Signature Verification Using Deep Learning and CNN

Signature verification plays a crucial role in authentication and fraud detection across various domains such as banking, legal documentation, and digital security. Traditional methods often struggle with intra-class variability, making deep learning approaches, particularly Convolutional Neural Networks (CNNs), a promising alternative. This study presents a CNN- based signature verification system that effectively distinguishes between genuine and forged signatures. The proposed model extracts spatial features from handwritten signatures using multiple convolutional layers, enabling robust feature learning. A Siamese network architecture is employed to compare signature pairs, utilizing contrastive or triplet loss to enhance verification accuracy. The system is trained on publicly available signature datasets and evaluated using performance metrics such as accuracy, precision, recall, and F1-score. Experimental results demonstrate that the CNN-based approach outperforms traditional feature-based methods, providing improved generalization to unseen signatures. This research highlights the potential of deep learning in enhancing signature verification reliability while reducing manual effort in forensic analysis. Index terms: Signature Verification, Convolutional Neural Networks, Deep Learning, Siamese Network, Authentication.