Face Morphing Detection Using Fourier Spectrum of Sensor Pattern Noise

Abstract

Morphing attack is becoming a serious challenge for the existing face recognition systems. Aiming at face morphing detection, a novel method is proposed by using Fourier spectrum of sensor pattern noise (FS-SPN). The sensor pattern noise of the facial image is first extracted based on guided image estimation, and the facial quantification statistics, which characterize the specific frequency difference in FS-SPN between the real face image and the morphed image, are obtained. With a linear support vector machine, morphed face image can be detected. Experimental results and analysis show that it outperforms the existing methods in detection accuracy for both complete morphing and splicing morphing.