Face and Liveliness Detection based Bank Locker Security System

Ensuring transaction security is one of the biggest challenges for banking systems today. The use of biometrics by users has attracted huge sums of money from banks around the world due to its convenience and acceptability. Especially in offline environments where facial images on ID documents are matched to digital selfies. In fact, comparing selfies and ID cards is also used in some more widespread programs these days such as in Automated Immigration Control. A major difficulty with such methods is narrowing down the differences between comparative facial images from different origins. We propose a new architecture for cross-domain matching problems based on deep features extracted from two well-referenced convolutional neural networks (CNNs). Results obtained from the collected data, called Face Bank, demonstrated the power of the proposed face-to-face comparison problem and its incorporation into real bank security systems, with an accuracy of over 93%.