AI image detection improves by tracking continuous authenticity signals
Learning Continuous Source Responses For Generalizable AI-Generated Image Detection
Computer Vision and Pattern Recognition
Summary
It is becoming very hard to tell AI-made images apart from real photos. The authors propose a new method called CuRe that detects AI-generated images by not just labeling images as real or fake but by measuring how much an image looks like each source. This approach helps the system learn more detailed clues about image authenticity beyond a simple yes-or-no decision. Their method works better than others across many tests and stays accurate even when images are degraded or altered.
What this means in practice
- •For social media platforms: Improve detection of manipulated or AI-generated images to reduce misinformation and increase user trust across varied sources.
- •For digital forensic teams: Enhance forensic image analysis by providing stronger generalization to new AI generators and robustness to image quality changes.
Authors
Manni Cui, Ruiqi Liu, Zijian Yu, Hao Tan, Zibo Wei, Zian Wang, Ziheng Qin, Huijia Zhu, Weiqiang Wang, Jun Lan, Shu Wu
Abstract
Advances in image generation have made synthetic images increasingly difficult to distinguish from real photographs, raising concerns about the trustworthiness of visual media. Existing AI-generated image detectors often perform well on in-domain data, but their robustness and cross-generator generalization remain limited. These limitations are commonly attributed to overfitting to shortcut cues. Although many methods seek to suppress shortcut learning, most retain binary classification as the training task without reconsidering how the task itself shapes the learned representations. We introduce CuRe, a framework for learning Continuous Source Responses that revisits authenticity detection from the perspective of the training task. CuRe reformulates backbone adaptation as regression of real-generated mixing ratios, providing finer supervision that encourages the model to capture authenticity-related variation beyond binary endpoint separation. We further select a compact source-response subspace to suppress nuisance variation and limit the final classifier's access to potential shortcut cues. Across ten public benchmarks, CuRe achieves an average balanced accuracy of 89.7%, exceeding the second-best method by 5.2 percentage points. Further experiments demonstrate consistent generalization gains across visual backbones and strong robustness to common image degradations. Code is available at https://github.com/manic-cui/CuRe