Evaluation of Iris Biometric Recognition Performance Using Machine Learning Methods

Abstract

Research Problem/Questions – Biometric recognition systems have gained increasing importance due to the growing demand for secure and reliable identity authentication. Among various biometric traits, the iris provides highly unique and stable patterns that remain unchanged throughout an individual’s lifetime. The main research question of this study is: How can machine learning and deep learning approaches improve the accuracy and reliability of iris-based biometric recognition systems? Short Literature Review – Previous studies have demonstrated that traditional machine learning methods such as Support Vector Machines (SVM) combined with statistical or texture-based feature extraction can achieve competitive accuracy in iris recognition. However, recent advances in deep learning, particularly convolutional neural networks (CNNs) such as VGG16 and DenseNet, have shown superior performance by automatically extracting complex features. While earlier works primarily focused on limited datasets and handcrafted features, current research emphasizes data augmentation and transfer learning techniques to address scalability and robustness challenges. Methodology – In this study, the CASIA-Iris-Thousand dataset, consisting of 20,000 near-infrared iris images from 1,000 individuals, was used. For traditional machine learning, SVM classifiers were trained using feature extraction methods including Histogram of Oriented Gradients (HOG), Local Binary Patterns (LBP), and Principal Component Analysis (PCA) for dimensionality reduction. In parallel, the VGG16 deep learning architecture was applied for classification tasks. Data augmentation techniques, such as Gaussian blur and Gaussian noise, were employed to increase sample diversity and improve generalization. Performance evaluation was conducted using accuracy, F1-score, False Acceptance Rate (FAR), and False Rejection Rate (FRR). Results and Conclusions – The results show that SVM with HOG+PCA features achieved 96.5% accuracy with a 3.5% FRR, while the VGG16 model, combined with Gaussian blur augmentation, reached 99% accuracy and 1% FRR, demonstrating superior performance. The low FAR values indicate high security, while the low FRR values highlight user-friendliness. Compared with previous works, the proposed augmentation strategies significantly improved recognition performance, particularly for larger class sizes. The findings confirm that deep learning models, when combined with data augmentation, outperform traditional machine learning approaches in iris biometric recognition. This study highlights the potential of deep learning for real-world biometric security systems and suggests that future work should focus on optimizing models for deployment in resource-constrained environments.

Keywords

• IRIS recognition
• Machine Learning
• Support Vector Machine
• VGG16
• Biometric authentication

Iris Biyometrik Tanıma Performansının Makine Öğrenmesi Yöntemleriyle Değerlendirilmesi

Abstract

Biyometrik tanıma sistemleri, güvenlik ve kişiye özel kimlik doğrulama gereksinimlerinin artmasıyla birlikte günümüzde yaygın şekilde kullanılmaktadır. Bu çalışmada, iris görüntülerine dayalı biyometrik tanıma sistemlerinin performansı, farklı makine öğrenmesi ve derin öğrenme yöntemleri kullanılarak kapsamlı bir şekilde analiz edilmiştir. Araştırmanın ilk aşamasında, açık kaynaklı bir iris veri seti üzerinde kişi tanıma algoritmalarına uygun olacak şekilde çeşitli ön işleme ve veri artırma teknikleri uygulanmıştır. Bu adımlar, sistemin tanıma başarımını optimize etmeyi ve sınıflandırma algoritmalarının genel doğruluğunu artırmayı hedeflemiştir. Ardından, geleneksel makine öğrenmesi yaklaşımlarından biri olan destek vektör makineleri (Support Vector Machine - SVM) ile derin öğrenme tabanlı VGG16 sinir ağı modeli kullanılarak sınıflandırma işlemleri gerçekleştirilmiştir. Modellerin performansı doğruluk (accuracy) ve sahte red oranı (False Rejection Rate - FRR) metrikleri üzerinden değerlendirilmiştir. VGG16 modeli gauss filtre kullanılarak artırılmış veri seti ile %99 doğruluk ve %1 FRR değeri ile en iyi performansı gösterirken, SVM ile aynı veri seti ve HOG+PCA öznitelik çıkarımı kullanılarak %96.5 doğruluk ve %3.5 FRR oranlarına ulaşmıştır. Çalışma kapsamında geliştirilen model dosyaları ve kodlar, açık erişimli olarak paylaşılmaktadır.

