Question
Why are we getting a high passive liveness score on images taken as a photo of a photo.
Answer
There are several factors that can affect passive liveness evaluation. To achieve better results, we recommend the following:
- Evaluate passive liveness (PL) on original, uncropped images taken by our autocapture components. This will lead to more accurate detection, as the background of the subject is important in the evaluation process.
- In the onboarding process, it is essential to ensure that the user cannot upload a photo themselves. It must be strictly enforced that users take a selfie at the time of enrolment. Whether on mobile or web, the use of the Innovatrics face auto-capture component is strongly recommended.
- We also recommend using Video Injection Detection to protect against video injection attacks. You can find out more here: https://developers.innovatrics.com/digital-onboarding/docs/functionalities/security/video-injection/.
- Evaluate PL on the server side. The models on the server are more robust, and the ability to correctly predict liveness is much better compared to on-device evaluation.
Additional Notes
Please refer to our documentation for further information about the PL. https://developers.innovatrics.com/digital-onboarding/docs/functionalities/face/passive-liveness-check/