July 24, 2016 -
A biometrics researcher specializing in iris “liveness” detection is developing new technologies to ensure that iris scanners won’t be duped by either a high-res photo of the authorized user’s eye or their severed eyeball, according to a report by IEEE Spectrum.
Adam Czajka is also studying post-mortem eyes at both the University of Notre Dame and Warsaw University. Contrary to popular belief, he discovered that irises are perfectly viable for iris identification even several hours after the individual’s death.
Czajka helped run an iris liveness detection competition at the International Conference on Biometrics last month in which various iris identification systems had to distinguish a real iris from a paper printout or a printed contact lens.
“A printout is enough to fool some commercial systems we had in the lab,” Czajka said in an interview.
To accurately identify the printout, common methods look for the reflection from a moist cornea, use thermal imaging to detect the real eye’s warmth, or analyze the sample for microscopic evidence of the dot-matrix patterns generated by printers.
The results of the competition have not yet been published, but Czajka said they showed a significant improvement over the previous competition in 2013 where the algorithm rejected 29% of live samples and incorrectly accepted 6% of printouts.
In his latest work, Czajka designed a more advanced method that could properly distinguish a real live eye from a real dead eye by using pupil dynamics to both identify an individual and detect liveness.
During the study he found that each person’s pupil contracts or dilates in somewhat different ways in response to lighting changes.
When initially enrolling people in an ID system, Czajka’s scanner records their pupils’ individual responses to flashes of light. Later, when the same subjects were being identified, the iris scanner emits a pulse of light to determine if their pupil dynamics match those previously captured in their profiles.
The study successfully achieved perfect performance in under 3 seconds, although Czajka emphasized that the scanner was working in lab conditions and it is unclear whether it would perform in the real world under varying lighting condition.
Czajka said he is also interested in studying other complicating factors, such as whether pupil dynamics change if the individual has consumed alcohol or is under stress.
Next, Czajka expanded his research to the properties of post-mortem eyes, collaborating with a colleague from the University of Warsaw’s medical school, Piotr Maciejewicz, who photographed eyes in the morgue.
Using both infrared and visible light cameras to collect images, they took the first set at 5 hours after death, as well as two additional sessions at about 16 and 27 hours after death. Czajka and PhD student Mateusz Trokielewicz tested the eye photo samples using a few commercial products and one open-source iris matching system.
The dead eyes were easily recognizable in the first session, while some of the photos could still be used for identification as late as 27 hours after the individual’s death.
Though the cornea eventually became cloudy over time as the corneal cells deteriorated, its opaque nature did not affect the infrared imaging’s accuracy.
Czajka is currently conducting research using eyes that are a few days or even weeks after the individual’s death.
“We are sure that irises can be used as biometric identifiers for at least a few days after death,” said Czajka. “That’s why liveness detection is so important.”