Humboldt-Universität zu Berlin - Mathematisch-Naturwissenschaftliche Fakultät - Visual Computing

Reflection Analysis for Face Morphing Attack Detection

A facial morph is a synthetically created image of a face that looks similar to two different individuals and even facial identification system cannot distinguish between this synthetic person and the two individuals. It can be generated by warping two face images such that facial features like eyes, nose and mouth are aligned and finally mixing the aligned images with an additive alpha blending.

We study the effects of face morphing on the physical correctness of the illumination to detect facial morphs. For this purpose, we estimate the direction to the light sources based on specular highlights in the eyes and generate a synthetic map for highlights on the skin. This map is compared with the highlights in the image that is suspected to be a fraud. Morphing faces with different geometries, a bad alignment of the source images or using images with different illuminations can lead to inconsistencies in reflections that indicate the presence of a morphing attack.


Overview of our morphing detection method based on reflection analysis.


In order to synthesize specular highlights, we need the light directions relative to the person in the image we want to analyze and a 3-D model of the face of the person in the image. Whereas the 3-D model can be captured at any time with additional devices, the light directions have to be extracted from the suspicious face image. To do so, we employ the vitreous body of the human eyes, which reflects light like a mirrored sphere and has approximate a radius of 7.8mm. After the estimation of the position of the vitreous body and fitting the 3-D face model to the suspicious image, the calculation of the light directions and synthesis of the highlights is a straightforward process.


Detected (left) and synthesized (center) specular highlights on the tip of the nose in a morphed face image. Right: detected (red) and synthesized (yellow) highlights in one image, overlap in green.



C. Seibold, A. Hilsmann, P. Eisert
Reflection Analysis for Face Morphing Attack Detection, Proc. European Signal Processing Conference (EUSIPCO), Rome, Italy, Sep. 2018