You’re hunched over the keypad at the ATM, trying to remember your PIN (was it Aunt Bess’ birthdate backward, or your first girlfriend’s phone number?), hoping the big guy behind you can’t see what you punch in. What if, instead, the machine could recognize you all by itself? That’s the idea behind an iris-imaging identification system produced by Moorestown, N.J.-based Sensar now in pilot testing at ATMs and bank tellers.
Iris-based identification, proponents of the technique say, has a number of features that make it highly accurate and broadly applicable. No two irises are identical-even an individual’s left and right irises are different. The tangled network of connective tissue that creates the random patterns of the iris is protected from environmental influences because it lies inside the eye. And computers can adjust for light conditions and glasses or contact lenses. Iris imaging is thus finding its way into an impressive variety of applications-from computer access control to automated fare collection on public transportation to the identification of thoroughbred horses.
Sensar, a three-year-old spinoff of the Sarnoff Corp. in Princeton, N.J., brings iris identification to the ATM by using technology adapted from Sarnoff military imaging systems to zero in on customers’ eyes. “You don’t have to do anything,” says Rob Van Naarden, the company’s vice president of marketing. “We find you in the scene and take over.” In Sensar’s setup, two video cameras first locate the customer’s head, then an eye, and finally the pupil. A third, higher-resolution camera makes an image of the iris. Using software licensed from Marlton, N.J.-based IriScan, the system digitally encodes the image and compares it with an image on file, typically taken when a customer opens an account.
In the first public tests of the system, begun last spring, Sensar and NCR, a Dayton, Ohio-based-based supplier of ATMs, installed iris-imaging units in ATMs at a savings and loan in Swindon, England. They also placed units at three tellers and a customer service desk as an alternative to signature-based identification. In the first seven months, approximately 1,600 volunteers tried iris identification; according to Van Naarden, there were no “false accepts” (nobody was improperly granted access to anybody else’s account), and bank customers overwhelmingly preferred iris imaging to PIN or signature identification.
Van Naarden estimates that incorporating Sensar’s identification system will add $4,000 to $5,000 to the cost of an ATM, which typically runs from $35,000 to $40,000. Sensar is conducting further pilot tests at banks in several countries; the company plans to begin full rollout of the product late this year at yet-to-be-disclosed locations.
Become an MIT Technology Review Insider for in-depth analysis and unparalleled perspective.Subscribe today