There are two key elements to make the clever stuff work
To be able to run an electronic facial ID check, the client clicks on the ID check task, takes a photo of themselves on their web cam or on their phone through the app.
They follow the instructions, providing a picture of their ID document and perform several moves to ensure it is them.
We then do all the clever and taxing stuff; validating the documents using advanced machine learning and visual recognition algorithms.
The evaluation of a document has two criteria to satisfy:
Is the image the same as the person presenting the document?
Is the document presented genuine? Is it a counterfeit/fake?
Minerva uses Credas technology to evaluate these two criteria completely independently.
Verifying the authenticity of the visual on the document presented – it is in the right position? Is all of the information present that would be expected? More in-depth analysis is then conducted; digital watermarks, and embossed text and bar -codes, identifying any flaws in the ID integrity.
Analysing all of the data on the document then gets verified through the database.
The client’s selfie is then evaluated. Texture analysis is used to ensure it is a genuine selfie. The selfie is then compared to the ID photo to make sure the faces match. To verify the movements made by the various pictures and actions taken, a combination of physical body moves and face tracking is used.
While machines do a lot of the work in verifying an ID, there is of course some level of human involvement. The machine learning algorithms need to be trained by humans and, although rare, whenever there is a high level of uncertainty the case is reviewed by a human.
With over 150,000 documents checked to date and a 97% accuracy rate, then the answer is yes, it works and it saves you time.