Fair Deepfake Detectors Can Generalize

Harry Cheng, Ming-Hui Liu, Yangyang Guo, Tianyi Wang, Liqiang Nie, Mohan Kankanhalli

National University of Singapore· Shandong University· Harbin Institute of Technology

Deepfake Detection; Causal Inference; Generalization; Fairness; Back-door Adjustment

⋅ NeurIPS ⋅ Project Page ⋅Slides ⋅Poster ⋅OpenReview

Abstract

Deepfake detection models face two critical challenges: generalization to unseen manipulations and demographic fairness among population groups. However, existing approaches often demonstrate that these two objectives are inherently conflicting, revealing a trade-off between them.

In this paper, we, for the first time, uncover and formally define a causal relationship between fairness and generalization. Building on the back-door adjustment, we show that controlling for confounders (data distribution and model capacity) enables improved generalization via fairness interventions.

Motivated by this insight, we propose Demographic Attribute-insensitive Intervention Detection (DAID), a plug-and-play framework composed of:

Demographic-aware data rebalancing, which employs inverse-propensity weighting and subgroup-wise feature normalization to neutralize distributional biases; and
Demographic-agnostic feature aggregation, which uses a novel alignment loss to suppress sensitive-attribute signals.

Across three cross-domain benchmarks, DAID consistently achieves superior performance in both fairness and generalization compared to several state-of-the-art detectors, validating both its theoretical foundation and practical effectiveness.