DEEP FAKE CHALLENGES IN E-KYC:A REALISTIC DATASET FOR TRAINING AND TESTING VERIFICATION MODELS

Abstract

Due to deepfake technology's broad availability, evaluating digital registration processes, such as eKYC verifications, has become more difficult. So, it's getting harder and harder to verify digital registration processes. For the purpose of testing and improving facial recognition systems against deepfake attacks, the eKYC-DF corpus is a specific dataset. Both applications are doable. There are a lot of fake face recordings in this sample that are part of the collection. These fake recordings sound almost identical to the real ones. Light, editing, and racial composition vary greatly between the recordings, making them easily identifiable from one another. Better identity verification methods can be developed by researchers and developers to protect consumers' trust in the internet and prevent unauthorized access to services. By fortifying the reliability of eKYC systems, they will improve consumer safety.