Researchers develop liveness detection for iris biometrics

A researcher from the Warsaw University of Technology has trained a machine-learning algorithm to perform liveness detection for iris scans, MIT Technology Review reports.

The system developed by the research team, led by Mateusz Trokielewicz, can distinguish a living iris from a dead one with 99 percent accuracy, with all errors observed in testing from misclassification of live subjects, but with an important caveat. The changes in iris appearance that allow liveness detection are only complete after the subject has been dead for about 16 hours.

“Samples collected briefly after death (i.e., five hours in our study) can fail to provide post-mortem changes that are pronounced enough to serve as cues for liveness detection,” the researchers say in their paper “Presentation Attack Detection for Cadaver Iris.”

Training was performed with the Warsaw BioBase PostMortem Iris dataset, which consists of 574 images collected from 17 people with near-infrared imaging between 5 hours and 34 days after their deaths. Another dataset of live iris images was created with the same camera to remove variables, and the images were cropped to remove the means of holding the eyes of the deceased open.

The Warsaw University of Technology has been part of the ongoing effort to develop liveness detection for iris authentication, including as a co-presenter of the Liveness Detection-Iris Competition 2017.

Article Topics

biometric liveness detection  |  biometrics  |  biometrics research  |  iris recognition