[4M3-GS-10] AI application: Optimization / Visualization

[4M3-GS-10-03] Research on detection methods of spoofing attacks for eKYC using Deepfake

〇Non Kawana¹, Keishi Ooshima¹, Akane Suzuki¹, Masayuki Yoshino¹ (1. Hitachi, Ltd.)

Keywords:Deepfake, eKYC, Detection

eKYC is a method of online identity verification. In 2021, we presented the potential risk of identity spoofing attacks against eKYC using Deepfake. In this paper, we report the results of our evaluation of countermeasure techniques against this spoofing attack. First, we created a real video that simulates eKYC and a fake video using Deepfake. Fake videos can be roughly classified into two types: FaceSwap, which replaces a person's face with someone else's, and Reenactment, which changes a person's expression. We evaluated the performance of the Deepfake detection technology, which is publicly available as OSS, using the videos we created. As a result of our evaluation, we found that no Deepfake detection technology could correctly detect both types of fake videos. There is no all-purpose Deepfake detection technology for all fake videos, and it is important to combine multiple detection technologies to deal with various types of Deepfake.