ML Security PapersFrequency-Aware Semantic Fusion with Gated Injection for AI-generated Image Detection
Shuchang Zhou 1, Shangkun Wu 1, Jiwei Wei 1, Ke Liu 1, Ran Ran 1, Caiyan Qin 2, Yang Yang 1
Published on arXiv
2604.27875
Output Integrity Attack
OWASP ML Top 10 — ML09
Key Finding
Achieves state-of-the-art performance and strong generalization across multiple challenging datasets by reducing frequency shortcut bias and aligning semantic-frequency representations
FGINet
Novel technique introduced
AI-generated images are becoming increasingly realistic and diverse, posing significant challenges for generalizable detection. While Vision Foundation Models (VFMs) provide rich semantic representations and frequency-based methods capture complementary artifact cues, existing approaches that combine these modalities still suffer from limited generalization, with notable performance degradation on unseen generative models. We attribute this limitation to two key factors: frequency shortcut bias toward easily distinguishable cues associated with specific generators and cross-domain representation conflict between high-level semantics and low-level frequency patterns. To address these issues, we propose a Frequency-aware Gated Injection Network (FGINet) to improve generalization. Specifically, we design a Band-Masked Frequency Encoder (BMFE) that applies cross-band masking in the frequency domain to reduce reliance on generator-specific patterns and encourage more diverse and generalizable representations. We further introduce a Layer-wise Gated Frequency Injection (LGFI) mechanism to progressively inject frequency cues into the VFM backbone with adaptive gating, aligning with its hierarchical abstraction and alleviating representation conflict. Moreover, we propose a Hyperspherical Compactness Learning (HCL) framework with a cosine margin objective to learn compact and well-separated representations. Extensive experiments demonstrate that FGINet achieves state-of-the-art performance and strong generalization across multiple challenging datasets.
Key Contributions
- Band-Masked Frequency Encoder (BMFE) that applies cross-band masking to reduce reliance on generator-specific frequency shortcuts
- Layer-wise Gated Frequency Injection (LGFI) mechanism to progressively fuse frequency cues with VFM semantic features while alleviating representation conflict
- Hyperspherical Compactness Learning (HCL) framework with cosine margin objective for compact and well-separated representations
🛡️ Threat Analysis
Paper proposes a detection method for AI-generated images (deepfake/synthetic image detection), which is a core output integrity and content authenticity problem. The goal is to verify whether images are real or AI-generated, directly addressing content provenance and authentication.