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

2025-12-19

AI Summary

AI-generated image detectors suffer from poor generalization, primarily due to over-reliance on model-specific semantic cues rather than universal generative artifacts. Method: We propose an unsupervised pixel-wise nonlinear mapping preprocessing technique that actively perturbs pixel distributions to eliminate semantic shortcuts, thereby forcing detectors to attend to high-frequency domain artifacts common across diverse generators. This approach explicitly integrates semantic shortcut suppression into the robustness enhancement framework and combines frequency-domain analysis with fine-tuning of ResNet- and ViT-based detectors. Contribution/Results: To our knowledge, this is the first work to explicitly incorporate semantic shortcut mitigation into detection robustness design. Our method enables cross-generator evaluation between GANs and diffusion models. Experiments demonstrate a 12.7% average improvement in zero-shot detection accuracy on unseen generators, significantly enhancing the generalization capability of state-of-the-art detectors without increasing inference overhead.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation
• Machine Learning: Adversarial Learning & Robustness

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

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.

Problem

• Detects AI-generated images across unseen generative models.
• Disrupts pixel distribution to break semantic shortcut patterns.
• Forces focus on generalizable high-frequency generative artifacts.

Innovation

• Pixel-level mapping disrupts image distribution
• Forces focus on high-frequency generative traces
• Enhances cross-generator detection generalization