Two-stage morph detection scheme for face and iris biometrics

Morphing attacks presented by image and/or feature information of more than one individual are a major concern of biometric systems. Penetrating the morphed biometric information to biometric recognition systems may cause successful verification of all presented individuals to the morphed image against a registered template. Therefore, designing a robust biometric system to detect this kind of attack is necessary. This paper aims to implement a novel solution for distinguishing the morph and bona fide identities using a hybrid detection system for face-iris biometrics. The anti-morphing pipeline concentrates on two steps of detection using handcrafted and deep-learning techniques. Firstly, the detection is done according to the texture information of images and then the morphing attack detection framework is applied once more for images recognized as bona fide using deep learning technique to boost the reliability of the face-iris biometric systems. The effectiveness of proposed morph detection method is examined on FERET, FRGC, and FRLL morph datasets using three different morphing algorithms for face biometric. In order to testify the robustness of proposed anti-morphing technique, CASIAv4-Interval database is used on one morphing algorithm for iris biometric. Demonstration of results clarifies the robustness of the proposed solution for morph detection.