Biometric identification: the eyes have it

A retinal scan maps the unique pattern of capillaries on the retina with an error rate of only one in 10 million. This compares with fingerprint identification error rates as high as one in 500.

A retinal scan is performed by casting a beam of low energy infrared light into the eye. Because retinal blood vessels absorb more light than the rest of the eye, the amount of reflection varies as the beam traces a standardised path on the retina.

The network of blood vessels is so complex that even identical twins do not share the same pattern.

Fooling a retinal scanner is difficult — the retina degrades quickly after death so one taken from a cadaver could not be used to fake an identity. And any attempt at surgical alteration would be detected by modern scanning equipment.

But because retinal patterns can be altered by diabetes, glaucoma, cataracts or retinal degenerative disorders they cannot be used as a universal security measure without making allowances for age- or disease-related physiological changes.

It took nearly 50 years for technology to catch up with the theory and commercial retinal scanners were not available until the 1980s. But they were soon superseded by iris scanners as a biometric identification tool. The use of iris patterns was patented in 1987 and scanning technology became available in 1994.

The main advantage of iris scans is that they can be performed from a distance of up to 90cm with an initial scan taking 20 seconds and subsequent identifications only two seconds. In contrast, retinal scanners must be within 1.25cm of the eye, the initial scan taking around 45 seconds and authentication scans 10–15 seconds.

Plus, an individual’s iris pattern remains constant throughout their life, barring trauma, so an initial scan lasts a lifetime.

Iris scanning records the crypts, furrows, ridges, striations, ligaments and collarette in the coloured part of the eye using camera technology. It creates images of 512 bytes, whereas retinal scans are only 35 bytes.

The iris has been described as the ideal part of the human body for biometric identification. It is well protected from damage (unlike fingers), and its unique fine texture is determined randomly during embryonic gestation. IrisCode, the first commercially deployed iris recognition algorithm, developed by John Daugman of the University of Cambridge, has a false match rate better than one in 10^–11.

The main applications of iris recognition technology have been automated border crossings, controlling access to restricted areas, mother-infant pairing in maternity wards, and prisoner booking and release. It is being considered for biometric identity cards in several countries.

Over a million frequent flyers have registered with project IRIS (Iris Recognition Immigration System) for automatic border crossing at 10 UK airport terminals.