Recently, the proliferation of highly realistic synthetic images, facilitated through a variety of GANs and Diffusions, has significantly heightened the susceptibility to misuse. While the primary focus of deepfake detection has traditionally centered on the design of detection algorithms, an investigative inquiry into the generator architectures has remained conspicuously absent in recent years. This paper contributes to this lacuna by rethinking the architectures of CNN-based generators, thereby establishing a generalized representation of synthetic artifacts. Our findings illuminate that the up-sampling operator can, beyond frequency-based artifacts, produce generalized forgery artifacts. In particular, the local interdependence among image pixels caused by upsampling operators is significantly demonstrated in synthetic images generated by GAN or diffusion. Building upon this observation, we introduce the concept of Neighboring Pixel Relationships(NPR) as a means to capture and characterize the generalized structural artifacts stemming from up-sampling operations. A comprehensive analysis is conducted on an open-world dataset, comprising samples generated by \tft{28 distinct generative models}. This analysis culminates in the establishment of a novel state-of-the-art performance, showcasing a remarkable \tft{12.8\%} improvement over existing methods. The code is available at https://github.com/chuangchuangtan/NPR-DeepfakeDetection.

翻译：近年来，通过多种生成对抗网络（GAN）和扩散模型产生的高度逼真的合成图像迅速泛滥，显著增加了被滥用的风险。尽管深度伪造检测的主要关注点传统上集中于检测算法的设计，但近年来针对生成器架构的调查研究却明显缺失。本文通过重新思考基于CNN的生成器架构，弥补了这一空白，从而建立了合成伪影的通用表示。我们的研究结果表明，上采样算子除了产生基于频率的伪影外，还能产生通用的伪造伪影。特别是，由GAN或扩散模型生成的合成图像中，上采样算子导致的图像像素间的局部相互依赖性显著体现。基于这一观察，我们引入了相邻像素关系（Neighboring Pixel Relationships, NPR）的概念，以此捕捉并表征由上采样操作产生的通用结构伪影。我们在一个开放世界数据集上进行了全面分析，该数据集包含由\tft{28种不同生成模型}生成的样本。这一分析最终确立了新的最先进性能，比现有方法提升了\tft{12.8\%}的显著进步。代码可在https://github.com/chuangchuangtan/NPR-DeepfakeDetection获取。