arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2210.14571cs.CV
收藏

面向扩散模型深度伪造的检测研究

Towards the Detection of Diffusion Model Deepfakes

Ruhr University Bochum(波鸿鲁尔大学);CISPA Helmholtz Center for Information Security(CISPA亥姆霍兹信息安全中心)

Jonas Ricker, Simon Damm, Thorsten Holz, Asja Fischer

AI总结:

本文评估现有GAN检测器对扩散模型生成图像的检测能力,发现重训练后可实现近乎完美检测且能泛化到GAN,并指出扩散模型因缺少网格状频率伪影而更难检测。

摘要:

在过去几年中,扩散模型(DMs)已达到前所未有的视觉质量水平。然而,针对扩散模型生成图像的检测研究相对较少受到关注,而这对于防止其对社会造成不利影响至关重要。相比之下,生成对抗网络(GANs)已从取证角度得到了广泛研究。因此,在本工作中,我们顺理成章地迈出下一步,评估以往的方法是否可用于检测由扩散模型生成的图像。我们的实验得出两个关键发现:(1)最先进的GAN检测器无法可靠地区分真实图像与扩散模型生成的图像,但(2)在扩散模型生成图像上重新训练这些检测器后,可以实现几乎完美的检测效果,并且该效果显著地甚至能泛化到GAN生成图像。结合特征空间分析,我们的结果引出一个假设:与GAN相比,扩散模型产生的可检测伪影更少,因此更难被检测。一个可能的原因是扩散模型生成的图像中不存在网格状频率伪影,而这类伪影是GAN的已知弱点。然而,我们有一个有趣的观察:扩散模型倾向于低估高频分量,我们将其归因于学习目标。

英文摘要:

In the course of the past few years, diffusion models (DMs) have reached an unprecedented level of visual quality. However, relatively little attention has been paid to the detection of DM-generated images, which is critical to prevent adverse impacts on our society. In contrast, generative adversarial networks (GANs), have been extensively studied from a forensic perspective. In this work, we therefore take the natural next step to evaluate whether previous methods can be used to detect images generated by DMs. Our experiments yield two key findings: (1) state-of-the-art GAN detectors are unable to reliably distinguish real from DM-generated images, but (2) re-training them on DM-generated images allows for almost perfect detection, which remarkably even generalizes to GANs. Together with a feature space analysis, our results lead to the hypothesis that DMs produce fewer detectable artifacts and are thus more difficult to detect compared to GANs. One possible reason for this is the absence of grid-like frequency artifacts in DM-generated images, which are a known weakness of GANs. However, we make the interesting observation that diffusion models tend to underestimate high frequencies, which we attribute to the learning objective.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