Synthbuster: Towards Detection of Diffusion Model Generated Images

Abstract

Synthetically-generated images are getting increasingly popular. Diffusion models have advanced to the stage where even non-experts can generate photo-realistic images from a simple text prompt. They expand creative horizons but also open a Pandora's box of potential disinformation risks. In this context, the present corpus of synthetic image detection techniques, primarily focusing on older generative models like Generative Adversarial Networks, finds itself ill-equipped to deal with this emerging trend. Recognizing this challenge, we introduce a method specifically designed to detect synthetic images produced by diffusion models. Our approach capitalizes on the inherent frequency artefacts left behind during the diffusion process. Spectral analysis is used to highlight the artefacts in the Fourier transform of a residual image, which are used to distinguish real from fake images. The proposed method can detect diffusion-model-generated images even under mild JPEG compression, and generalizes relatively well to unknown models. By pioneering this novel approach, we aim to fortify forensic methodologies and ignite further research into the detection of AI-generated images.

Are Deepfakes a Game-changer in Digital Images Steganography Leveraging the Cover-Source-Mismatch?

“…Steganalysis is performed with EfficientNet-v2 against the MiPOD embedding algorithm with payload 0.4 bpp (Bits Per Pixels). Similar conclusions can be drawn from [3]: the inconsistency between deepfake cover-sources is very important even for models supposedly close to each other (see the family of StableDiffusion for instance). However, one can observe that the only common cover-source, namely StableDiffusion-XL, exhibits similar results as the one reported in Table 6.…”

“…Note that these important inconsistencies do not stem from different prompts; indeed, we used the same prompts for every generator; therefore, high inconsistencies can only be explained by the deepfake cover-sources. We further confirmed these observations using an additional dataset from [3]. Table 8 shows a selective subset of inconsistencies obtained when training on the same 15 generators as in Table 7 and 6, but testing on these additional images, as for Table 6.…”