Thanks to leapfrogging advancement in deep learning, the facial recognition API is quickly becoming mainstream. The vast availability of facial recognition API over cloud-based subscription, is making it almost utility-like now.
However, not all facial recognition APIs are born equal. Some may not be commercially available for on-premise deployment option when required, while some remain less commercially tested for varying skin tones, and most importantly – many commercially available options are not supported with highly secure liveness detection component.
OkayFace is applied to following use cases
Real-time Facial Sighting
Substitution to Password
OkayFace is powered by facial verification algorithm that satisfies 0.000001 false match rate. This is further strengthen with liveness detection component that supports financial-grade anti-spoofing techniques.
Liveness detection (also known as anti-spoofing) is a process that should take place before facial recognition, in use cases that requires stringent verification of a person’s facial identity. It makes sure that the person (user/customer) you are interacting with, remotely over an app, is indeed a live person and not a photo or recorded screenplay. It prevents spoofing attacks.
The liveness detection technique offered is:
- Passive liveness detection: Also known as non-interactive mode, or silent mode, this deep-learning-based technology only requires a single-frame portrait photo from live camera to determine if a face is live or otherwise.
- Prevent spoofing attack from printed photo, portrait displayed from screen, deepfake, animated talking head, and printed face mask with edges cut-out
- Supports multi-channel UI, without dependency on native mobile app technology
- Proven implementation that involved multi-ethnic population in Southeast Asia
- Support and technical expertise in Southeast Asian timezone
- Readily integrated with ID capture technology to support facial ID verification with various identity documents, such as passport, national identity card and driver’s license.
- Presentation attack detection methods that adhere to ISO 30107-3 guidelines.