Use of Computer Vision to Authenticate Retail Invoices with the Convolution-Neural Networks

Abstract

Digital signatures are a new trend when signing electronic documents in shopping cart platforms. The mundane process involves a login application where a user is authenticated using login credentials and then proceeding to a cart application to produce invoices. In this process, a user is required to authenticate invoices created using a digital signature by using a text input. However, intruders could easily impersonate the user by login to the application and creating a digital signature whereby the authorized user is responsible for invoices created. This impersonation process has caused several breaches in the confidentiality of data. Therefore, this research proposes a system that uses the webcam image of a user in the invoice producing process. The image gathered is validated as a human using a convolutional neural network and then a watermark is created using the system’s date and added to the invoice instead of the current digital signature mechanism. Results demonstrated that the performance of invoice creation was high and less CPU and time was required under high brightness.