GrDT: Towards Robust Deepfake Detection Using Geometric Representation Distribution and Texture

In recent years, deepfake images and videos have rapidly spread across social media platforms. This poses signifi-cant threats to public privacy, property, and safety. Several detection methods have been proposed, with the most common approaches being those based on deep learning and the analysis of biometric signals. However, these methods generally suffer from poor generalization capa-bilities, struggle to detect high-quality deepfakes, and de-pend on high-resolution training data. Based on these observations, we propose a detection method based on a Graph Attention Network (GAT) and biometric features, referred to as GrDT. The core idea of GrDT is to iden-tify deepfake face images by leveraging facial texture representations and the geometric relationships of key facial points. Cross-validation on the DF40 and ForgeryNet datasets shows that GrDT outperforms other methods in terms of the AP and AUC metrics. The code is available at https://github.com/SIPLab24IGrDT.