Paper Title
Signature Verification Using Convolutional Neural Network

Abstract
A number of existing authentication systems use biometric information, more specifically, signatures, for authentication. However, authentication based on signatures is not infallible, as it is possible to forge signatures with a convincing degree of similarity. There needs to be a reliable method of detecting fake signatures in order to avoid forgery and fraud. In the proposed method, we use a convolutional neural network as a binary classifier. The datasets used were sourced from SigComp 2011 and Kaggle. The network is made to learn and extract features from the signatures that are pre-labelled either as fake or genuine. This network is then tested on a previously unseen set of signatures. This method gave varying results; the accuracy achieved from the method described varied in the range 15-60%. There could be a significant improvement in the results in case of availability of larger datasets for training. Keywords - Signature Verification, Convolutional Neural Networks