A Framework for Offline Signature Forgery Detection using FSICNN with Preprocessing Techniques

This research introduces an innovative approach to detect forgery signatures, leveraging the capabilities of Convolutional Neural Networks (CNN). In response to the increasing prevalence of signature forgery in various sectors, the study addresses the critical need for an efficient and reliable detection system. For this purpose, in this paper, a Fake Signature Convolutional Neural Network (FSICNN) was proposed. Initially, in the proposed system, input images undergo preprocessing, those preprocessed images enhance their quality, ensuring that the subsequent analysis is conducted on clear and refined data. Then in the preprocessed image for the detection of the forgery signature, the proposed FSICNN and VGG16 are applied. In addition, the results of both methods are analyzed using a confusion matrix measures such as accuracy, recall, precision, and F1-score. The study results indicate that the proposed FSICNN has outperformed when compared to VGG16 in the accuracy metrics. This study contributes to the field of digital security by providing a feasible solution to a common and challenging problem, employing accessible and less complex methodologies suitable for various practical applications.