State of the Art Single Image Morphing Attack Detection Based on Image Forensics

A face morphing attack is an identity sharing scheme with which usually two persons, criminal and accomplice, can use the same identity document. A morphed face image can be created with computer software by blending the face images of usually two persons. Under certain circumstances the morphed image resembles both contributors and can easily fool officers and automated face recognition systems. A lot of research effort has been put into automated morphing attack detection (MAD), which can be categorized as single image MAD (SMAD) and differential MAD (DMAD). Here, we create and evaluate an automated SMAD algorithm which is based on an image forensics method, namely JPEG compression error. Thus, the method can also be used to visualize morphing related traces, something very important in a forensic context. By extracting simple statistical image features from questioned images and several compressed versions, and training a support vector machine, we achieved top performance at the corresponding benchmark of the Bologna online evaluation platform with an equal error rate of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\sim 5 \%$</tex>.