Problem
Research questions and friction points this paper is trying to address.
Deepfake Detection
Explainability
Concept Bottleneck Models
Forensic Analysis
DF-CBM: Region-Aware Concept Bottleneck Models for Deepfake Detection
🤖 AI Summary
This study addresses the opacity of prediction rationales and the lack of semantic-level visual evidence localization in deepfake detection by proposing a region-aware concept bottleneck model. The core innovation lies in, for the first time, mapping text-annotated manipulation concepts to specific facial regions, achieving spatially grounded concept prediction through the integration of parsed facial masks with a masked attention mechanism. Experimental results demonstrate that the proposed model outperforms existing baselines in both concept prediction and classification tasks. Furthermore, it generates counterfactual explanations to enhance transparency while achieving detection performance comparable to mainstream black-box detectors, thereby effectively unifying high interpretability with strong detection capability.
📝 Abstract
Deepfake detection methods have become increasingly effective yet most provide limited insight into the evidence behind their predictions. However, in forensic settings users also need to know which manipulation cues support the decision and where they appear. Existing explainability methods only partially address this need since localization-based approaches lack semantic descriptions while language-based explanation methods are only weakly grounded in visual evidence. In this work, we propose DF-CBM, a region-aware concept bottleneck model for explainable deepfake detection. DF-CBM builds a compact vocabulary of manipulation-related concepts from textual artifact annotations and links each concept to plausible facial and boundary regions. It then predicts these concepts from visual features using a concept-specific masked attention mechanism guided by parsed facial masks and the final real/fake decision is made from the predicted concept bottleneck. Our experiments show that DF-CBM outperforms concept-based baselines in concept prediction and deepfake classification while remaining competitive with state-of-the-art black-box detectors. Finally, qualitative results and intervention analyses demonstrate that DF-CBM provides spatially grounded concept evidence and enables counterfactual explanations of how individual manipulation concepts influence the final prediction. Our code is available at: https://github.com/GeorgeTsoumplekas/DF-CBM.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Concept Bottleneck Model
Deepfake Detection
Region-Aware
Masked Attention Mechanism
Counterfactual Explanations
🔎 Similar Papers
No similar papers found.
