Security evaluation and implementation of facial recognition algorithms

Abstract

The following report discusses various facial recognition spoofing techniques in use today, and compares various algorithms, that have been developed for detecting the same. Continuing towards the previous semester’s goal of developing a robust facial facial recognition system, an anti-spoofing module becomes a necessity as it plays a very important role of differentiating between genuine face and fake faces. Even the most advanced algorithms today, such as the FaceNet are prone to spoofing attacks, like the warped photo attack, cut-photo attack etc. Further, various deep-learning based architectures have been implemented and their performance on the task of detecting spoofed images compared.