E-Voting with Iris Recognition Using ResNet

Abstract

This project presents an intelligent and secure E-voting system based on iris recognition using deep learning techniques. The system leverages the unique and permanent patterns of the human iris to authenticate voters accurately, eliminating the risks associated with traditional methods such as voter ID cards or OTP verification. The proposed system preprocesses captured iris images through segmentation and normalization to remove noise and standardize the region for consistent analysis. Advanced feature extraction is performed using a Residual Neural Network (ResNet), which enhances recognition accuracy by learning deeper and more distinctive iris features while reducing computational complexity. Extracted features are stored as compact digital templates in the database and matched against new input samples to verify voter identity. The system provides fast, real-time authentication through a simple and user-friendly interface, ensuring reliability and ease of use even in large-scale elections. This approach strengthens the overall security of the voting process by preventing duplication, impersonation, and fraud. Furthermore, the model achieves high recognition accuracy under varying lighting and environmental conditions, demonstrating the potential for real-world implementation in secure electronic voting and other identity verification applications.

Introduction

The document presents a secure AI-based E-Voting System using Iris Recognition to address vulnerabilities in traditional voting, such as identity fraud, duplicate voting, and human errors. The system integrates biometric iris authentication with deep learning (ResNet-18) to ensure accurate voter identification, prevent impersonation, and maintain a tamper-proof digital voting process. Iris recognition is highlighted for its uniqueness, permanence, and resistance to external factors, making it more reliable than fingerprints or facial recognition.

The workflow involves preprocessing iris images (resizing, normalization, augmentation, masking) to ensure consistent, high-quality inputs for the ResNet-18 model, which extracts 512-dimensional feature embeddings. These embeddings are compared via cosine similarity against stored templates in a MySQL database to authenticate voters. Successful authentication allows vote recording while preventing repeated voting. The system is implemented as a Streamlit web application, enabling real-time voter registration, verification, and voting.

Mathematical foundations, including weighted vote aggregation, accuracy evaluation, and probabilistic validation, underpin vote integrity, fraud detection, and reliable tallying. Extensive preprocessing and validation ensure data consistency, correct duplicate detection, and anomaly handling.

Results show the system achieves 99.2% accuracy, outperforming traditional and basic electronic voting systems, demonstrating robustness, scalability, and reliability. Overall, the framework modernizes elections by combining AI, biometric security, and secure database management, enhancing transparency, efficiency, and voter confidence while laying the groundwork for future AI-driven governance systems.

Conclusion

The proposed E-voting framework, in summary, demonstrates a highly reliable and efficient approach for secure and accurate election management. The system effectively ensures vote integrity, accurate tallying, and robust anomaly detection by leveraging weighted aggregation, probabilistic validation, and structured preprocessing. Advanced verification and validation techniques, such as probabilistic checks and automated cross-verification, improve system reliability, reduce errors, and prevent fraudulent or duplicate entries. Experimental results indicate that the system outperforms traditional manual and basic electronic voting methods, achieving overall vote counting accuracy of 99.2% while maintaining fast processing and minimal validation errors. Furthermore, the framework demonstrates excellent generalization across different election datasets, including external or simulated real-world voting scenarios, highlighting its applicability for large-scale elections. This E-voting system provides a promising tool for election authorities, ensuring secure, transparent, and efficient election processes. Its scalability, robustness, and high accuracy make it well-suited for modern democratic applications. Future work could focus on deployment in real-time election monitoring systems, integration with blockchain-based audit trails, multi-layered security enhancements, and support for multi-modal voter verification mechanisms to further strengthen reliability and trust in electoral processes.

References

[1] Putro M. D., Priadana A., et al. (2025). “A High-Accuracy and Faster Face Recognizer Supporting Biometric Continuous Authentication for Smart Factory Workers.” IEEE Transactions on Industrial Informatics, Vol. 21, No. 5. [2] Shouwu He, Xiaoying Li (2024). “EnhanceDeepIris Model for Iris Recognition Applications.” IEEE Access, Vol. 12, DOI: 10.1109/ACCESS.2024.3388169. [3] J. M. Choudhary, R. Gupta, and S. Dey (2020). “Enhancing Human Iris Recognition Performance Using Ensemble of Deep Neural Networks.” Soft Computing, Vol. 24, pp. 11477–11491. [4] Y. Chen, X. Chen, and Z. Liu (2022). “DADCNet: Dual Attention Densely Connected Network for Accurate Iris Segmentation.” International Journal of Intelligent Systems, Vol. 37, pp. 829–858. [5] A. Ullah, K. Shaukat, M. A. Khan (2022). “Hybrid Approach for Iris Recognition Using Feature Extraction with ResNet and Classifier.” International Journal of Reconfigurable Embedded Systems, Vol. 11(1), pp. 59–70. [6] S. V. Sheela and P. A. Vijaya (2019). “Iris Recognition Based on Gabor and Wavelet Features Using Machine Learning.” IEEE Access, Vol. 7, pp. 128343–128356. [7] A. Al-Waisy, R. Qahwaji, and S. Ipson (2021). “A Multimodal Biometric System Based on Iris, Face, and Fingerprint Recognition for Secure E-Voting Applications.” Computers & Security, Elsevier, Vol. 105, pp. 102–118. [8] G. Li, Z. Sun, and T. Tan (2018). “Efficient Iris Recognition through Deep Feature Learning and Matching.” Pattern Recognition Letters, Vol. 82, pp. 43–50. [9] N. Othman, A. Ross (2017). “On Exploring the Feasibility of Iris Recognition in Visible Spectrum Using Smartphone Cameras.” IEEE Transactions on Information Forensics and Security, Vol. 12, No. 9, pp. 2046–2057. [10] P. Grother, M. Ngan, and K. Hanaoka (2021). “IREX 10: Evaluation of Iris Recognition Algorithms for Mobile and Web Applications.” NIST Interagency Report 8367 [11] R. Daugman (2016). “How Iris Recognition Works.” IEEE Transactions on Circuits and Systems for Video Technology, Vol. 14, No. 1, pp. 21–30. [12] S. Yadav and P. Singh (2023). “Deep Residual Learning-Based Iris Recognition for Biometric Security Systems.” International Journal of Advanced Computer Science and Applications, Vol. 14(3), pp. 112–119. [13] L. Zhang, W. Zuo, and D. Zhang (2020). “Learning a Deep Representation for Iris Recognition with Sparse Convolutional Features.” Neurocomputing, Vol. 408, pp. 268–279. [14] M. G. Kumar, A. Thomas, and R. Rajesh (2022). “Secure E-Voting System Using Blockchain and Biometric Authentication.” IEEE Access, Vol. 10, pp. 98643–98654. [15] S. Badrinath, K. Wahi, and P. Reddy (2021). “Real-Time Iris Recognition for High-Security Applications Using Deep CNN Models.” Journal of Intelligent & Fuzzy Systems, IOS Press, Vol. 40(4), pp. 7283–7294.

Copyright

Copyright © 2025 T.B. Karthikeyan, M.G.I. Jithamanyubabu, Dr. S. Subashini. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.