Date of Graduation
Statler College of Engineering and Mineral Resources
Lane Department of Computer Science and Electrical Engineering
Face recognition systems operate on the assumption that a person's face serves as the unique link to their identity. In this thesis, we explore the problem of morph attacks, which have become a viable threat to face verification scenarios precisely because of their inherent ability to break this unique link. A morph attack occurs when two people who share similar facial features morph their faces together such that the resulting face image is recognized as either of two contributing individuals. Morphs inherit enough visual features from both individuals that both humans and automatic algorithms confuse them. The contributions of this thesis are two-fold: first, we investigate a morph detection methodology that utilizes wavelet sub-bands to differentiate bona fide and morph images. Second, we investigate the usefulness of morphing identical twins to train a network robustly.
Although not always discernible in the image domain, many morphing algorithms introduce artifacts in the final image that can be leveraged for morph attack detection. Because wavelet decomposition allows us to separately examine low and high frequency data, we can identify and isolate these morphing artifacts in the spatial frequency domain. To this end, a wavelet-based deep learning approach to detect morph imagery is proposed and evaluated. We examine the efficacy of wavelet sub-bands for both single and differential morph attack detection and compare performance to other methods in the literature.
Finally, experiments are done on a large scale morph dataset created using twins. This high quality morph twins dataset is used to train a single morph detector. The details of this detector are explained and the resulting morph detector is submitted to the NIST FRVT test for objective evaluation, where our detector exhibited promising results.
Chaudhary, Baaria, "Single and Differential Morph Attack Detection" (2021). Graduate Theses, Dissertations, and Problem Reports. 10290.