Dermalog dominates in iris biometrics liveness detection competition

Dermalog has won a convincing victory in the LivDet-Iris 2025 competition, with the highest accuracy for detection iris biometric spoofs in all four evaluations.

The competition was held as part of the IEEE International Joint Conference on Biometrics (IJCB), held last week in Osaka, Japan.

Aside from Dermalog, there were half a dozen other entries, from universities in Germany, China, the U.S., India and Norway. The competition was organized by experts from the University of Notre Dame, University of North Carolina and Clarkson University in the U.S., and the National Research Institute (NASK) of Poland.

The competition proceeded in two parts; the first made up of three tasks, each dedicated to a different kind of spoof attack scenario, and the second consisting of systematic attempts to spoof the system by laboratory staff.

One of the three challenges during part one was an evaluation of iris liveness in an operational payments scenario by PayEye, based on its sequestered dataset made up of the most common physical attacks observed or anticipated. Those attacks include paper printouts, e-book reader presentations and doll eyes.

The second task in part one was to identify biometric samples morphed with deep-learning models.

The third evaluated the robustness of iris PAD software to textured contact lens patterns created with modern manufacturing methods. “As modern high-resolution printing, multi-layered texturing, and enhanced pigmentation techniques make textured lenses increasingly difficult to distinguish from natural irises, PAD methods trained on older datasets may struggle to maintain accuracy,” the organizers explain. Dermalog’s iris PAD was found to be 99.99 percent accurate in catching fraud attempts using advanced contact lenses.

The runner-up in task one is a team from Michigan State University, while a team from the Indian Institute of Technology Mandi was runner-up in both task two and task three.

The system-level evaluation consisted of testing with real people and presentation attacks. Dermalog was the only participant in part two.

In the pre-print paper describing the results, the organizers observe that the state of the art in iris PAD is sufficient to detect advanced contact lenses, and Dermalog’s model “demonstrated strong performance on both bona fide and presentation attack samples.”  The real-world payment conditions of task one and the biometric morphing attack in task two are both still challenging to the industry.

The low participation from iris biometrics providers raises the prospect that the modality may not have sufficient commercial interest to quickly overcome those challenges.

Article Topics

biometric liveness detection  |  biometrics  |  DERMALOG  |  IEEE  |  IJCB  |  iris biometrics  |  LivDet  |  presentation attack detection  |  spoof detection