Face Anti-spoofing Technique Using CNN and SVM
Deepak B. Desai, S.N. Kavitha · 2019
Nowadays spoofing of face biometrics systems have become very common. The attack is carried out by printed photos, video, masks etc. In a recent study, well known commercial face recognition systems (Face-lock Pro, Visidon, Veriface, Luxand Blink and fast-Access) were easily targeted and fooled with the spoofed images of the targeted person and it can be obtained easily from social networks. The main task in an image classification is to extract the important features. To overcome the problem of spoofing attacks, many came up with different techniques but all those need an expert to extract the important features from the image or video and classify to that accordingly. Also it is a very time consuming process as they use trial and error methods for extracting important features. The paper emphasizes on how to extract all the important features using a CNN model and to classify whether the image belongs to spoofed (fake) or un-spoofed (real) class using SVM.