Towards Bias-Aware and Interpretable DeepFake Detection through Semi-Supervised Facial Attribute Labeling

This study investigates the correlation between misclassifications in DeepFake detection and high-level facial attributes. A pre-trained frame-level classifier is used to distinguish between manipulated and authentic video contents and its wrong predictions are analyzed in detail. To enrich the dataset, we automatically annotate each video with additional labels, including gender, hair color, hair length, ear visibility and ethnicity, using a semi-supervised facial attribute recognition pipeline. An analysis of how misclassified video segments cross-reference with the visual attribute labels to identify emerging patterns is provided. Valuable insights for future bias-aware training strategies and more interpretable DeepFake detection systems are finally given.