Thinking Racial Bias in Fair Forgery Detection: Models, Datasets and Evaluations

2024-07-19

Proceedings of the AAAI Conference on Artificial Intelligence

AI Summary

This work addresses the pervasive racial bias in deepfake detection. We introduce FairFD, the first large-scale, multi-ethnic balanced benchmark dataset, systematically exposing cross-racial performance disparities across state-of-the-art (SOTA) detectors. To rigorously quantify fairness, we propose two novel evaluation metrics: average performance metrics and utility-regularized fairness metrics. Furthermore, we design BPFA (Bias Pruning with Fair Activations), a post-hoc, training-free debiasing method that jointly improves detection accuracy and fairness—marking the first such approach. Extensive evaluation across 12 mainstream detectors demonstrates that FairFD effectively reveals latent racial biases in existing methods, while BPFA elevates the best-performing detector to a new SOTA, achieving both high overall accuracy and significantly enhanced cross-racial robustness and fairness balance.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Ethics, Bias, and Fairness
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Social Networks and Social Media: Fairness and bias in social network and social media analysis
• Responsible Web: Fairness, accountability, transparency and ethics of web technologies

Abstract

Due to the successful development of deep image generation technology, forgery detection plays a more important role in social and economic security. Racial bias has not been explored thoroughly in the deep forgery detection field. In the paper, we first contribute a dedicated dataset called the Fair Forgery Detection (FairFD) dataset, where we prove the racial bias of public state-of-the-art (SOTA) methods. Different from existing forgery detection datasets, the self-constructed FairFD dataset contains a balanced racial ratio and diverse forgery generation images with the largest-scale subjects. Additionally, we identify the problems with naive fairness metrics when benchmarking forgery detection models. To comprehensively evaluate fairness, we design novel metrics including Approach Averaged Metric and Utility Regularized Metric, which can avoid deceptive results. We also present an effective and robust post-processing technique, Bias Pruning with Fair Activations (BPFA), which improves fairness without requiring retraining or weight updates. Extensive experiments conducted with 12 representative forgery detection models demonstrate the value of the proposed dataset and the reasonability of the designed fairness metrics. By applying the BPFA to the existing fairest detector, we achieve a new SOTA. Furthermore, we conduct more in-depth analyses to offer more insights to inspire researchers in the community.

Problem

• Explores racial bias in deep forgery detection models
• Introduces FairFD dataset with balanced racial representation
• Proposes novel fairness metrics and Bias Pruning technique

Innovation

• Developed FairFD dataset with balanced racial ratio
• Designed novel fairness metrics for evaluation
• Proposed BPFA technique for bias mitigation