The paper introduces a novel method for AI-generated image detection, focusing on the noise patterns of simple patches in images. The authors propose a Single Simple Patch (SSP) network that extracts and analyzes the noise fingerprints of these patches to distinguish between real and fake images. They argue that generative models tend to focus on generating rich textures while neglecting the hidden noise present in simple patches, which can serve as strong clues for detection. To address the issue of low-quality images, they introduce an enhancement module and a perception module to improve the quality of the patches and reduce interference from artifacts like blurring and compression. Extensive experiments on the GenImage and ForensSynths datasets demonstrate that their method outperforms existing techniques, achieving state-of-the-art performance across different generators. The contributions of the paper include the SSP network, which leverages noise fingerprints, and the enhanced SSP network, which improves detection accuracy on low-quality images.The paper introduces a novel method for AI-generated image detection, focusing on the noise patterns of simple patches in images. The authors propose a Single Simple Patch (SSP) network that extracts and analyzes the noise fingerprints of these patches to distinguish between real and fake images. They argue that generative models tend to focus on generating rich textures while neglecting the hidden noise present in simple patches, which can serve as strong clues for detection. To address the issue of low-quality images, they introduce an enhancement module and a perception module to improve the quality of the patches and reduce interference from artifacts like blurring and compression. Extensive experiments on the GenImage and ForensSynths datasets demonstrate that their method outperforms existing techniques, achieving state-of-the-art performance across different generators. The contributions of the paper include the SSP network, which leverages noise fingerprints, and the enhanced SSP network, which improves detection accuracy on low-quality images.