DEEP TRANSFER LEARNING-BASED FORGERY DETECTION IN HANDWRITTEN SIGNATURES

Abstract

Forgery detection in handwritten signatures is a critical task in document verification, legal systems, and financial institutions. Conventional techniques struggle to handle the diverse patterns and variabilities in signatures. This paper presents a deep transfer learning based approach, utilizing pre-trained convolutional neural networks (CNNs) such as VGG16 and ResNet50, to improve forged signature detection. ResNet50 achieved superior performance with an accuracy of 98.1%, precision of 96.7%, recall of 97.2%, and an F1-score of 96.9%, outperforming VGG16's 97.5% accuracy and 96.0% F1-score. The proposed method effectively balances accuracy and computational efficiency, offering a powerful solution for forgery detection with minimal resource requirements.