Explainable Synthetic Image Detection through Diffusion Timestep Ensembling
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909707664359424 |
|---|---|
| author | Wu, Yixin Zhang, Feiran Shi, Tianyuan Yin, Ruicheng Wang, Zhenghua Gan, Zhenliang Wang, Xiaohua Lv, Changze Zheng, Xiaoqing Huang, Xuanjing |
| author_facet | Wu, Yixin Zhang, Feiran Shi, Tianyuan Yin, Ruicheng Wang, Zhenghua Gan, Zhenliang Wang, Xiaohua Lv, Changze Zheng, Xiaoqing Huang, Xuanjing |
| contents | Recent advances in diffusion models have enabled the creation of deceptively real images, posing significant security risks when misused. In this study, we empirically show that different timesteps of DDIM inversion reveal varying subtle distinctions between synthetic and real images that are extractable for detection, in the forms of such as Fourier power spectrum high-frequency discrepancies and inter-pixel variance distributions. Based on these observations, we propose a novel synthetic image detection method that directly utilizes features of intermediately noised images by training an ensemble on multiple noised timesteps, circumventing conventional reconstruction-based strategies. To enhance human comprehension, we introduce a metric-grounded explanation generation and refinement module to identify and explain AI-generated flaws. Additionally, we construct the GenHard and GenExplain benchmarks to provide detection samples of greater difficulty and high-quality rationales for fake images. Extensive experiments show that our method achieves state-of-the-art performance with 98.91% and 95.89% detection accuracy on regular and challenging samples respectively, and demonstrates generalizability and robustness. Our code and datasets are available at https://github.com/Shadowlized/ESIDE. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_06201 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Explainable Synthetic Image Detection through Diffusion Timestep Ensembling Wu, Yixin Zhang, Feiran Shi, Tianyuan Yin, Ruicheng Wang, Zhenghua Gan, Zhenliang Wang, Xiaohua Lv, Changze Zheng, Xiaoqing Huang, Xuanjing Computer Vision and Pattern Recognition Artificial Intelligence Computation and Language Recent advances in diffusion models have enabled the creation of deceptively real images, posing significant security risks when misused. In this study, we empirically show that different timesteps of DDIM inversion reveal varying subtle distinctions between synthetic and real images that are extractable for detection, in the forms of such as Fourier power spectrum high-frequency discrepancies and inter-pixel variance distributions. Based on these observations, we propose a novel synthetic image detection method that directly utilizes features of intermediately noised images by training an ensemble on multiple noised timesteps, circumventing conventional reconstruction-based strategies. To enhance human comprehension, we introduce a metric-grounded explanation generation and refinement module to identify and explain AI-generated flaws. Additionally, we construct the GenHard and GenExplain benchmarks to provide detection samples of greater difficulty and high-quality rationales for fake images. Extensive experiments show that our method achieves state-of-the-art performance with 98.91% and 95.89% detection accuracy on regular and challenging samples respectively, and demonstrates generalizability and robustness. Our code and datasets are available at https://github.com/Shadowlized/ESIDE. |
| title | Explainable Synthetic Image Detection through Diffusion Timestep Ensembling |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2503.06201 |