Face forgery generating algorithms that produce a range of manipulated videos/images have developed quickly. As a result, this causes an increase in the production of fake information, making it difficult to identify. Since facial manipulation technologies raise severe concerns, face forgery detection is gaining steadily more attention in the area of computer vision. In real-world applications, face forgery detection systems frequently encounter and perform poorly in unseen domains due to poor generalization. In this paper, we propose a deepfake detection method based on meta learning named Meta Deepfake Detection (MDD). The goal of the model is to develop a generalized model capable of solving new unseen domain directly without the need for model updates. The MDD algorithm that establishes various weights for face images from various domains. Specifically, the MDD use meta-weight learning to shift information from source domains to target domains with meta-optimization steps, which aims for the model to generate effective representations of the source and target domains. We build multi-domain sets by meta splitting strategy to create meta-train set and meta-test set. Based on these sets, the model determines the gradient decent and get backpropagation. The inner loop gradient and outer loop gradient are aggregated to update the model to enhance generalization. By introducing pair-attention loss, average-center alignment loss, the detection capabilities of the system are substantially enhanced. Besides, we use some evaluation benchmarks established from several popular deepfake datasets to compare generalization of our proposal in several baselines and assess its effectiveness.

The rapid development of deep learning models that can produce and synthesize hyper-realistic videos are known as DeepFakes. Moreover, the growth of forgery data has prompted concerns about malevolent intent usage. Detecting forgery videos are a crucial subject in the field of digital media. Nowadays, most models are based on deep learning neural networks and vision transformer, SOTA model with EfficientNetB7 backbone. However, due to the usage of excessively large backbones, these models have the intrinsic drawback of being too heavy. In our research, a high performance DeepFake detection model for manipulated video is proposed, ensuring accuracy of the model while keeping an appropriate weight. We inherited content from previous research projects related to distillation methodology but our proposal approached in a different way with manual distillation extraction, target-specific regions extraction, data augmentation, frame and multi-region ensemble, along with suggesting a CNN-based model as well as flexible classification with a dynamic threshold. Our proposal can reduce the overfitting problem, a common and particularly important problem affecting the quality of many models. So as to analyze the quality of our model, we performed tests on two datasets. DeepFake Detection Dataset (DFDC) with our model obtains 0.958 of AUC and 0.9243 of F1-score, compared with the SOTA model which obtains 0.972 of AUC and 0.906 of F1-score, and the smaller dataset Celeb-DF v2 with 0.978 of AUC and 0.9628 of F1-score.

Deepfakes are digitally manipulated videos that appear realistic but are actually fake. With the rapid advances in deep generative models, the accessibility and sophistication of such manipulation technologies are increasing, making it more challenging to detect fake content. Different facial forgery techniques result in complex data distributions, and most existing deepfake detection approaches rely on convolutional neural networks (CNNs) that treat the task as a binary classification problem. While these methods achieve high accuracy on specific datasets, their generalization performance across datasets is often poor due to overfitting to manipulation techniques seen during training. In this study, we propose a model called MEViT, which integrates the EfficientNet Vision Transformer with a meta-learning framework to enhance generalization in deepfake detection. Furthermore, we introduce a pair-discrimination loss to push the feature representations of fake samples away from those of real samples, and a domain adjustment loss to reduce domain shifts across different manipulation methods. The MEViT model is trained on a specific manipulation method in the FaceForensics++ dataset and evaluated on other unseen methods from the same dataset. Additionally, we conduct extensive experiments on multiple deepfake benchmarks, including FaceForensics++ and CelebDF-v2, and compare our method with various state-of-the-art approaches to demonstrate its effectiveness.

Diffusion-based generative models have significantly advanced the field of image synthesis, presenting additional challenges regarding the integrity and authenticity of digital images. Consequently, the identification of AI-generated images has become a critical problem in image forensics. However, there is a lack of literature addressing the detection of images generated by diffusion models. In this paper, our focus is on developing a model capable of detecting images generated through both GAN techniques and diffusion models. We propose DiffCoR, a novel detection method for identifying AI-generated images. It consists of two main modules: Stable Diffusion Processing (SDP) and Image Representation Learning (IRL). The SDP module uses a pre-trained Stable Diffusion model to reconstruct input images via reverse diffusion and captures subtle manipulations through reconstruction discrepancies. The IRL module applies self-supervised learning with Latent Consistency Loss (LCL) to extract robust, invariant features, ensuring consistent latent representations across augmented views. We also incorporate frequency domain analysis using Discrete Fourier Transform (DFT) to enhance manipulation detection. Additionally, we introduce ForensicsImage, a publicly available dataset of over 400,000 real and AI-generated images from LSUN-Bedroom, CelebA-HQ, CelebDFv2, and various diffusion models. Experiments on ForensicsImage and GenImage show that DiffCoR achieves state-of-the-art performance, with strong cross-dataset generalization, making it suitable for real-world use in digital forensics, content verification, and social media moderation. Our code and dataset are available at https://github.com/nhantran214/diffcor.