When Maria Macumi’s husband was brutally killed in Burundi, she escaped. She now has a kind of identity, like all other refugees at the Oruchinga refugee settlement in Uganda, that allows her to obtain the necessary amount of food from the World Food Program to feed herself and her children. Iris recognition technology, pioneered by John Daugman, Professor of Computer Vision and Pattern Recognition at the University of Cambridge, is used in part of the identifying procedure.
“The remarkable thing about an iris design is that it includes a lot of unpredictability, which makes each one incredibly distinct,” Daugman explains.
The concept that the iris may be used as a fingerprint dates back to 1949, when J. H. Doggart, a British ophthalmologist, saw the iris’s immensely rich and intricate pattern.
“I’d define it in terms of entropy,” Daugman adds, referring to a basic notion in information theory that evaluates the degree of unpredictability in a group of patterns. “I recognized that iris patterns had a lot of entropy, which means that each one may be encoded with a globally unique signature.”
An ‘impossible task’
“Despite the old tradition saying that ‘the eye is the doorway to the soul,’ recognizing someone only by gazing at their eyes seems like an unsolvable challenge,” says Daugman.
“At first, computer vision concentrated on predictable things with basic shapes, such as manufactured components,” argues Daugman. “Natural things, on the other hand, aren’t necessarily so, and the area is now much more focused on probabilistic reasoning and learning. The important innovation in my iris recognition algorithms was to see unpredictability as the key to the solution, rather than as noise.”
Based on this concept, Daugman created the IrisCode algorithm suite, which offered an automated and quick technique for identifying a person’s identification.
“When the codes for two distinct iris patterns are compared, the probability that they would agree by chance in more than a third of their bits (binary digits, which are the fundamental units of data in computing) is less than one in a million,” Daugman continues. “It’s like throwing a fair coin 250 times in a row—the odds of receiving less than one-third ‘heads’ are less than one in a million.”
Surprisingly, the IrisCode method can do millions of iris pattern comparisons per second due to the ease of parallel bit comparisons. This was vital for the technology’s commercialisation.
Commercialization and patents
In 1994, Daugman’s patents were awarded, and firms including Panasonic, Oki, and Sensar were given licenses to produce customized iris cameras. These cameras capture pictures in the near-infrared band (700–900nm), which display rich iris texture even in dark brown eyes, which are found in the majority of the world’s population. System integrators such as Sagem and Morpho, data centers such as Google, banks, and designers of airport security gates and systems, such as the UK Home Office project IRIS (Iris Recognition Immigration System) enabling passport-free international border crossing, were among the other licensees.
Daugman used information theory to show that IrisCode was very resistant to erroneous matches, even in datasets that were in the billions. Prior to IrisCode, no such strong biometric technology had been seen, and no such technology has been produced since. Today, Daugman’s methods are used in all iris recognition systems throughout the globe.
“In India, if you don’t have the means of establishing your identity, you don’t exist,” Daugman said. Individuals are unable to obtain State benefits and services without proper paperwork. This is a significant issue in a nation where just 5.15 percent of Indians have a passport and a small percentage of the population has a bank account.
