Using Deep Learning Technology to Prevent Deepfake in eKYC

Deepfake is a rising threat, particularly for its effectiveness in impersonation. By combining and superimposing existing images and videos onto source images or videos using a machine learning technique known as generative adversarial network, it is feasible to create a good quality animated face which can satisfy liveness test checked through facial landmark movement (e.g. eye blink, mouth movement, head yaw/pitch), and impersonate as another facial identity.

This, is evidently a threat in digital ID verification of many eKYC processes that we have witnessed. Furthermore, deepfake can be produced by anyone without knowledge in AI or 3D modelling, simply by using off-the-shelf facial animation tool.


The following video, is an example of face impersonation demo that our Product Department has produced in our eKYC Lab using commercially available software. It is not a sophisticated 2D animation, but effective enough to crack many interactive-mode live face detection SDK that we have tried.

The only way to tackle rising challenges like this, is to use technology to solve problems created from technology - i.e. using deep learning technique to mitigate problems resulted from machine learning advancement.

We have successfully deployed face anti-spoofing API to 2 of our EMAS eKYC Cloud customers. The enhanced OkayFace liveness detection:

1. does not rely on any SDK, thus, effectively reducing the size of mobile app
2. works simply as a JSON API, thus, provides omni-channel support - from mobile app, to mobile web and even desktop PC
3. does not require any facial landmark movement; simply a selfie portrait from front-facing camera will do.

Committing to continuous improvements and safe-guarding KYC compliance of our customers, is the purpose of existence of Innov8tif eKYC Lab.