities from the illumination to find the iris. It processes the resultant data to map the iris and check it against a template. It sounds complex, but the system can effectively identify as many as 20 subjects per minute.
K.L.
Information on biometrics-related activities of the U.S. government is available at www.biometrics.gov.
The Biometrics Consortium website, www.biometrics.org, provides a number of helpful links to biometrics activities, resources, and organizations around the world.
