Iris recognition

The iris's texture—including crypts, furrows, and collarette—is formed by about the eighth month of gestation and remains stable throughout life, barring trauma or disease. John Daugman's algorithm, developed in 1993, is the foundation of most commercial systems: it uses Gabor filters to demodulate the iris image into a 2048-bit binary code, then compares codes using the Hamming distance. A match is declared if the distance is below a threshold, typically around 0.32. The algorithm's accuracy is enhanced by its ability to handle pupil dilation changes and off-axis gaze, though extreme dilation can cause false non-matches.

Modern iris recognition systems use near-infrared (NIR) illumination (700–900 nm) to reveal iris features even in darkly pigmented eyes, where visible light fails. Capture devices range from fixed cameras requiring the user to stand at a specific distance to mobile and handheld units that acquire images at up to 2 meters. The ISO/IEC 19794-6 standard defines the data interchange format for iris images, ensuring interoperability. For non-cooperative or surveillance scenarios, 'iris-at-a-distance' systems have been developed, but they face challenges from motion blur and specular reflections, often requiring multiple frames and quality assessment algorithms.

Iris recognition is deployed in border control (e.g., the UAE's iris screening at airports), national ID programs (India's Aadhaar, which enrolled over 1.2 billion people), and refugee registration by the UNHCR. It is also used in banking for ATM authentication and in mobile devices for secure unlock. The technology's high accuracy makes it suitable for high-security facilities, but its reliance on user cooperation and the need for specialized hardware limit its ubiquity. In 2020, iris recognition was used in contactless payment trials, and it is being explored for continuous authentication in smart environments.

One lesser-known fact is that iris recognition can work on the iris of a deceased person for a short period, which has implications for forensic identification. Another is that the iris changes with age, but not in texture; rather, the pupil size and the visible area of the iris change, which can affect matching. The first iris recognition patent was filed by Leonard Flom and Aran Safir in 1987, but the concept dates to 1936 when ophthalmologist Frank Burch proposed it. Additionally, the technology has been used to identify individuals with aniridia (absence of iris) is impossible, but those with iris coloboma (a hole) can still be matched if the remaining pattern is sufficient. Finally, some systems use the iris's 'chaotic' nature to generate cryptographic keys, a field known as biometric cryptosystems.

Iris

The colored, circular part of the eye that controls pupil size.

Biometric

A measurable physical or behavioral characteristic used for identification.

Hamming distance

A measure of the difference between two binary strings; used to compare iris codes.

Gabor filter

A linear filter used for texture analysis, applied to iris images to extract features.

Near-infrared (NIR)

Light in the 700–900 nm wavelength range, used to illuminate the iris for imaging.

Aadhaar

India's biometric identity program that includes iris scans.

Aniridia

A rare condition where the iris is partially or completely absent.

Iris coloboma

A congenital defect causing a hole or notch in the iris.