Current deepfake detection models achieve state-of-the-art performance on pristine academic datasets but suffer severe spatial attention drift under real-world compound degradations, such as blurring and severe lossy compression. To address this vulnerability, we propose a foundation-driven forensic framework that integrates an extreme compound degradation engine with a structurally constrained, multi-stream architecture. During training, our degradation pipeline systematically destroys high-frequency artifacts, optimizing the DINOv2-Giant backbone to extract invariant geometric and semantic priors. We then process images through three specialized pathways: a Global Texture stream, a Localized Facial stream, and a Hybrid Semantic Fusion stream incorporating CLIP. Through analyzing spatial attribution via Score-CAM and feature stability using Cosine Similarity, we quantitatively demonstrate that these streams extract non-redundant, complementary feature representations and stabilize attention entropy. By aggregating these predictions via a calibrated, discretized voting mechanism, our ensemble successfully suppresses background attention drift while acting as a robust geometric anchor. Our approach yields highly stable zero-shot generalization, achieving Fourth Place in the NTIRE 2026 Robust Deepfake Detection Challenge at CVPR. Code is available at https://github.com/khoalephanminh/ntire26-deepfake-challenge.

翻译：当前深度伪造检测模型在原始学术数据集上取得了最先进的性能，但在真实世界的复合退化（如模糊和严重有损压缩）下会出现严重的空间注意力漂移。为解决这一脆弱性，我们提出了一种基础驱动的取证框架，该框架集成了极端复合退化引擎与结构约束的多流架构。在训练过程中，我们的退化流水线系统性地破坏高频伪影，优化DINOv2-Giant骨干网络以提取不变几何和语义先验。随后，我们通过三条专门路径处理图像：全局纹理流、局部面部流以及融合CLIP的混合语义融合流。通过利用Score-CAM分析空间归因和余弦相似度分析特征稳定性，我们定量证明了这些流提取了非冗余、互补的特征表示，并稳定了注意力熵。通过经校准的离散化投票机制聚合这些预测，我们的集成成功抑制了背景注意力漂移，同时充当了鲁棒的几何锚点。我们的方法在零样本泛化中表现出高度稳定性，在CVPR的NTIRE 2026鲁棒深度伪造检测挑战赛中荣获第四名。代码发布在：https://github.com/khoalephanminh/ntire26-deepfake-challenge。