RoGA: Towards Generalizable Deepfake Detection through Robust Gradient Alignment
This work addresses domain shifts in deepfake detection, which is crucial for real-world applications, though it appears incremental as it builds on existing domain generalization techniques.
The paper tackles the problem of domain generalization in deepfake detection by proposing a novel learning objective that aligns generalization gradient updates with ERM gradient updates, resulting in improved performance over state-of-the-art methods on multiple datasets.
Recent advancements in domain generalization for deepfake detection have attracted significant attention, with previous methods often incorporating additional modules to prevent overfitting to domain-specific patterns. However, such regularization can hinder the optimization of the empirical risk minimization (ERM) objective, ultimately degrading model performance. In this paper, we propose a novel learning objective that aligns generalization gradient updates with ERM gradient updates. The key innovation is the application of perturbations to model parameters, aligning the ascending points across domains, which specifically enhances the robustness of deepfake detection models to domain shifts. This approach effectively preserves domain-invariant features while managing domain-specific characteristics, without introducing additional regularization. Experimental results on multiple challenging deepfake detection datasets demonstrate that our gradient alignment strategy outperforms state-of-the-art domain generalization techniques, confirming the efficacy of our method. The code is available at https://github.com/Lynn0925/RoGA.