The wide-eyed ID flaw

Author: John Monczunski

Unless you work in a high security environment, you may never have had your eye scanned to confirm your identity. But in recent years the eye has been increasingly harnessed as a biometric for identification purposes, and not just in spy movies.

To cite but two examples, a hospital chain in the New York metropolitan area uses it to check patient identity, and at the Schiphol Airport in the Netherlands the scans are used for passport-free immigration. Just as with fingerprints, the retina and iris possess stable features unique to an individual, making them ideal biometric identifiers.

Iris scans are more common and less invasive than retinal scans, which require a person to place their eye extremely close to the camera. In contrast, an image taken from several feet away will often suffice for an iris scan.

The technology works like this: A picture of the iris, the colored portion of the eye that controls the size of pupil dilation, is taken and then compared against a reference image taken earlier.

Pupil dilation, however, may be a source of problems. A study of iris scans recently conducted by research faculty member Karen Hollingsworth ’08M.S.,’10Ph.D. with her doctoral advisors Kevin Bowyer and Patrick Flynn, both ND professors of computer science, found that iris recognition software had trouble recognizing eyes with highly dilated pupils.

The likely reason for the poorer recognition results, the researchers speculate, is that a widely dilated pupil shrinks the iris. With a smaller portion of the iris available for imaging, the recognition software has less data to work with, yielding less reliable results. To circumvent the handicap, the researchers recommend that pupil dilation should be measured and recorded with each iris result, thus offering an indication of the match’s reliability. This recommendation has been adopted by the International Organization for Standardization, which has incorporated pupil radius as a measure of iris image quality into a forthcoming specification for iris image data.

John Monczunski is an associate editor of Notre Dame Magazine.