Beyond Semantic Features: Pixel-level Mapping for Generalized AI-Generated Image Detection

The rapid evolution of generative technologies necessitates reliable methods for detecting AI-generated images. A critical limitation of current detectors is their failure to generalize to images from unseen generative models, as they often overfit to source-specific semantic cues rather than learning universal generative artifacts. To overcome this, we introduce a simple yet remarkably effective pixel-level mapping pre-processing step to disrupt the pixel value distribution of images and break the fragile, non-essential semantic patterns that detectors commonly exploit as shortcuts. This forces the detector to focus on more fundamental and generalizable high-frequency traces inherent to the image generation process. Through comprehensive experiments on GAN and diffusion-based generators, we show that our approach significantly boosts the cross-generator performance of state-of-the-art detectors. Extensive analysis further verifies our hypothesis that the disruption of semantic cues is the key to generalization.

Key Contributions

Empirically demonstrates that existing semantic-reduction methods (high-pass filtering, patch shuffling) fail to fully eliminate semantic bias, limiting cross-generator generalization.
Proposes a pixel-level mapping pre-processing step that suppresses low-frequency semantic patterns while amplifying high-frequency generative artifacts inherent to the image generation process.
Validates consistent cross-generator performance gains across multiple benchmarks covering both GAN and diffusion-based generators.

🛡️ Threat Analysis

Output Integrity Attack

Directly proposes a novel technique to improve AI-generated image detection across unseen generative models, addressing output integrity and content authenticity — the core of ML09.