Core Functionality

Product Details, Limitations and Disclaimers

ADVANCE.AI's 3D Liveness Detection is a market leading product that offers a high level of security and accuracy against spoofing attempts that leverages on advanced 3D mapping technology and robust anti-spoofing algorithm.

Liveness Detection & Anti-Spoofing Algorithms: We have implemented robust liveness detection and anti-spoofing algorithms that have undergone extensive testing.
Usability and Success Rates
- Mobile Device Users: Running on latest version of 3D Liveness Detection. 95-98% of users successfully prove liveness within the first 3 scans and ~99% of users prove liveness within the first 5 scans. Results may vary if used under circumstances where there is an unusually high percentage of attacks or if user journeys inadequately motivates users to complete the journey.
- Web Browser Success Rates: Running on the latest version of 3D Liveness Detection on a compatible web browser and webcam. 95-98% of users successfully prove liveness within the first 3 scans and ~99% of users prove liveness within the first 5 scans. Results may vary if used under circumstances where there is an unusually high percentage of attacks or if user journeys inadequately motivates users to complete the journey.
Liveness FAR (False Accept Rate for Spoofs) is 0% - at which no known single Spoof artifact has bypassed the 3D Liveness Detection model.
Tampering Protection: SDK generated 3D face vectors are encrypted and tamper-proof, ensuring that it can not be altered or manipulated. Encryption has been verified as tamper-proof.
Camera Stream Tampering Protection: SDK has been rigorously tested, and no vulnerability to browser code tampering or use of virtual camera to produce false Liveness Success responses has been found.
Hill Climbing Prevention: 3D Liveness Detection aggregates match levels to prevent attackers from determining the proximity to a successful result, while still maintaining excellent user experience. By providing aggregated results instead of exact scores, flags or any other indicators, attackers will be unable to determine success proximity.
- Note: Guidance is only provided to help improve the overall quality of media captured for liveness detection check.
Replay Attack Detection: New face vectors are validated with an updated database of historically hashed face vectors to deter replay attacks. In the event replay attack is detected, the new face vector will be rejected.
- Hashed Face Vectors -
No Skin Tone Biasness: No observable biasness to skin tone (Light and Dark). Though camera sensors relies on reflected light to function, fair and dark skin tones may impact the amount of light captured by the sensors. However, in larger scale tests, there have been no observable performance differences regardless of skin tones.
Security and Usability: Regular updates are performed to improve both usability and security without compromising on user experience
For the user instruction, you can refer to Workflow Node description and configuration