A Deep Learning-Driven Multimodal Behavioral Biometric Authentication System Utilizing Signature Analysis and Keystroke Dynamics

Advancements in the digital world has also led to enormous threats such as credential theft, brute-force attacks, and spoofing. Authentication plays a significant role to ensure secure access to the digital world. Contemporary authentication systems are extremely vulnerable to security threats. New threats are inevitable in the digital era. Recently, lot of day-today work are using internet where the credibility of the user is of much importance. Therefore, a deep learning based multimodal authentication system has been proposed. The proposed work integrates signature analysis and keystroke dynamics to improve accuracy and robustness. This proposed work has three phases i) Signature verification using VGG16, ii) Keystroke dynamics verification using Bi-LSTM and iii) combining both using soft voting classifier. The proposed work has exhibited enhanced security based on the accuracy, precision and F1-score.