How Noise Benefits AI-generated Image Detection

TL;DR

This paper introduces PiN-CLIP, a novel noise-based training method that enhances the robustness and generalization of AI-generated image detection by suppressing shortcuts and amplifying stable forensic cues.

Contribution

It proposes a controllable positive-incentive noise technique that improves detection accuracy across diverse generative models, advancing out-of-distribution generalization.

Findings

Achieved a 5.4% improvement in average accuracy over existing methods.

Constructed feature-space noise via cross-attention to suppress shortcut-sensitive directions.

Demonstrated state-of-the-art performance on a dataset with 42 different generative models.

Abstract

The rapid advancement of generative models has made real and synthetic images increasingly indistinguishable. Although extensive efforts have been devoted to detecting AI-generated images, out-of-distribution generalization remains a persistent challenge. We trace this weakness to spurious shortcuts exploited during training and we also observe that small feature-space perturbations can mitigate shortcut dominance. To address this problem in a more controllable manner, we propose the Positive-Incentive Noise for CLIP (PiN-CLIP), which jointly trains a noise generator and a detection network under a variational positive-incentive principle. Specifically, we construct positive-incentive noise in the feature space via cross-attention fusion of visual and categorical semantic features. During optimization, the noise is injected into the feature space to fine-tune the visual encoder, suppressing shortcut-sensitive directions while amplifying stable forensic cues, thereby enabling the extraction of more robust and generalized artifact representations. Comparative experiments are conducted on an open-world dataset comprising synthetic images generated by 42 distinct generative models. Our method achieves new state-of-the-art performance, with notable improvements of 5.4 in average accuracy over existing approaches.