An Efficient and Robust Iris Spoof Detection Pipeline via Optimized Deep Features
Zeenat Zahra, Arvind Selwal, Deepika Sharma · 2024
This research provides a unique method that integrates an optimized deep feature pipeline to improve the iris spoof detection systems’ robustness and efficacy. Advanced countermeasures are necessary due to the growing vulnerability of iris recognition as a secure biometric authentication mechanism to spoof attacks. Our suggested pipeline optimizes iris image representation for reliable spoof detection by extracting discriminative features using cutting-edge deep learning techniques. The pipeline is divided into several stages, each of which is intended to handle a particular issue related to iris spoof attacks. For training and validation, we use a carefully selected dataset that covers a wide variety of fake materials and ambient circumstances. We provide real-time applicability in multiple security scenarios by optimizing the feature extraction process through the use of optimization algorithms. Furthermore, we have created a cutting-edge anti-spoofing system that is challenging for attackers to outperform. Since we implemented two pre-trained models, VGG-19 and ResNet-50, and then tweaked them to extract the deep level features from an iris picture, the suggested DeFusNet model is effective. The predictions from these improved models are fused at the feature level to produce the results. Finally, using benchmark datasets from Notre Dame 2017, IIITD-WVU, and NDCLD, we assessed the planned DeFusNet. Our test findings show notable gains in efficiency and accuracy over state-of-the-art iris spoof detection techniques. In conclusion, the iris spoof detection pipeline that has been shown provides a strong way to strengthen the security of iris recognition systems. Because of its effectiveness and resilience, it is a strong contender for use in practical applications where trustworthy biometric authentication is essential. Optimized deep feature integration raises the bar for iris spoof detection and supports continuous efforts to maintain the security of biometric security systems.