Biometric recognition and authentication systems

How it works

Iris recognition systems take images of the eye using infrared light. Iris melanin is transparent under infrared illumination, this allows iris detail to be revealed regardless of eye colour. 

Recognition algorithms locate the boundaries of the iris and then process the iris portion of the image to provide a distinct and concise representation of an individual’s iris pattern. This representation offers a very high level of discrimination between individuals within a population. 

Attacks

Presentation attack requires capturing an image of an individual’s iris with sufficient detail to allow the creation of an appropriate artefact. Capture of a suitable image is easiest with an iris recognition camera and subject cooperation - though attacks have been shown using a general purpose camera. 

Early iris recognition systems were susceptible to simple presentation attacks using photographs of eyes, but features in some modern iris systems allowing liveness detection now make this much more challenging. 

Implementation notes

Iris recognition is considered to have a very low false match rate and is often selected to be used where population sizes are large. 

Finding and segmenting the iris from larger images of the eye is a challenging task, which can lead to false non-match errors if done unsuccessfully. Patterned contact lenses also lead to false non-matches, limiting the use of iris recognition on populations using these cosmetic devices.

There are two use cases, close-up or iris at a distance. For close-up recognition, hardware to illuminate the eye region and capture the iris image is relatively easy to implement, while iris at a distance is more difficult. Imaging distances of 1-5m can be implemented using commercially available equipment.

Protections

A presentation attack may involve a simple photograph or video sequence of the target's face or eye region, or the use of an artefact, such as a patterned contact lens.

For each of these, liveness detection is required to identify a real person is present. Liveness detection methods may involve observing the location of reflections as illumination angles are changed, or movements of the pupil are measured.

In addition, the fact that an iris system will need to locate the face region and subsequently the eye region adds a certain amount of difficulty to an attacker trying to present an artificial image.

Summary

Iris recognition is a relatively secure biometric method of recognising people. This is in part due to the high level of discrimination across a population that can be achieved. Iris recognition can be fairly robust against presentation attacks for systems with an adequate liveness check, due to the difficulty in obtaining and presenting the necessary high quality images.