Explainable ArcFace-Based Deep Metric Learning for Multimodal Biometric Recognition and Authentication Systems
International Journal of Medical Toxicology and Forensic Medicine,
Vol. 16 (2026),
1 Dey 2026
,
Page 1-8
https://doi.org/10.22037/ijmtfm.v16.53370
Abstract
Background: Biometric recognition offers a reliable solution for recognizing individuals based on unique physiological traits such as fingerprints and iris patterns. Fingerprint-based systems are widely employed because of their simplicity and distinct ridge structures, but they are generally affected by partial, worn, and noisy captures. Iris recognition provides higher accuracy and stability because iris patterns remain largely unchanged over an individual’s lifetime. This study presents the Fingerprint-Iris Feature Fusion with Deep Metric Learning (FIF-DML) approach for multimodal biometric recognition and authentication.
Methods: The proposed model learns features using two parallel EfficientNet-B0 networks that independently extract deep representations from fingerprint and iris inputs. These extracted features are then fused at the feature level to form a unified representation. In addition, ArcFace-based metric learning maps the fused features into a compact embedding space, enhancing inter-class separability and intra-class compactness. In the identification stage, a query fingerprint-iris pair is converted into a 128-dimensional embedding and matched with gallery embeddings using cosine similarity to predict the identity. Finally, Grad-CAM visualizes the regions that most influence the model’s final decision.
Results: We conduct extensive experimental analysis of the proposed FIF-DML model on the Multimodal Iris Fingerprint Biometric dataset, achieving 97.33% precision and demonstrating superior performance compared to existing methods.
Conclusion: The results confirm that the proposed FIF-DML model provides a reliable and effective solution for secure biometric authentication.
- Multimodal biometric recognition, EfficientNet-B0, Computer vision, Explainable artificial intelligence
How to Cite
References
[1] Sheela SV, Radhika KR. Deep learning-driven fusion of iris biometrics for optimized multimodal authentication and informative security. Int J Inf Comput Secur. 2026;29(4):397-435. [DOI: 10.1504/IJICS.2026.153343]
[2] Nurdin D, Lawi A, Toaha S. From Biometric Surfaces to Contextual Intelligence: A Comprehensive Survey of Multimodal Authentication and the Trimodal Fusion Paradigm. In: 2026 International Conference on Current Research in Artificial Intelligence and Data Science (ICCRAIDS). Vol 1. IEEE; 2026:1-6. [DOI: 10.1109/ICCRAIDS67816.2026.11519743]
[3] Nurudeen BI, Sani AA, Oriyomi AM, Towobola IO. Facial Recognition System Using A Deep Learning Approach: a Review. J Inst Res Big Data Anal Innov. 2026;2(1):400-14.
[4] Gao C, Yang Z, Jia W, Leng L, Zhang B, Teoh ABJ. Deep Learning in Palmprint Recognition: A Comprehensive Survey. IEEE Trans Syst Man Cybern Syst. 2026. [DOI: 10.32388/YZ5TFY]
[5] Ali K, Ajan A, Imtiaz MH. FPGA-Based Implementations of Biometric Recognition: A Review. Electronics. 2026;15(10):2145. [DOI: 10.3390/electronics15102145]
[6] Kakade S, Raut U. Biometric Authentication Systems: Trends, Challenges, and Future Prospects-A Comprehensive Review. MethodsX. 2026:103908.
[7] Hu B. Tiny transformers: multimodal digital biometric authentication model using tiny enhanced dual transformers with quality-gated fusion network. PeerJ Comput Sci. 2026;12:e3657. [DOI: 10.7717/peerj-cs.3657]
[8] Binu D. Bimodal biometric authentication based on index level fusion for combined iris and fingerprint biometrics. Biomed Signal Process Control. 2026;117:109556. [DOI: 10.1016/j.bspc.2026.109556]
[9] Mohammed A, Salama A, Shebka N, Ismail A. Enhancing network access control using a multi-modal biometric authentication framework. Eng Technol Appl Sci Res. 2025;15(1):20144-50. [DOI: 10.48084/etasr.9554]
[10] Kaur P, Kant C. 2-Phase Multi-trait Biometric Authentication Model Against Spoofing Attack Using Deep Hash Model. SN Comput Sci. 2024;6(1):47. [DOI: 10.1007/s42979-024-03513-w]
[11] Maghrabi LA, Altwijri M, Binyamin SS, Alallah FS, Hamed D, Ragab M. Secure biometric identification using Orca predator’s algorithm with deep learning: Retinal iris image analysis. IEEE Access. 2024;12:18858-67. [DOI: 10.1109/ACCESS.2024.3360871]
[12] Nivedita V, Joseph JA, Varghese DT, Nithyanandh S, Sundaram J, Ali G. Multi-modal biometric authentication integrating gait and Face Recognition for Mobile Security. In: 2025 International Conference on Smart & Sustainable Technology (INCSST). IEEE; 2025:1-6. [DOI: 10.1109/INCSST64791.2025.11210361]
[13] Multimodal iris-fingerprint biometric data [Dataset]. Kaggle. Accessed October 13, 2026. https://www.kaggle.com/datasets/ninadmehendale/multimodal-iris-fingerprint-biometric-data
[14] Iqbal A, Jaffar MA, Jahangir R. Enhancing brain tumor multi-classification using EfficientNet B0-based intelligent diagnosis for Internet of Medical Things (IoMT) applications. Information. 2024;15(8):489. [DOI: 10.3390/info15080489]
[15] Peng N, Chen M, Feng G. Feature extraction and recognition of Chinese mitten crab carapace based on improved MobileNetV2. Appl Sci. 2024;14(12):4982. [DOI: 10.3390/app14124982]
[16] Alduhailan A, Kamarudin NH, Sheik Abdullah SNH, Dau A. Deep learning in biometric authentication: Challenges, recent advancements, and future trends. J Adv Inf Technol. 2025;16:458-77. [DOI: 10.12720/jait.16.4.458-477]
[17] Kaur V, Bhatt DP, Tharewal S, Tiwari PK. Intelligent multimodal 3D biometric recognition using PointNet++ for robust face-ear authentication. Sci Rep. 2025. [DOI: 10.1038/s41598-025-27757-5]
[18] Khatri M, Sharma A. Deep learning approach based on Iris, face, and palmprint fusion for multimodal biometric recognition system. Int J Performability Eng. 2023;19(6):407. [DOI: 10.23940/ijpe.23.06.p6.407416]
- Abstract Viewed: 19 times
- 53370 Downloaded: 3 times