Keywords

• IRIS Tanıma
• Makina Öğrenmesi
• Destek Vektör Makinaları
• VGG16
• Biyometrik Tanıma

References

1. [1] Vivek kumar; K Nageshwara Rao. “Analysis of Biometric Authentication Techniques: A Review".” Volume. 9 Issue.3, March – 2024 International Journal of Innovative Science and Research Technology (IJISRT), www.ijisrt.com. ISSN - 2456-2165, PP :1736-1746, doi:10.38124/ijisrt/IJISRT24MAR886
2. [2] R. A. Hamaamin, “Biometric Systems: A Comprehensive Review,” BASRA JOURNAL OF SCIENCE, vol. 42, no. 1, Jun. 2024, doi: 10.29072/basjs.20240110.
3. [3] C. Raghavendra, A. Kumaravel and S. Sivasubramanian, "Iris technology: A review on iris based biometric systems for unique human identification," 2017 International Conference on Algorithms, Methodology, Models and Applications in Emerging Technologies (ICAMMAET), Chennai, India, 2017, pp. 1-6, doi: 10.1109/ICAMMAET.2017.8186679.
4. [4] Suwardi, Y.N., Emami, M.S. (2024). "An Effective Iris Identification Software System for Hospitals’ Emergency Unit". 2024 IEEE 12th Conference on Systems, Process and Control (ICSPC), 53-58.
5. [5] R. Harrabi, "A Combined Support Vector Machine and Statistical Method for Iris Recognition," 2025 4th International Conference on Computing and Information Technology (ICCIT), Tabuk, Saudi Arabia, 2025, pp. 738-745, doi: 10.1109/ICCIT63348.2025.10989426.
6. [6] O. M. Moslhi, “New full iris recognition system and iris segmentation technique using image processing and deep convolutional neural network,” vol. 6, pp. 20–27, 3 2020.
7. [7] S. Minaee and A. Abdolrashidi, “Deepiris: Iris recognition using a deep learning approach,” arXiv preprint arXiv:1907.09380, 2019.
8. [8] K. Simonyan, A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” in 3rd International Conference on Learning Representations, 2015, pp. 1–14.

9. [9] M. Turk and A. Pentland. Eigenfaces for recognition. J Cogn Neurosci, vol. 3, no. 1, pp. 71–86, 1991.
10. [10] T. Ojala, M. Pietikäinen, and D. Harwood (1994), "Performance evaluation of texture measures with classification based on Kullback discrimination of distributions", Proceedings of the 12th IAPR International Conference on Pattern Recognition (ICPR 1994), vol. 1, pp. 582 - 585.
11. [11] N. Dalal and B. Triggs, "Histograms of oriented gradients for human detection," 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’05), San Diego, CA, USA, 2005, pp. 886-893 vol. 1, doi: 10.1109/CVPR.2005.177.
12. [12] Institute of Automation, Chinese Academy of Sciences, “CASIA-Iris-Thousand,” [Online]. Available: http://www.cbsr.ia.ac.cn/english/IrisDatabase.asp. [Accessed: Jun. 8, 2025].
13. [13] Szymkowski, M., Jasinski, P. and Saeed, K. "Iris-based human identity recognition with machine learning methods and discrete fast Fourier transform". Innovations Syst Softw Eng 17, 309–317 (2021). doi:10.1007/s11334-021-00392-9
14. [14] Daugman J (2004) How iris recognition works. IEEE Trans Circuits Syst Video Technol 14(1):21–30.
15. [15] Kuhifayegh, F. and Rajabi, R. (2024). "Iris Recognition via Deep Learning Using Capsule Networks with Enhanced Routing Algorithm". Sustainable Energy and Artificial Intelligence, 1(2), 91-99. doi: 10.61186/seai.2410-1012
16. [16] Adegboye Olujoba James and Awodoye Olufemi Olayanju. “An Iris Recognition System Using Enhanced Convolution Neural Network". Acta Scientific Computer Sciences 4.10 (2022): 24-29.