Bank Locker Security System using Machine Learning with Face and Liveliness Detection

Bank Locker Security System using Machine Learning with Face and Liveliness Detection

Prof. Sunil M. Kale, Anuja Nair, Manasi Pagar, Kiran Pagar, Esha Kamble

ABSTRACT

One of the main problems that financial systems are currently facing is ensuring the security of transactions. Banks from all over the world spend a significant amount of money using biometric authentication of consumers because it is convenient and widely used. Particularly in offline settings where digital selfies and ID document facial photos are matched. In reality, nowadays, more extensive programs like automatic immigration control also use selfie-ID comparisons. Limiting the discrepancies between comparative facial photos given their various origins is the procedure's greatest challenge. We suggest a unique architecture based on deep features derived by two well-referenced convolutional neural networks for the cross-domain matching problem (CNN). The results from the data collection, known as Face Bank, show that the proposed face-to-face comparison problem is strong and should be included into actual banking security systems with more than 93% accuracy.

Keywords: Face Bank, Convolutional Neural Networks (CNN), automatic immigration control, digital selfies, face to face comparison problem.