Effective Synthetic Image Detection via Noise Residual Clustering
2026-07-12 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionCryptography and Security
AI summaryⓘ
The authors address the challenge of detecting AI-made fake images without needing large labeled datasets for training. They use a pre-trained model to find subtle noise patterns in images and extract features with a frozen Vision Transformer, then group images using a small sample of real images and K-Means clustering. This method works well across different types of fake images, especially those made by diffusion models, and does not require retraining for new AI generators. Their approach performs better than existing methods on tests and they plan to share their code publicly.
Generative AISynthetic ImagesImage AuthenticationNoise ResidualNoiseprint++Vision TransformerUnsupervised LearningK-Means ClusteringDiffusion ModelsGeneralization
Authors
Caihui Yan, Gang Cao, Huawei Tian, Zhen Li, Yuhang Zhai
Abstract
The rapid advancement of generative artificial intelligence (AI) has made synthetic images remarkably realistic, posing security threats such as misinformation and fraud. It is significant to detect the synthetic image in the manner of passive and blind image authentication. Most existing detectors rely on supervised training with large labeled datasets, leading to high costs and degraded performance on unknown generative models. To attenuate such deficiencies, we propose a training-free detection method. Specifically, noise residual fingerprints are first extracted by a simple yet effective pre-trained Noiseprint++ model. Then multi-scale features are further extracted from such residual by a frozen Vision Transformer (ViT), followed by adaptive weighted fusion. Only a few real image samples are used needed to initialize the clustering centers for unsupervised K-Means, distinguishing real and synthetic images without training. Extensive evaluations on four benchmark datasets show that our proposed scheme achieves an average accuracy of 82.2%, outperforming the state-of-the-art detectors on generalization ability. Superior performance is gained on the popular diffusion type of synthetic images, and the effectiveness of each module is validated by ablation studies. Source code will be publicly available at https://github.com/multimediaFor/NoiseCluSID.