Biometrics: Profiling the Future (Part One)
Geoffrey M. Gluckman
May 31, 2007
A curious incident occurred involving a thin, middle-aged man of medium height wearing a tweed cap atop gray hair. He had taken a Declaration form, but didn’t fill it out. Instead, approaching the Custom’s counter, he beckoned to the AIC, and said, “Would you, dear?”
To my surprise she began to run through the questions. When the monetary declaration came up, the man stated he had on his person $40,000 in cash.
To my untrained ears this was certainly a red flag. After all, hadn’t our instructors at the training center drilled into our heads to follow through on irregularities, and most of all listen to our gut?
With the continuing digitization of government-issued documents, the world of profiling is about to be turned on its head. The science of biometrics--stored personal data such as fingerprints--is creating a revolution. The way people are processed for identification, verification, or given restricted access, in and out of public and private structures, will be forever changed.
As of 2006 the US State Department has issued several hundred employee passports embedded with this type of date storage device. This will be a test of new technology already in use by Australia.
But what about on the national level?
So far, twelve states have adopted implementation of enhanced security measures on state-issued documents, such as Minnesota’s use of a reflective row of loons that seem to float over the driver’s license surface. The new licenses also include an invisible digital watermark detectable by a computer or police scanner, and a ‘status check’ for all immigrant drivers. These are just a few of the features to prevent fraudulent duplication and aid in positive identification. In the next year Virginia will be the first to incorporate a radio frequency identification device (RFID) into state licenses.
The horizon of state-issued identification envisions increases in the multilayered approach of security features, of which biometrics is just a part.
Ideally, there should be 12 to 20 separate security features on a license,” says Reed Stager, Vice President, Public Policy, Digimarc Corporation of Beaverton, Oregon (www.digimarc.com).
Security measures come in both overt and covert forms. Overt are visible to the naked eye, while covert are not. Examples of overt are a bar code, a magnetic strip, a hologram, a sturdy polycarbonate lamination, a ghosted duplicate image, a kinegram, and a reflective image, such as Minnesota’s provided by 3M Company of St. Paul (www.3M.com). Covert features are laser engraving, ultraviolet and infrared images, and digital signatures and watermarks (see photo).
Digital watermarks are covert digital security features that transform multiple elements of driver licenses, such as photo, into machine-readable security tokens,” says Stager.
These marks contain a digital code, which could carry other pertinent security demographics, and even tri-dimensional (3-D) facial biometric information.
Biometrics
The primary biometric factors are fingerprint, retinal (iris), 3-D facial, and hand patterns. Originally developed in the late 1990’s at Bauman University in Moscow, Russia, 3-D facial recognition technology raises the bar in the identification and verification arena. At present the product is only available from A4Vision (www.A4Vision.com) and Geometrix, Inc. (www.geometrixinc.com), both of San Jose, California. The former company appears to be the technological leader.
In brief, based on A4Vision technology, the 3-D process involves a camera the size of a cigarette pack (see photo) that projects an invisible structured light (near infrared) pattern onto the subject’s face, revealing the face's surface geometry, which is captured on streaming video at 30 frames per second. The addition of another axis of measurement--depth--allows 3-D to measure point to point, such as forehead to cheekbone (see diagram). This step takes only three seconds.
Next, that face pattern is plugged into an algorithm to generate a 3-D "mesh" created from measurements smaller than a millimeter. Then, a biometric template -- based on bone structures that don't change over time -- is created from the image and is stored in the database, requiring only 4 kilobytes (see diagram). Thus, the information may be imbedded into a digital watermark or RFID tag.
Identification (one to many) is performed by matching the biometric template against an enrollment database. Verification (one to one) is accomplished through matching the biometric template against a template stored on a smart card.
3-D systems are merely computers that mimic the natural way for humans to recognize one another--the faces,” says Grant Evans, CEO of A4Vision.
Currently, most state-issued licenses have a 2-D facial photograph, which has been easily thwarted despite increased industry efforts at fraud protection. 3-D facial technology proponents claim to solve the shortcomings of the 2-D photo, which relate to a dependence on ideal lighting, camera angle, and amount of facial hair. While 2-D photos are less accurate than fingerprint or iris recognition, it is the current standard the world over and cheaper to implement.
Despite the lack of a 3-D facial photograph database, there is value in identification and verification using this new technology because a 3-D image maps back to stored 2-D images, and improves accuracy.
For example,” says Evans, “if someone suspected of criminal activity presents a 2-D drivers license photo, a 3-D image of the same individual will provide data points about the 2-D image, verifying that the holder is the same person in the 2-D photo. In other words, it makes it harder to fake it.”
Nevertheless, drawbacks to 3-D do exist, such as the absence of completed broad scale testing, especially in areas of wide surveillance. However, in 2003, the system produced a 90 percent success rate at a test run at Logan airport. In addition, as with 2-D, false positive identification is possible, where a face is recognized in the database, but it is not the correct person. Finally, 3-D facial biometrics is not as accurate as fingerprint or iris, but 3-D requires the least amount of cooperation from the person being assessed. It is considerably cheaper than retinal scan systems. Another benefit of these systems over the 2-D is the ability to capture the required facial information while a target is in motion. Perhaps most importantly, facial biometrics can’t be reversed engineered to produce the original face.
Iris recognition may be the trademark of sci-fi thrillers, but these systems are real. In essence, the scanner reads the person’s iris pattern, which is as unique as a fingerprint. This type of system requires more time and becoming comfortable with the technology, a laser passing over one’s eyeball. Moreover, it is not conducive to processing high volumes of people.
Fingerprinting began in the 1880’s with the scientific research of Sir Francis Galton, a British anthropologist. Fingerprints are the most common form of definitive identification and have the widest established databases. Even identical twins do not have the same fingerprints. Again, the process requires full cooperation of the individual, which many object to for reasons of religion and hygiene. In Europe, the use of fingerprints is limited due to the potential for identity theft. In California, this method of identification has been combined with 2-D photos on state-issued identification cards since 1990.
Overall, one type of biometric is not better than another. Instead, it is application dependent, i.e., mass screening versus restricted area access. Most often the combination of several types, such as face with finger, iris with finger, are required to maximize security and reduce fraudulent attempts.
By the way, the man with the forty grand cash was actor Richard Harris.
(In Part Two, I will cover RFID, including sub-dermal devices for humans.)
Resources:
Digimarc Corporation www.digimarc.com
A4Vision www.A4Vision.com
Geometrix, Inc. www.geometrixinc.com
2007 ©Geoffrey M. Gluckman www.geoffreygluckman.com
Print
Email
