Evaluation Techniques to Detect Face Morphing Vulnerabilities for Differential Images

Real-time facial recognition has been effectively implemented in a variety of secure applications, including border control. The biometric community has expressed a significant deal of interest in the facial recognition systems’ (FRSs) susceptibility to many types of attacks, including face morphing attacks and direct and indirect attacks. A morphing attack aims to disable a Facial Recognition System (FRS) at an Automated Border Control (ABC) gate by displaying an e-passport or electronic machine-readable travel document (eMRTD) that is generated using a morphed facial image. Single Image-Based MAD (S-MAD) and Differential Image-Based MAD ($D-M A D)$ are two different categories of face morphing detection techniques. The main focus of this most recent research is D-MAD. Obtaining the suspected morph images from the passport and comparing it with the trusted image or live caught face image from the ABC gates makes the D-MAD approach a good fit for border crossing scenarios. A comparison of several methods, techniques, and databases is discussed in this work.