ML Security PapersDeepfake Detection with Multi-Artifact Subspace Fine-Tuning and Selective Layer Masking
Xiang Zhang , Wenliang Weng , Daoyong Fu , Ziqiang Li , Zhangjie Fu
Published on arXiv
2601.01041
Output Integrity Attack
OWASP ML Top 10 — ML09
Key Finding
MASM achieves significant cross-dataset generalization advantages over existing methods across diverse forgery types, compression conditions, and complex distortion scenarios.
MASM (Multi-Artifact Subspaces and selective layer Masks)
Novel technique introduced
Deepfake detection still faces significant challenges in cross-dataset and real-world complex scenarios. The root cause lies in the high diversity of artifact distributions introduced by different forgery methods, while pretrained models tend to disrupt their original general semantic structures when adapting to new artifacts. Existing approaches usually rely on indiscriminate global parameter updates or introduce additional supervision signals, making it difficult to effectively model diverse forgery artifacts while preserving semantic stability. To address these issues, this paper proposes a deepfake detection method based on Multi-Artifact Subspaces and selective layer masks (MASM), which explicitly decouples semantic representations from artifact representations and constrains the fitting strength of artifact subspaces, thereby improving generalization robustness in cross-dataset scenarios. Specifically, MASM applies singular value decomposition to model weights, partitioning pretrained weights into a stable semantic principal subspace and multiple learnable artifact subspaces. This design enables decoupled modeling of different forgery artifact patterns while preserving the general semantic subspace. On this basis, a selective layer mask strategy is introduced to adaptively regulate the update behavior of corresponding network layers according to the learning state of each artifact subspace, suppressing overfitting to any single forgery characteristic. Furthermore, orthogonality constraints and spectral consistency constraints are imposed to jointly regularize multiple artifact subspaces, guiding them to learn complementary and diverse artifact representations while maintaining a stable overall spectral structure.
Key Contributions
- MASM applies SVD to pretrained model weights, partitioning them into a stable semantic principal subspace and multiple learnable artifact subspaces to decouple semantic representations from forgery artifact representations.
- A selective layer mask strategy adaptively regulates per-layer weight updates based on each artifact subspace's learning state, suppressing overfitting to any single forgery type.
- Orthogonality constraints and spectral consistency constraints jointly regularize multiple artifact subspaces to learn complementary, diverse artifact representations while preserving overall spectral structure.
🛡️ Threat Analysis
Paper proposes a novel deepfake face detection architecture — a core ML09 task (AI-generated/manipulated content detection). The primary contribution is a new forensic detection method (MASM) that improves cross-dataset generalization, not merely applying existing methods to a domain.