Mouse behaviour based multi-factor authentication using neural networks

Abstract

Increased usage of multilevel security is noticed along with a rise in vulnerabilities for password based authentication. This is a level III security protection by authentication, which uses handwritten signature using mouse movement. User mouse behavioral pattern based security is a paradigm for online security layers. It uses digital signature to establish authentication with a secure system. The hypothesis is that even if signature verification and respective issuing authorities exists today, their penetration is not felt among a wider audience. To make it affordable and common man, friendly this paper describes user behavior based security. The recorded behavior is transformed as inputs to construct system understandable features. These recordings are carried out in a training system. During the training phase, every position of mouse movement and its curvature is recorded. This data is processed by a data character classifier and provided to an analyzer such that it computes a feature based on distance between points to build a user profile. A secure messaging token mechanism is integrated for multi factor secure authentication. It uses tokenizer cloud service and its mobile app to receive the token.