AI-Enhanced Multi-Modal Biometric Authentication System for Banking Application

With the continuous evolution of financial services, ensuring robust user authentication has become more crucial than ever to safeguard sensitive data from unauthorized access. This study introduces an innovative platform that integrates sophisticated biometric authentication techniques, combining voice and facial recognition, powered by self-supervised learning. These advanced techniques help mitigate key security risks, like spoofing attacks and environmental inconsistencies, which often weaken traditional biometric systems. In voice recognition, self-supervised learning is applied to identify and prevent voice spoofing attempts, where fraudsters try to impersonate legitimate users. Training on extensive datasets of unlabeled speech enables the model to recognize a diverse range of accents, vocal characteristics, and background noise, enhancing accuracy across various real-world conditions. For facial recognition, the system employs Convolutional Neural Networks (CNNs) alongside self-supervised learning to improve user identification, even when variations in facial expressions, head positions, or lighting conditions are present. The integration of these techniques strengthens the platform’s ability to differentiate between authentic users and fraudulent attempts, thereby ensuring a secure and dependable authentication process. Evaluation results demonstrate a notable improvement in the platform’s overall performance, particularly in recognizing users under challenging conditions. These findings highlight the system’s effectiveness in securing sensitive financial data while offering a reliable and adaptive authentication method.