|
eTrue Inc. combines face and fingerprint recognition to verify identity. The company's TrueFace software circumscribes a face to find relationships between features such as the chin, temples, and eyebrows. |
|
TrueFace software identifies a person's face from a database of stored images. |
|
The EntrePad fingerprint sensor from Authentec Inc. looks below the surface layer of skin to capture a fingerprint's unique ridge and valley patterns. At the center of the chip is a rectangular sensor matrix. The light-colored area surrounding the sensor matrix is a drive ring. When a finger is placed on the sensor matrix, the drive ring couples a small signal onto the finger's subdermal layer, creating a digital image that reflects the fingerprint's underlying surface. |
|
Iris-recognition technology from Iridian Technologies can be used to control access to computer rooms, bank vaults, and labs. To be identified, a person simply looks into a camera mounted near the entrance. |
Biometric systems identify or authenticate a person's identity, an important distinction because different technologies usually work better for one or the other. Identification asks the question, "Who are you?" The system selects a person's individual features and searches a database of stored images or templates to see if a match can be found. Authentication asks if a person is who he or she claims to be. This method compares a person's characteristics against an enrolled image stored on a database and provides a yes or no answer.
The technology sounds better suited for spy movies, but it's proving its value in the real world. MIT's Technology Review recently named it one of the top 10 technologies that will change the world. Biometric systems can prevent unauthorized access or use of cellular phones, PCs, computer networks, and ATMs. The technology is also cropping up in law-enforcement applications such as prison security or criminal identification in public places.
"We like to think we're science fact, not science fiction," says Amy Whilldin, manager of marketing communications at Iridian Technologies, Moorestown, N.J., developers of an iris-recognition technology. "Biometrics isn't the next new thing, it's the next now thing."
Face to face
One of the fastestgrowing areas of biometrics is facial recognition, which uses facial features to distinguish among different people.
With facial recognition, a video capture of a person's face is translated into numerical data and then compared against an enrolled-user database. The technology can be used to control access to buildings and entrances, as well as access to sensitive information on computer networks.
It's the only biometric suitable for long-range surveillance, so it's particularly useful for identifying criminals in retail stores, stadiums, casinos, and other public places. The technology gained attention, and stirred up a civil-rights controversy, when the Tampa Police Dept. used it at the last Super Bowl. Video cameras captured images of attendees' faces as they entered the stadium. The images were then compared to a database of criminal suspects and known terrorists.
One major player in face recognition is eTrue Inc., Southboro, Mass. Its TrueFace software can verify that a person matches his or her stored image or identify whether a person matches anybody in an existing database.
Face images can come from surveillance cameras or transmitted digital images. Captured images may include one or more faces and a background, so TrueFace must first isolate a single face. The software does this by circumscribing below the chin, on either side of the temples, and above the eyebrows. It clips out the face and applies neural-network technology, a proprietary method of eTrue that looks for relationships between facial features. The relationships are encoded and stored in a template code.
"Neural net is able to match two patterns together and come up with a measure of similarity between them. The thing about neural net that's different than most forms of measurement is that it doesn't use rules, it uses patterns. Rules are very brittle in the sense that a rule can say a chair has four legs and a back. But what if the chair had five legs? Under that rule, it wouldn't be a chair. The level of mismatch between two patterns can be significant and can still be called the same object. There's more flexibility in matching patterns than having rules," says Michael Kuperstein, founder, chairman, and CTO of eTrue. This flexibility lets TrueFace accommodate different head orientations, lighting conditions, makeup, hairstyles, new facial hair, and different facial expressions.
To verify an identity, a person uses a PIN, card, or fingerprint to retrieve the preexisting referenceface template representing his or her true identity from a database. TrueFace matches the known reference template against the liveface template and produces a match score. When a match score exceeds a preset threshold, the two faces are deemed to come from the same person. Verification takes about a second.
For identification of an unknown person against a database of people, TrueFace compares the live-face template against all reference templates in a database and sorts all the match scores from the best score down. If the top score is above a preset matching threshold, it's the best candidate to identify the unknown person. TrueFace can compare 500 faces/sec on a Pentium 200-MHz PC.
NASA is putting eTrue's face recognition technology to the test. Some NASA employees use eTrue's biometric log-on service to access data on network servers from their homes. Fingerprint and face recognition authenticate employee identities, and the biometric templates link to the individuals via their e-mail addresses.
Fingerprints leave a mark
The unique line patterns on a person's fingertip have long been used for identification. Fingerprint-imaging systems have found a home in law enforcement, as well as welfare programs where prints identify participants who try to apply for benefits under false identities. As military, government, and financial institutions get in on the game, fingerprint imaging is also being used in security and accesscontrol applications.
