Iris recognition is widely used in biometric identification due to its high accuracy and reliability. However, it is vulnerable to presentation attacks where adversaries attempt to deceive the system using fake iris artifacts such as printed images or contact lenses. This research focuses on enhancing the security of iris recognition systems by addressing the critical issue of iris presentation attack detection. The proposed methodology employs an 18-layer deep convolutional neural network specifically designed for iris presentation attack detection. The input images are processed to isolate the iris region, normalized to standard image dimensions, enhanced to improve image quality and resized. These preprocessing steps are crucial for deep convolutional neural network to effectively learn and distinguish between genuine and fraudulent irises. The results demonstrated the efficacy of our method, achieving an APCER of 0.5, a BPCER of 0.0015, an HTER of 0.25, and an accuracy of 98.61%. These findings indicate that our deep learning based approach provides a significant improvement in detecting iris presentation attacks, thereby enhancing the security of iris recognition systems.
Farhan et al. (Sun,) studied this question.