Digital customer onboarding using video identification is now firmly established in the financial sector. The detection of attempted fraud is of great importance. The use of technologies from the field of artificial intelligence is now setting new standards in digital ID recognition.
Many banks and savings banks now practice digital customer onboarding using video identification. With the help of a procedure from the field of artificial intelligence, the identification provider of IDs is now setting new standards in digital ID recognition. For the video identification process, it uses algorithms for image recognition, which help to automatically check the ID document during the identification process. With this procedure, which is based on the deep learning principle, attempted fraud can be reliably detected and the security of the established procedure can be increased.
The procedure for online legitimation is now used in many countries by companies that are obliged under the Money Laundering Act to identify their customers before onboarding. A trained employee, the so-called identification specialist, checks the user’s identification document in a video chat. In addition to human expertise, highly specialized software is used to check that both the ID card and the identity of the user are genuine. This carries out an automatic image comparison and checks the validity of the ID using various features such as the code numbers.
Digital identities are valuable and therefore easily call for fraudsters. Some use deep learning algorithms for image recognition in order to detect forged IDs and technical manipulations during the identification process. The company developed the software behind it itself, as none of the existing technologies, for example, based on the principle of edge detection, have delivered satisfactory results in practice. Since video identification takes place in uncontrolled environments, the light, the background, and the resolution of the web or smartphone camera used to change from case to case.
In the next step, the program classifies the document, i.e. determines the country of issue, whether it is an identity card or passport, and the version of the card. Finally, the algorithm is able to read and check the data of the identification document such as the name of the holder and the check digits. These automatisms help the identification specialists to identify any discrepancies or attempted fraud. The algorithms for image recognition analyze a large amount of data and gradually learn to recognize relationships and patterns and to make corresponding predictions.
The technology in making ID can also be used for automatic face matching, depending on the requirements like convert2mp3 it will recognize the entered link before it shows the matching results. The algorithm uses 128 points on the face to identify a degree of similarity between the person in the video chat and the person on the identification document. The identification is only successful if the algorithm has determined that it is the same person. This comparison is also used to reduce attempted fraud, as already known fraudsters are filtered out at this point in the process.