Authors: Professor S. Rama Rao, Associate Professor M.Radhika, Associate Professor M.Prasanna Kuamr
Abstract: Online signature verification (OSV) systems are vital for enabling secure authentication in digital environments that depend on electronic signature collection and verification. Not only do these systems examine the static structure of a signature, but they also use deep learning and machine learning techniques to assess its distinct dynamic properties, such the pen's pressure, speed, and stroke sequence. These days, researchers look at the distinct static and dynamic aspects of signatures using methods like hybrid models, RNNs, CNNs, and Generative Adversarial Networks (GANs). A major drawback of the aforementioned deep learning approaches for building dependable and responsive real-time systems is the constant training of these models in response to database updates involving new users. Our plan is to create an OSV system that uses a Convolutional Neural Network (CNN) and a ReactJS website so that users may save or upload their signatures to a database. Based on the degree to which the user's uploaded signature matches their original signature, the model uses the spatial information gathered from the signatures to make judgements. In order for the system to handle both genuine and fake signatures, it is regularly trained. This system aims to provide identity verification for critical digital applications and online transactions in a secure, reliable, and resilient manner. To do this, we combine methods for preprocessing signatures, extracting features, and building classification models.
