ATM identity validation with face recognition
A cross-platform prototype exploring face recognition as a second identity check at ATMs.
Recognition rate above 90% in the project's tests
The problem
Card-and-PIN authentication can benefit from an additional, non-contact identity signal when security risk is high.
My contribution
I designed, implemented, tested and documented the prototype as my master's project.
Approach
Haar Cascade face detection combined with Local Binary Pattern recognition, persistent training data, motion detection and a desktop application workflow.
What changed
The project report states a recognition rate above 90% under its testing strategy, and recommends future work on speed and recognition performance.
Evidence
- Biometric systems need attention to testing conditions, failure handling, privacy and layered authentication, not accuracy alone
Tools and platforms