Presentation Attack Detection in Facial Authentication using Small Training Dataset Obtained by Multiple Devices
Anna Yu. Denisova, Victor Fedoseev · 2021 International Conference on Information Technology and Nanotechnology (ITNT) · 2021
The paper proposes a way to construct a feature set in the problem of countering presentation (or spoofing) attacks on facial biometrics authentication systems. In such a type of attack, an intruder disguises as an authorized user using his photo or mask. To detect such attacks, we use data from special sensors in addition to RGB camera, including thermal, depth, infrared, or other data. Our method is fast to train, does not depend on the consistency of the input data, and can be efficient in the conditions of a lack of training data. The conducted experiments have shown its performance over some algorithms based on the classical learning approach on the WMCA dataset.