“…Face forgery detection has gained momentum recently in the biometric community owing to its vast application, especially in commercial face recognition systems [27,19,53,37,18,10]. Photo-realistic forged images tamper with the functionality and integrity of security checkpoints, where, ideally, there must be a zero-tolerance policy to false acceptance [24,51,12,40]. Introduced in [12], facial morph images, as one of the categories of the forged face images, can bypass established automated face recognition systems, as well as border control officers, where both struggle to distinguish a bona fide image from a morphed one [25] due to delicacy in synthesizing morphed sam-Figure 1: Group Lasso regularization, as a representation learning, leads to selecting the most discriminative subbands for detecting a morphed image.…”