Generative Adversarial Network (GAN): Discriminator Model for Online Signature Verification

Abstract

The rapid development of modern secure authentication system plays significant roles over traditional mechanism like enable password, PIN, code and tokens. Conventional mechanism has possibility of loss, theft or misuse. Today, modern biometric, behavioural and physiological systems are alternative for robust authentication. In addition, online signature authentication has become popular due to time based features like velocity, pressure and stroke sequence which is difficult to forge. Existing approaches to online verification encompass statistical models, template matching, and machine learning techniques, including Hidden Markov Models (HMMs), Support Vector Machines (SVMs), and neural networks. Recent advancements in deep learning have further enhanced accuracy and robustness in signature-based authentication.

In this paper, the application of Generative Adversarial Networks (GANs) for online signature verification using the SVC 2004 dataset is explored. GANs is consists of two neural networks architecture which are generator and discriminator that able to generate synthetic signature and its verifications.The discriminator of GANs is reliable component to distinguish real and fake signatures. Through the combining of GAN architecture concepts that offers reliable scalable and secure authentication.