Publication: Finger vein presentation attack detection using explainable slim network
Loading...
Date
2024-07
Authors
Tan, Ying Jun
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The project titled "Finger vein presentation attack detection using explainable slim network" focuses on developing a compact deep learning-based algorithm to
detect presentation attacks of finger veins. This research is crucial, as biometric identification has become increasingly popular for security purposes due to its
reliability and convenience. However, the vulnerability to presentation attacks poses a serious threat to the integrity of these systems. Existing methods face challenges such as computational complexity, database dependency, and limited generalizability. To address these issues, this project aims to develop a compact deep learning algorithm using explainable residual slim networks through the application of PyTorch. This research contributes to enhancing the security and reliability of biometric systems, ensuring their effectiveness in distinguishing between presentation attacks and bonafide finger veins. The objective of this project is to develop a deep learning-based algorithm to detect finger vein presentation attacks using an explainable residual slim network. The second objective is to minimize the parameters used while maintaining the accuracy of the deep-learning model for detecting finger vein presentation attacks. The third objective is to reduce classification errors with an accuracy of 95%. The ESPRESSNET model was successfully designed with total 17015154 parameters used, it able to obtain the testing accuracy of 97.92% on IDIAP datasets and 97.63% on SCUT-SFVD datasets. As a conclusion, a deep learning-based algorithm to detect finger vein presentation attacks using an explainable residual slim network was successfully developed and it is effective across different datasets.