Face Verification for Onboarding and Transaction in Banking

This project, in collaboration with CIMB Bank Vietnam, focuses on developing a face verification system for use in onboarding and transactions. As a digital bank, CIMB aims to improve security and customer experience by implementing AI technologies. The goal is to create a solution that meets the State Bank of Vietnam's requirements, featuring face verification and liveness detection to identify spoof images and enhance security.

The project is designed with a modular architecture, using pretrained models and GPU acceleration to ensure quick processing times. Key components include APIs for liveness detection, image extraction, and verification, all secured with API keys. The system can handle hundreds of transactions per second, maintaining a balance between false acceptance and rejection rates.

The methodology involved extensive research and testing of face detection and comparison techniques, leading to the selection of facial embedding methods for their effectiveness. The project also addressed challenges related to model disparities and data preprocessing, ultimately achieving a scalable and secure deployment using Docker and Kubernetes. This initiative not only meets current banking needs but also establishes a foundation for future improvements in AI-driven security solutions