Researchers have introduced a new metric called Cross-AUC to better evaluate the performance of deepfake detectors. Traditional methods using Area Under the ROC Curve (AUC) can be misleading when detectors encounter data from different sources or varying artifact types. Cross-AUC addresses this by averaging per-domain AUCs with a measure of prediction polarization, quantified by the Wasserstein Distance, to more realistically assess generalization capabilities under domain shifts. Experiments on seven benchmark datasets indicate that Cross-AUC provides a more interpretable and practical evaluation of deepfake detection robustness. AI
IMPACT This new metric could lead to more robust deepfake detection systems by providing a more realistic assessment of their performance across diverse data sources.
RANK_REASON The cluster contains a research paper introducing a novel evaluation metric for deepfake detectors.
- arXiv
- Cross-AUC
- deepfake
- deepfake detectors
- Diffusion Models
- face-swapping tools
- generative artificial intelligence
- Wasserstein metric
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
