EKYC-DF: A REALISTIC DEEPFAKE CORPUS FOR TESTING AND TRAINING EKYC VERIFICATION MODELS

Abstract

Digital registration methods, such as electronic Know Your Customer (eKYC) checks, have become increasingly difficult to validate as a result of the widespread availability of deepfake computer technology. This has made the task of validating digital registration procedures more challenging. Within the context of deepfake attacks, the eKYCDF corpus is a particular dataset that has the potential to be employed for the purpose of evaluating and strengthening facial recognition systems. There are opportunities for both of these uses. This sample contains a considerable number of phony facial recordings, which are included in the collection. These phony recordings bear a strong resemblance to the ones that were actually available. The lighting, editing, and racial composition of the recordings are all notably different from one another, which makes it simple to discern amongst the recordings. By developing more effective methods of identity verification, researchers and developers have the power to safeguard the trust that users have in the internet and stop persons from gaining access to systems without authorization. This will allow them to enhance the security of electronic know-yourcustomer (eKYC) systems, which will allow them to better protect their customers.