The Alpha Blending Hypothesis: Compositing Shortcut in Deepfake Detection
For deepfake detection researchers, this work reveals a critical shortcut in current detectors and provides a more robust training paradigm.
The paper identifies that deepfake detectors rely on low-level compositing artifacts (alpha blending) rather than semantic anomalies, and proposes BlenD, a method using self-blended images that achieves state-of-the-art cross-dataset generalization (94.0% AUROC in ensemble) without training on generated deepfakes.
Recent deepfake detection methods demonstrate improved cross-dataset generalization, yet the underlying mechanisms remain underexplored. We introduce the Alpha Blending Hypothesis, positing that state-of-the-art frame-based detectors primarily function as alpha blending searchers; rather than learning semantic anomalies or specific generative neural fingerprints, they localize low-level compositing artifacts introduced during the integration of manipulated faces into target frames. We experimentally validate the hypothesis, demonstrating that deepfake detectors exhibit high sensitivity to the so-called self-blended images (SBI) and non-generative manipulations. We propose the method BlenD that leverages a large-scale, diverse dataset of real-only facial images augmented with SBI. This approach achieves the best average cross-dataset generalization on 15 compositional deepfake datasets released between 2019 and 2025 without utilizing explicitly generated deepfakes during training. Furthermore, we show that predictions from explicit blending searchers and models resilient to blending shortcuts are highly complementary, yielding a state-of-the-art AUROC of 94.0% in an ensemble configuration. The code with experiments and the trained model will be publicly released.