Hybrid Deep Learning Approach for Signature Forgery Detection Using CNN Feature Extraction and Machine Learning Classifiers

Iqrar Ali; Nazia

Authors

Iqrar Ali Department of Data Science, NED University of Engineering Science and Technology Karachi, Pakistan
Nazia Department of Data Science, Mehran University of Engineering & Technology, Jamshoro, Pakistan

Keywords:

Signature Verification, CEDAR Dataset, Offline Signature Verification, Model Hybridization, CNN Hybrid Models, Transfer Learning, Deep Learning, Feature Extraction, Biometric Authentication, CNN Architecture

Abstract

Handwritten signature verification remains crucial for authenticating financial transactions and legal documents. While recent studies have explored deep learning approaches, systematic evaluation of different CNN architectures combined with multiple classifiers remains limited. This study examines model hybridization that integrates multiple feature extractors with various classifiers for the detection of signature forgery using the CEDAR benchmark dataset. We tested four modern CNN architectures (VGG16, ResNet50, EfficientNetB0, and MobileNetV2) as frozen feature extractors. Each one was paired with five classification algorithms (XGBoost, SVM, Random Forest, KNN, and Logistic Regression), making twenty different hybrid models. Our experiments demonstrate that certain combinations yield excellent results, with four setups achieving perfect classification: EfficientNetB0 with both SVM and Logistic Regression, and ResNet50 with both SVM and Logistic Regression, each achieving 100% accuracy with no false acceptances or rejections. EfficientNetB0+LogisticRegression stands out as the best hybrid because it needs only 5 million parameters and takes 9.3 milliseconds to decide for each signature. These results demonstrate that strategic model hybridization, leveraging transfer learning from ImageNet-pretrained networks, can achieve exceptional accuracy without task-specific fine-tuning, making the approach suitable for practical deployment in banking and legal authentication systems.

References

K. Stergiou, S. Ougiaroglou, and A. Sidiropoulos, “Signature forgery detection using deep and machine learning,” Intell. Decis. Technol., vol. 19, no. 3, pp. 1429–1444, May 2025, doi: 10.1177/18724981251330068.

Mustafa Servet Kiran, Gokhan Seyfi, “Feature Fusion Using Deep Learning Algorithms in Image Classification for Security Purposes by Random Weight Network,” Appl. Sci., vol. 15, no. 16, p. 9053, 2025, doi: https://doi.org/10.3390/app15169053.

S. N. Srihari, A. Xu, and M. K. Kalera, “Learning strategies and classification methods for off-line signature verification,” Proc. - Int. Work. Front. Handwrit. Recognition, IWFHR, pp. 161–166, 2004, doi: 10.1109/IWFHR.2004.61.

J. D. S. Rajesh Kumar, “Writer-independent off-line signature verification using surroundedness feature,” Pattern Recognit. Lett., 2012, [Online]. Available: https://dl.acm.org/doi/10.1016/j.patrec.2011.10.009

J. Deng, W. Dong, R. Socher, L. J. Li, K. Li, and L. Fei-Fei, “ImageNet: A Large-Scale Hierarchical Image Database,” 2009 IEEE Conf. Comput. Vis. Pattern Recognition, CVPR 2009, pp. 248–255, 2009, doi: 10.1109/CVPR.2009.5206848.

T. Ojala, M. Pietikäinen, and T. Mäenpää, “Multiresolution gray-scale and rotation invariant texture classification with local binary patterns,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 24, no. 7, 2002, doi: 10.1109/TPAMI.2002.1017623.

N. Dalal and B. Triggs, “Histograms of oriented gradients for human detection,” Proc. - 2005 IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognition, CVPR 2005, vol. I, pp. 886–893, 2005, doi: 10.1109/CVPR.2005.177.

D. G. Lowe, “Distinctive image features from scale-invariant keypoints,” Int. J. Comput. Vis., vol. 60, no. 2, pp. 91–110, Nov. 2004, doi: 10.1023/B:VISI.0000029664.99615.94.

S. Pal, S. Chanda, U. Pal, K. Franke, and M. Blumenstein, “Off-line signature verification using G-SURF,” Int. Conf. Intell. Syst. Des. Appl. ISDA, pp. 586–591, 2012, doi: 10.1109/ISDA.2012.6416603.

J. Ruiz-Del-Solar, C. Devia, P. Loncomilla, and F. Concha, “Offline Signature Verification Using Local Interest Points and Descriptors,” Lect. Notes Comput. Sci. (including Subser. Lect. Notes Artif. Intell. Lect. Notes Bioinformatics), vol. 5197 LNCS, pp. 22–29, 2008, doi: 10.1007/978-3-540-85920-8_3.

D. Impedovo and G. Pirlo, “Automatic signature verification: The state of the art,” IEEE Trans. Syst. Man Cybern. Part C Appl. Rev., vol. 38, no. 5, pp. 609–635, 2008, doi: 10.1109/TSMCC.2008.923866.

I. Masi, Y. Wu, T. Hassner, and P. Natarajan, “Deep Face Recognition: A Survey,” Proc. - 31st Conf. Graph. Patterns Images, SIBGRAPI 2018, pp. 471–478, Jul. 2018, doi: 10.1109/SIBGRAPI.2018.00067.

J. P. Florian Schroff, Dmitry Kalenichenko, “FaceNet: A Unified Embedding for Face Recognition and Clustering,” arXiv:1503.03832, 2015, [Online]. Available: https://arxiv.org/abs/1503.03832

M. E. ; M. H E ; Kim, “Lightweight Visual Transformers Outperform Convolutional Neural Networks for Gram-Stained Image Classification: An Empirical Study,” Biomedicines, vol. 11, no. 5, 2023, doi: https://doi.org/10.3390/biomedicines11051333.

F. Boutros, N. Damer, F. Kirchbuchner, and A. Kuijper, “ElasticFace: Elastic Margin Loss for Deep Face Recognition,” IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit. Work., vol. 2022-June, pp. 1577–1586, 2022, doi: 10.1109/CVPRW56347.2022.00164.

L. G. Hafemann, R. Sabourin, and L. S. Oliveira, “Offline handwritten signature verification - Literature review,” Proc. 7th Int. Conf. Image Process. Theory, Tools Appl. IPTA 2017, vol. 2018-January, pp. 1–8, Jul. 2017, doi: 10.1109/IPTA.2017.8310112.

H. K. Pathak, U. Chaurasia, R. R. Upadhyay, K. K. Singh, and A. Kaplan, “Transfer learning based approach for reliable handwritten signature verification with SignaLearn,” Discov. Artif. Intell. 2026, Feb. 2026, doi: 10.1007/s44163-026-00929-6.

S. Dey, A. Dutta, J. I. Toledo, S. K. Ghosh, J. Llados, and U. Pal, “SigNet: Convolutional Siamese Network for Writer Independent Offline Signature Verification,” Sep. 2017, Accessed: Feb. 13, 2026. [Online]. Available: http://arxiv.org/abs/1707.02131

Wanghui Xiao & Hao Wu, “Learning features for offline handwritten signature verification using spatial transformer network,” Sci. Rep., vol. 15, 2025, [Online]. Available: https://www.nature.com/articles/s41598-025-92704-3