One company on the playing field is Authentec Inc., Melbourne, Fla., which offers FingerLoc and EntrePad solid-state fingerprintsensing integrated circuits. The chips use TruePrint technology, which looks beneath the surface layer of the skin — to what is called the live skin layer — to capture a fingerprint's unique ridge and valley patterns and produce a digital image. This means skin surface conditions such as calluses, dryness, moisture, or the effects of aging don't limit the chips' ability to capture an image. Ink, paint, and glue have little or no effect either.
At the center of the chips is a rectangular sensor matrix comprised of numerous individual elements. Surrounding the sensor matrix is a drive ring, which is excited by onchip direct digital synthesis components that generate a sinusoidal signal. The elements of the sensor matrix receive this signal when a finger is placed on the sensor. The drive-ring signal couples onto the finger's subdermal layer to create a digital pattern that reflects the fingerprint's underlying surface.
The eyes have it
Another unique feature for identification is the human iris. With more than 400 identifiable features in the iris, the probability of two being exactly alike is estimated at 1 in 1072. In addition, research has shown that features of the iris remain stable over a lifetime, beginning around age one.
Users first enroll, which involves capturing a picture of the iris using standard black-and-white video imaging. The features of the iris are measured and encoded into an IrisCode record, which is used as a reference template for future comparisons. The IrisCode consists of 512 bytes; half describes the iris features and half contains data that controls the comparison process.
In use, a person stands 8 to 40 in. away from the camera. The captured image is processed into an IrisCode. The resulting IrisCode record is compared to each record enrolled in the database for recognition, a process that takes less than 2 sec.
IRT works as long as the camera can see the iris — even if the person is wearing glasses or colored contacts. This proves especially advantageous in situations where protective eyewear must be worn, such as clean rooms. Another pro to IRT is privacy. Unlike retinal scanning, iris recognition doesn't give away any personal information such as whether the person takes certain prescription or illicit drugs, has hypertension, or is pregnant.
"From the privacy standpoint, IRT actually enhances personal privacy because there are no PIN numbers. There's nothing to lose. There's nothing to forget. With IRT, you need to know nothing more than your iris pattern," says Whilldin.
Iridian's technology can be used to restrict access to physical locations such as bank vaults and labs, as well as in information security and electronic commerce, where it can control access to workstations, networks, and databases. Consumers may soon see the benefits of IRT. EyeTicket Corp., based in McLean, Va., is using Iridian's technology to speed airline travel. The EyeTicket system identifies air travelers only by their irises, so they can check themselves in and board the airplane without using a credit card or other ID. After a onetime enrollment, passengers can be electronically processed for checkin, baggage check, and boarding simply by looking into an ordinary video camera for about 1 sec.
Hitting the right key
Another biometric on the market today identifies a person by the way he or she types at a keyboard. Neither enrollment nor verification interrupt regular workflow. In fact, a person won't realize the biometric is in place unless told. While it's a fairly new biometric, the idea has been around since the invention of the telegraph, when people could figure out who was transmitting Morse-code messages by the specific way the dots and dashes sounded.
Net Nanny Software International Inc., Bellevue, Wash., offers BioPassword software, which recognizes people by how they type their user name and password. To enroll, a person types in his or her user name and password up to 15 times to create a reference template that is stored on the server.
BioPassword adds an additional layer of security to the ones already in place within NT/2000. It adds typing rhythm to the password system and prevents passwords from being misused. Passwords are useless if they're typed by anyone other than the proper person. Network administrators can lock out a user account after three or five log-on attempts, whatever number they designate. An Event Log provides the network administrator with information about who tried to log on, who was successful, and who was rejected.
When someone attempts to log on, BioPassword checks the NT/2000 user account databases to verify that the account information is valid, sees if there is a biometric template on file for that account, and then decides if the typing rhythm matches the template within acceptable perimeters.
And the list goes on
Other physical and behavioral characteristics can be used to identify and authenticate. Handgeometry identification is based on the fact that a person's hand is uniquely shaped, and the shape remains fairly stable over the course of a lifetime. The biometric involves measuring physical geometric features of the hand, such as length and width of fingers and thumb. Voice verification has been explored and developed over the past 30 years. Two approaches to this biometric include dedicated hardware and software at the access point and dial-up of a PC host using regular phones. Signature verification tries to differentiate between parts of the signature that are consistent and those that change with each signing. Some devices also analyze the static image of a signature. This involves acquiring an image of the signature and saving it for future comparisons.
|
GET A GRIP
Using LiveGrip technology, the ChronoLog measures and analyzes features such as veins, arteries, and fatty tissues. Infrared scanners map the hand's internal surface, and a standard computer processor performs an identity-match analysis against a database of prescanned hand images. Initial registration involves a preauthorized individual entering a personal numeric code. Scanners then take 16 scans of the hand. When the data is entered, the computer learns the person's unique characteristics. The process takes only a few minutes, and the information is then stored in a secure database.
|