Databases the size of entire nations could be searched in parallel to make a confident identification decision

John Daugman University of Cambridge

Researchers at the University of Cambridge, UK, compared over 2,000 iris images - a total of over 2.3 million possible pairs - and found that the chances of one person's iris being mistaken for another person's were extremely low.

Iris patterns are good candidates for biometric systems because they are more distinctive than fingerprints. And, just as with fingerprints, even identical twins have different irises.

The researchers used small video cameras to take pictures of their volunteers' eyes and then calculated a 2048-bit digital signature or IrisCode for each subject.

No false positives

They found that the chances of two people having iris signatures that are even two-thirds identical are one in 10 million.

There is a popular belief that the iris systematically reflects one's health or personality... but such claims have been discredited as medical fraud

John Daugman and Cathryn Downing

Iris recognition systems have to allow for some variation in iris signature because eyes appear different under different conditions.

But the study shows that if a computer finds that only three quarters of a person's iris signature matches the signature it has on record, the chances of this being a false identification are still only one in a thousand, million, million.

The University of Cambridge's John Daugman says that the system which he helped develop produced no false positives at all when it was tested on the eyeballs of thousands of volunteers in the US, UK and Japan over a three-year period.

Tests of the system by the British government's National Physical Laboratory also found no false positives in more than 2.75 million comparisons, he says.

Changes over time

He and his colleague, Cathryn Downing, are also confident that their system will cope with any variations over time.

"There is a popular belief that the iris systematically reflects one's health or personality, and even that its detailed features reveal the state of individual organs (iridology), but such claims have been discredited as medical fraud," they write.

When the scientists looked at the differences between signatures calculated for the same iris in the same person but at different times, they found differences averaging around 11%, well below the level that would cause confusion.

The Cambridge system has already been tested in the High Street and was well received by customers of the Nationwide Building Society, but the then cost of the cameras used prevented more wide-scale use.

For those who fear gruesome tricks by would be thieves, Dr Daugman has some comfort: "Soon after death, the pupil dilates considerably, and the cornea becomes cloudy.

"There are several tests for proving that an iris is living tissue and not a photograph or printed contact lens," he says.

"Some have to do with the way living tissue looks in infrared light; others have to do with motion of the pupil (always oscillating slightly, even under steady light levels), or eyelid movement and blinks."

The work is reported in the journal Proceedings of the Royal Society: Biological Sciences Series B.