From Spurious to Causal: Low-rank Orthogonal Subspace Intervention for Generalizable Face Forgery Detection

2026-01-17

AI Summary

This work addresses the limited generalization of face forgery detection models caused by spurious correlations induced by unobservable confounding factors in their representations. The authors propose a novel representation-space intervention paradigm that, for the first time, unifies diverse spurious correlations into a low-rank subspace and removes them from the original features via orthogonal low-rank projection. The model instead learns from the orthogonal complement subspace to capture genuine forgery cues. This approach obviates the need to explicitly model each confounding factor individually and introduces only 0.43 million trainable parameters. Despite its minimal overhead, it achieves state-of-the-art performance across multiple benchmarks, significantly enhancing both robustness and generalization capability.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

The generalization problem remains a critical challenge in face forgery detection. Some researches have discovered that ``a backdoor path"in the representations from forgery-irrelevant information to labels induces biased learning, thereby hindering the generalization. In this paper, these forgery-irrelevant information are collectively termed spurious correlations factors. Previous methods predominantly focused on identifying concrete, specific spurious correlation and designing corresponding solutions to address them. However, spurious correlations arise from unobservable confounding factors, making it impractical to identify and address each one individually. To address this, we propose an intervention paradigm for representation space. Instead of tracking and blocking various instance-level spurious correlation one by one, we uniformly model them as a low-rank subspace and intervene in them. Specifically, we decompose spurious correlation features into a low-rank subspace via orthogonal low-rank projection, subsequently removing this subspace from the original representation and training its orthogonal complement to capture forgery-related features. This low-rank projection removal effectively eliminates spurious correlation factors, ensuring that classification decision is based on authentic forgery cues. With only 0.43M trainable parameters, our method achieves state-of-the-art performance across several benchmarks, demonstrating excellent robustness and generalization.

Problem

• face forgery detection
• spurious correlations
• generalization
• confounding factors
• representation bias

Innovation

• low-rank subspace
• orthogonal projection
• spurious correlation
• causal intervention
• face forgery detection