Abstract
With the development of Metaverse technology, the avatar in Metaverse has faced serious security andprivacy concerns. Analyzing facial features to distinguish between genuine and manipulated facial videosholds significant research importance for ensuring the authenticity of characters in the virtual world and formitigating discrimination as well as preventing malicious use of facial data. To address this issue, the FacialFeature Points and Class-head-Transformer (FFP-ChT) deepfake video detection model is designed based onthe clues of different FFPs distribution in real and fake videos and different displacement distances of real andfake FFPs between frames. The face video input is first detected by the BlazeFace model, and the face detectionresults are fed into the FaceMesh model to extract 468 FFPs. Then, the Lucas–Kanade (LK) optical flow methodis used to track the points of the face, the face calibration algorithm is introduced to re-calibrate the FFPs, andthe jitter displacement is calculated by tracking the FFPs between frames. Finally, the Ch is designed in thetransformer, and the FFPs and FFP displacement are jointly classified through the ChT model. In this way, thedesigned ChT classifier is able to accurately and effectively identify deepfake videos. Experiments on opendatasets clearly demonstrate the effectiveness and generalization capabilities of our approach.
NETL
NSTL
Assoc Computing Machinery