FA-Net: A Deep Face Anti-Spoofing Framework using Optical Maps
Aaradhya Wadhwa, Abhishek Dinkarnath Garg, Chhavi Dhiman · 2022 2nd International Conference on Intelligent Technologies (CONIT) · 2022
In recent times, with growing technology, the authentication systems all around the world are being shifted to Face recognition, touch IDs, voice commands or other forms of biometrics because these features are unique to every person and are also much easy and convenient to use compared to the traditional password or PIN systems. But in spite of these features being unique, people have found different ways to commit identity fraud thefts such as Replay attacks, photo or video impersonations, fake 3D modeling of the face, voice modulations, etc. So it is of utmost importance that the security systems responsible for tackling these attacks should be cutting edge and as accurate as possible. In this paper, a deep face anti-spoofing architecture, FA-Net is developed. It detects different attacks such print attack or replay/video attack. The performance analysis of the proposed work is evaluated on two benchmarks: CASIA FASD dataset and Replay Attack dataset.