Home / Current Issue / Paper 1723629
Development of an Improved Fingerprint Biometric Model for Automated Teller Machine Security System
Subject area: Science,Engineering and Technology · Area of research: Cybersecurity, Machine Learning
Abstract
Fingerprint authentication provides a biometric approach for improving user verification in ATM systems. However, fingerprint images may contain variations and alterations that can affect reliable classification. This study developed a Hybrid Autoencoder–CNN–ANN model for fingerprint-based ATM authentication using the Sokoto Coventry Fingerprint Dataset (SOCOFing). The model used an autoencoder for feature representation, convolutional neural network (CNN) layers for spatial feature extraction, and artificial neural network (ANN) layers for classification. Fingerprint images were converted to grayscale, resized to 128 × 128 pixels, and normalized before training. The autoencoder was first trained to learn compact fingerprint representations, after which the learned features were used by the CNN–ANN classification stage. Experimental results showed that the developed model achieved a test accuracy of 89.48%, with 95% recall and 82% F1-score for the real fingerprint class, while the weighted F1-score was 90%. The confusion matrix recorded 3,934 altered fingerprints correctly classified as altered and 1,435 real fingerprints correctly classified as real. However, 563 altered fingerprints were classified as real, indicating that false acceptance remains a limitation. Compared with the CNN-based study of Okorie et al. [1], the developed model achieved higher accuracy, recall, and F1-score, although its precision was lower. The findings demonstrate the feasibility of combining autoencoder, CNN, and ANN techniques for fingerprint classification in a simulated ATM authentication environment.
Keywords
Fingerprint authentication; biometric security; autoencoder; convolutional neural network; artificial neural network; SOCOFing; ATM security; deep learning.
References
[1] M. K. Okorie, N. J. Ngene, U. C. Ugwa, and G. O. Ota, “Fingerprint Recognition Using Convolutional Neural Network (CNN) For Secure Authentication Systems,” European Journal of Theoretical and Applied Sciences, vol. 4, no. 3, 2026. European Journal of Theoretical and Applied Sciences
[2] H. Chiroma, “Deep Learning Algorithms based Fingerprint Authentication: Systematic Literature Review,” Journal of Artificial Intelligence and Systems, 2021. Institute of Electronics and Computer
[3] S. Minaee, E. Azimi, and A. A. Abdolrashidi, “FingerNet: Pushing the Limits of Fingerprint Recognition Using Convolutional Neural Network,” 2019. arXiv
[4] S. Saponara, A. Elhanashi, and Q. Zheng, “Recreating Fingerprint Images by Convolutional Neural Network Autoencoder Architecture,” IEEE Access, 2021. IEEE
[5] A. Hussian, F. Murshed, M. Al-Andoli, and G. Aljafari, “A Hybrid Deep Learning Approach for Secure Biometric Authentication Using Fingerprint Data,” Computers, vol. 14, p. 178, 2025.
[6] H. G. Muhammad and Z. A. Khalaf, “A Survey of Fingerprint Identification System Using Deep Learning,” International Journal of Computing and Digital Systems, 2025. International Journal of Computing and Digital Systems
[7] L. Efrizoni, S. Armoogum, and M. Z. Zakaria, “Deep Learning Innovations in Fingerprint Recognition: A Comparative Study of Model Efficiencies,” International Journal of Advances in Artificial Intelligence and Machine Learning, 2024. International Journal of Advances in Artificial Intelligence and Machine Learning
[8] G. H. Aljafary and A. Hussian, “Deep Convolutional Neural Networks for Fingerprint Classification,” 2025.
[9] M. A. A. Jabbar, A. Radhi, S. A. Z. Mghames, S. Behadili, and M. Alsaedi, “Fingerprint Forgery Detection and Person Identification Based on Deep Learning,” Iraqi Journal of Science, 2025. Iraqi Journal of Science
[10] S. S. Hameed, I. T. Ahmed, and O. M. Al Okashi, “Real and Altered Fingerprint Classification Based on Various Features and Classifiers,” Computers, Materials & Continua, 2022.
[11] M. S. Ali, A. Akram, J. Rashid, M. A. Jaffar, D. Shah, S. Ali, and M. Tahir, “Fake Fingerprint Classification Using Hybrid Features Learning With Gradient Boosting,” Applied Computational Intelligence and Soft Computing, 2025. Wiley
[12] Y. Madhav, S. China Venkateshwarlu, and D. Nagaraju, “Advanced Fingerprint Alteration Detection: A Comparative Analysis of Real and Synthetic Modifications Using InceptionV3 on the SOCOFing Dataset,” 2025.
How to cite this paper
@article{1723629,
author = {Samuel, O. D., Prof. Afolabi, A. O., Prof. Adeosun O. O., Akande, O. V., Adepoju, J. A.},
title = {Development of an Improved Fingerprint Biometric Model for Automated Teller Machine Security System},
journal = {Iconic Research And Engineering Journals},
year = {2026},
volume = {10},
number = {4},
pages = {249-259},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1723629.pdf},
abstract = {Fingerprint authentication provides a biometric approach for improving user verification in ATM systems. However, fingerprint images may contain variations and alterations that can affect reliable classification. This study developed a Hybrid Autoencoder–CNN–ANN model for fingerprint-based ATM authentication using the Sokoto Coventry Fingerprint Dataset (SOCOFing). The model used an autoencoder for feature representation, convolutional neural network (CNN) layers for spatial feature extraction, and artificial neural network (ANN) layers for classification. Fingerprint images were converted to grayscale, resized to 128 × 128 pixels, and normalized before training. The autoencoder was first trained to learn compact fingerprint representations, after which the learned features were used by the CNN–ANN classification stage. Experimental results showed that the developed model achieved a test accuracy of 89.48%, with 95% recall and 82% F1-score for the real fingerprint class, while the weighted F1-score was 90%. The confusion matrix recorded 3,934 altered fingerprints correctly classified as altered and 1,435 real fingerprints correctly classified as real. However, 563 altered fingerprints were classified as real, indicating that false acceptance remains a limitation. Compared with the CNN-based study of Okorie et al. [1], the developed model achieved higher accuracy, recall, and F1-score, although its precision was lower. The findings demonstrate the feasibility of combining autoencoder, CNN, and ANN techniques for fingerprint classification in a simulated ATM authentication environment.},
keywords = {Fingerprint authentication; biometric security; autoencoder; convolutional neural network; artificial neural network; SOCOFing; ATM security; deep learning.},
month = {October},
}