The fast growth of artificial intelligence (AI) raises much concern about the misinformation brought by AI-generated content (AIGC), especially Deepfake techniques that generate fake human faces. The recent development of Diffusion Models (DMs) moves another critical step forward to generate high-resolution and realistic human faces, which has become a challenge for existing Deepfake detectors. In this paper, we propose a DM-generated image detector by looking into the generation pipeline of DMs and the details of DM-generated images. The detector is based on the observation that DM-generated human faces show over-smooth textures and do not contain details as real human faces. Through a comprehensive analysis of DM-generated faces in spatial and frequency domains, we noticed that the over-smoothness improves the tolerance of Gaussian noise since excessive smoothness mitigates some of the impact of noise. Inspired by the observations, we propose a Deepfake detector capable of recognizing challenging DM-generated faces. We mainly propose the Noise Residual Unit (NRD) in our framework to collect the frequency response of images to Gaussian noise as distinctive features for classification. In detail, for an input face image, we add Gaussian noise to it and get the noise-degraded image. Then, the NRU generates the Noise Residual Image (NRI) by calculating the residual of the high-pass-filtered original image and the high-pass-filtered degraded image. The NRI indicates the high-frequency impact brought by the Gaussian noise and, therefore, suggests the tolerance of the original image to noise degradation. The original image and NRI are encoded and fused to obtain the joint representation, which is then fed to a classifier to predict the binary label. We conducted comprehensive experiments to evaluate the effectiveness of the proposed detector. The results indicate that our proposed detector achieves state-of-the-art detection performance on DM-generated faces and generalizes well to unseen DM-generated and GAN-generated face datasets.

The rise of Deepfake technology has sparked significant concerns due to its potential for misuse and malicious manipulation of multimedia content. Various detection approaches aimed at detecting Deepfake videos have been proposed, mostly relying on the identification of spatial and temporal artifacts. However, due to the different contexts of source images and the variety of generation techniques, current Deepfake detection methods usually perform well on training datasets, and yet generalize poorly to those unseen identities in new datasets. This issue is widely known as the generalization challenge of Deepfake detection. To address this challenge, this paper proposes an advanced spatiotemporal Deepfake video detector, named M otion-enhanced e nhanced S patiotemporal T ransformer (MeST-Former). MeST-Former is based on the spatiotemporal modeling capacity the video Swin Transformer. The spatial and temporal features are obtained from the RGB and motion images, respectively. To enhance the generalization ability of MeST-Former to unseen identities in unseen datasets, the ID-related components in the spatial and temporal features are detached. Specifically, MeST-Former adopts the newly proposed Identity-Decoupling Attention (IDC-Att) module to disentangle the ID-related and ID-unrelated components. Only the ID-unrelated components are used to construct more generalizable spatiotemporal representations. This process makes the constructed spatiotemporal features identity-agnostic and more generalizable to unseen identities. We conducted extensive experiments to evaluate the performance of the MeST-Former. Our results indicate that MeST-Former achieves accurate and generalizable Deepfake detection performance. Notably, MeST-Former also demonstrates high efficacy in detecting AI-animated talking-head videos.

The explosive progress of Deepfake techniques poses unprecedented privacy and security risks to our society by creating real-looking but fake visual content. The current Deepfake detection studies are still in their infancy because they mainly rely on capturing artifacts left by a Deepfake synthesis process as detection clues, which can be easily removed by various distortions (e.g. blurring) or advanced Deepfake techniques. In this paper, we propose a novel method that does not depend on identifying the artifacts but resorts to the mechanism of anti-counterfeit labels to protect face images from malicious Deepfake tampering. Specifically, we design a neural network with an encoder-decoder structure to embed watermarks as anti-Deepfake labels into the facial identity features. The injected label is entangled with the facial identity feature, so it will be sensitive to face swap translations (i.e., Deepfake) and robust to conventional image modifications (e.g., resize and compress). Therefore, we can identify whether watermarked images have been tampered with by Deepfake methods according to the label's existence. Experimental results demonstrate that our method can achieve average detection accuracy of more than 80%, which validates the proposed method's effectiveness in implementing Deepfake detection.

In this paper, we propose the Temporal Identity Inconsistency Network (TI 2 Net), a Deepfake detector that focuses on temporal identity inconsistency. Specifically, TI 2 Net recognizes fake videos by capturing the dissimilarities of human faces among video frames of the same identity. Therefore, TI 2 Net is a reference-agnostic detector and can be used on unseen datasets. For a video clip of a given identity, identity information in all frames will first be encoded to identity vectors. TI 2 Net learns the temporal identity embedding from the temporal difference of the identity vectors. The temporal embedding, representing the identity inconsistency in the video clip, is finally used to determine the authenticity of the video clip. During training, TI 2 Net incorporates triplet loss to learn more discriminative temporal embeddings. We conduct comprehensive experiments to evaluate the performance of the proposed TI 2 Net. Experimental results indicate that TI 2 Net generalizes well to unseen manipulations and datasets with unseen identities. Besides, TI 2 Net also shows robust performance against compression and additive noise.