On-device Face Authentication System for ATMs and Privacy Preservation
Abhijeet Boragule, Kin‐Choong Yow, Moongu Jeon
Abstract
Security authentication for financial transactions in the banking sector has emerged with modern technological advancements. Currently, many corporations are using smart cards to protect their information and transactions. Recently, in the case of card theft and password breach, many incidents and frauds have been reported widely. In this paper, we propose a face recognition system associated with smart cards to address the breaches caused by card and password theft. We develop a deep learning-based mini face recognition device to authenticate the actual user without any need for large infrastructure, expensive servers, and cloud servers. To protect the individual's face privacy, we utilize smart card memory to store and retrieve the encoded face information. Specifically, we train both face detection and recognition deep learning models and integrate them into the System on Module(SoM). We demonstrate performance on the collected dataset and show that our system is reliable and can be deployed into centralized or decentralized applications.