This repository provides an open-source silent face anti-spoofing solution that detects whether a face is genuine or fake (e.g., printed photos, screen displays, silicone masks, 3D prints). It includes model training architecture, data preprocessing methods, and test scripts, along with a pruned MobileFaceNet model optimized from 0.224G to 0.081G FLOPs with minimal accuracy loss. The system offers both a classification main branch and a Fourier spectrum auxiliary supervision branch, with an Android APK and SDK for deployment.