Fair and Interpretable Deepfake Detection in Videos

2025-10-20

AI Summary

Existing deepfake detection methods suffer from demographic bias, inadequate temporal modeling, and poor interpretability. To address these issues, we propose a fairness-aware, temporally explicit, and interpretable detection framework. First, we introduce sequence clustering to model dynamic inter-frame patterns, enhancing temporal robustness. Second, we integrate frequency-domain transformations with concept-based explanation mechanisms to generate human-understandable forensic evidence. Third, we design a demographic-aware data augmentation strategy that preserves authentic manipulation artifacts while balancing sample distributions across demographic groups. Evaluated on FaceForensics++, DFD, Celeb-DF, and DFDC using Xception and ResNet backbones, our method achieves superior trade-offs between overall accuracy and group fairness—measured by metrics such as equalized odds—outperforming state-of-the-art approaches.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Humans and AI: Explainable AI (XAI) for Human Understanding
• Philosophy and Ethics of AI: Accountability, Interpretability & Explainability

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Responsible Web: Algorithmic accountability and transparency on the web
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Existing deepfake detection methods often exhibit bias, lack transparency, and fail to capture temporal information, leading to biased decisions and unreliable results across different demographic groups. In this paper, we propose a fairness-aware deepfake detection framework that integrates temporal feature learning and demographic-aware data augmentation to enhance fairness and interpretability. Our method leverages sequence-based clustering for temporal modeling of deepfake videos and concept extraction to improve detection reliability while also facilitating interpretable decisions for non-expert users. Additionally, we introduce a demography-aware data augmentation method that balances underrepresented groups and applies frequency-domain transformations to preserve deepfake artifacts, thereby mitigating bias and improving generalization. Extensive experiments on FaceForensics++, DFD, Celeb-DF, and DFDC datasets using state-of-the-art (SoTA) architectures (Xception, ResNet) demonstrate the efficacy of the proposed method in obtaining the best tradeoff between fairness and accuracy when compared to SoTA.

Problem

• Detecting biased deepfake videos across demographic groups
• Improving fairness and interpretability in video deepfake detection
• Addressing temporal modeling and transparency in fake video identification

Innovation

• Fairness-aware framework integrating temporal feature learning
• Demographic-aware data augmentation balancing underrepresented groups
• Concept extraction and sequence clustering for interpretable decisions