Evaluating the Integration of Morph Attack Detection in Automated Face Recognition Systems

Due to the possibility of automatically verifying an individual’s identity by comparing his/her face with that present in a personal identification document, systems providing identification must be equipped with digital manipulation detectors. Morphed facial images can be considered a threat among other manipulations because they are visually indistinguishable from authentic facial photos. They can have characteristics of many possible subjects due to the nature of the attack. Thus, morphing attack detection methods (MADs) must be integrated into automated face recognition. Following the recent advances in MADs, we investigate their effectiveness by proposing an integrated system simulator of real application contexts, moving from known to never-seen-before attacks.

1. Introduction

The phenomenon of face spoofing has become increasingly important in biometrics and cybersecurity over the years. Advances in technology, especially in image and video manipulation, have seen the birth of phenomena such as deepfakes, face synthesis, and morphing techniques. The latter consists of gradually transforming a face image into another (Figure 1) and can be used for malicious purposes, for example, to deceive a face recognition system (FRS). In fact, it is possible to obtain false faces containing the characteristics of multiple real faces. The result of this operation can be maliciously exploited to share an identity document [8], as the face resulting from a well-made morphing process can be associated with all the contributing identities by a human operator and an automatic FRS [23]. The problem is even more evident if we consider that such a document can also be used by a terrorist to evade border control.