Existing zero-shot deepfake detection methods are often constrained to specific scenarios and struggle in diverse, complex scenarios. To address this limitation, we propose a universal zero-shot deepfake detection method. This method models the common forgery traces across different domains as domain-invariant features and introduces a novel domain-invariant meta-learning strategy. This strategy embeds the mechanism of domain-invariant learning into a meta-learning framework, enabling the model not only to extract specific domain-invariant features from certain domains, but also to leverage the meta-learning mechanism of fast adaptation to new domains. As a result, the model is capable of effectively capturing the intrinsic domain-invariant characteristics of deepfake images, thereby achieving universal zero-shot deepfake detection. Extensive comparative experiments demonstrate that the proposed method achieves the highest average detection AUC (86.96%) across 28 unseen datasets, representing an improvement of 8.02% over the second-best method (78.94%). Moreover, it is the only method that is effective in all four zero-shot scenarios, which strongly validates its superior zero-shot detection performance and universality. Code is released at https://github.com/QinQin741/DIML.

In practical applications of social media and the Internet, deepfake face images involve a plethora of unlabeled samples. To effectively identify unlabeled deepfake images, the domain adaptation technique has gained significant attention. It applies the knowledge learned from labeled samples (source domain) to unlabeled samples (target domain) in a cross-domain manner. However, the existing domain adaptation-based deepfake detection methods primarily focus on intra-type cross-domain scenarios. In this study, we propose an unsupervised domain adaptation-based deepfake face image detection method for extra-type cross-domain scenarios. The core idea of our approach lies in the development of a domain adaptation model that consists of Domain Tag Adversarial (DTA) and Domain Feature Alignment (DFA) algorithms, called DTA-DFA, which empowers the proposed method with strong cross-domain capability. The DTA is utilized to weaken the specificity within each domain, while DFA aligns the distribution between the source and target domains. Compared with the existing deepfake detection methods, the experimental results demonstrate that the proposed method dramatically enhances the extra-type cross-domain detection performance. Moreover, the DTA-DFA model also exhibits a remarkable ability to perform cross-domain detection from large-shot labeled samples to few-shot labeled samples, further verifying its powerful cross-domain capability. Code is released at https://github.com/QinQin741/DTA-DFA-DA-model.

Time-series anomaly detection is a critical task with significant impact as it serves a pivotal role in the field of data mining and quality management. Current anomaly detection methods are typically based on reconstruction or forecasting algorithms, as these methods have the capability to learn compressed data representations and model time dependencies. However, most methods rely on learning normal distribution patterns, which can be difficult to achieve in real-world engineering applications. Furthermore, real-world time-series data is highly imbalanced, with a severe lack of representative samples for anomalous data, which can lead to model learning failure. In this article, we propose a novel end-to-end unsupervised framework called the parallel-attention transformer (PAFormer), which discriminates anomalies by modeling both the global characteristics and local patterns of time series. Specifically, we construct parallel-attention (PA), which includes two core modules: the global enhanced representation module (GERM) and the local perception module (LPM). GERM consists of two pattern units and a normalization module, with attention weights that indicate the relationship of each data point to the whole series (global). Due to the rarity of anomalous points, they have strong associations with adjacent data points. LPM is composed of a learnable Laplace kernel function that learns the neighborhood relevancies through the distributional properties of the kernel function (local). We employ the PA to learn the global-local distributional differences for each data point, which enables us to discriminate anomalies. Finally, we propose a two-stage adversarial loss to optimize the model. We conduct experiments on five public benchmark datasets (real-world datasets) and one synthetic dataset. The results show that PAFormer outperforms state-of-the-art baselines.

With the rapid development of artificial intelligence technology, a variety of GAN generated deepfake face images/videos have emerged endlessly. The abuse of deepfake has brought serious negative effects to many industries. Therefore, there is an urgent need to develop advanced methods to combat the abuse of deepfake. As far as we know, there are almost no techniques that can distinguish multiple types of homologous deepfake face images. In this study, we propose a method based on the multi-classification task to address this issue. The proposed method relies on a novel network framework named FCD-Net that consists of the facial synaptic saliency module (FSS), the contour detail feature extraction module (CDFE), and the distinguishing feature fusion module (DFF). Utilizing this method, the imperceptible features introduced by deepfake can be exposed, and the differences caused by different types of deepfake can be distinguished, even if deepfake images are homologous. To test the proposed method and compare it with other SOTA methods, we establish a new homologous dataset named HDFD that contains real face images, entire face synthesis images, face swap images, and facial attribute manipulation images. Among them, the three types of deepfake images are all generated from the same real face images through different deepfake techniques. Abundant experiment results demonstrate that the proposed method has a high-level detection accuracy and relatively strong robustness against content-preserving manipulations. Moreover, the generalization of our method is superior to other SOTA methods.