In recent years, despite its wide use in various fields, deepfake has been abused to generate hazardous contents such as fake movies, rumors, and fake news by manipulating or replacing facial information of the original sources and, thus, exerts huge security threats to the society. Facing the continuous evolution of deepfake, research on active detection and prevention technology becomes particularly important. In this paper, we propose a new deepfake detection method based on cross-domain fusion, which, on the basis of traditional spatial domain features, realizes the fusion of cross-domain image features by introducing edge geometric features of the frequency domain and, therefore, achieves considerable improvements on classification accuracy. Further evaluations of this method have been performed on publicly deepfake datasets, and the results show that our method is effective particularly on the Meso-4 DeepFake Database.
Federated learning with data free knowledge distillation enables effective and privacy-preserving knowledge aggregation by employing generators to produce local pseudo samples during client-side model migration. However, in practical applications, data distributions across different institutions are often non-independent and identically distributed (Non-IID), which introduces bias in local models and consequently impedes the effective transfer of knowledge to the global model. In addition, insufficient local training can further exacerbate model bias, undermining overall performance. To address these challenges, we propose a heterogeneous federated learning framework that enhances knowledge transfer through guidance from global proxy data. Specifically, a noise filter is incorporated into the training of local generators to mitigate the negative impact of low-quality pseudo proxy samples on local knowledge distillation. Furthermore, a global generator is introduced to produce global pseudo proxy samples, which, together with local pseudo proxy data, are used to construct a cross attention matrix. This design effectively alleviates overfitting and underfitting issues in local models caused by data heterogeneity. Extensive experiments on publicly available datasets with heterogeneous data distributions demonstrate the superiority of the proposed framework. Results show that when the Dirichlet distribution coefficient is 0.05, our method achieves an average accuracy improvement of 5.77% over popular baselines; when the coefficient is 0.1, the improvement reaches 6.54%. Even under uniformly distributed sample classes, our model still achieves an average accuracy improvement of 7.07% compared to other methods.
Deepfake detection in video data typically relies on mining deep embedded representations across multiple modalities to obtain discriminative fused features and thereby improve detection accuracy. However, existing approaches predominantly focus on how to exploit complementary information across modalities to ensure effective fusion, while often overlooking the impact of noise and interference present in the data. For instance, issues such as small objects, blurring, and occlusions in the visual modality can disrupt the semantic consistency of the fused features. To address this, we propose a Multimodality Semantic Consistency Fusion model for video forgery detection. The model introduces a semantic consistency gating mechanism to enhance the embedding of semantically aligned information across modalities, thereby improving the discriminability of the fused representations. Furthermore, we incorporate an event-level weakly supervised loss to strengthen the global semantic discrimination of the video data. Extensive experiments on standard video forgery detection benchmarks demonstrate the effectiveness of the proposed method, achieving superior performance in both forgery event detection and localization compared to state-of-the-art approaches.
Face Forgery Detection (FFD) involves excavating more discriminative representations of face images, training recognition models to detect fake-faces, which is an important technology that ensures the authenticity, reliability, and security of facial information. However, the currently popular methods, due to excessive reliance on training data, only exhibit satisfactory detection performances within the same domain, but typically fail when applied to cross-domain scenarios. Therefore, how to achieve high generalization in cross-domain FFD with limited training-data has become a hot topic. Based on this, we propose the latent feature augmentation model for face forgery detection (LFAD) in cross-domain datasets. By constructing the Siamese Auto-Encoding Network, we can obtain label-consistent latent multi-view features, and introduce the contrast constraint to improve the discriminability of hard-samples. Furthermore, we simultaneously generate more diverse fake-features according to construction rules for high-generalization in efficiently training. Meanwhile, we induce the adaptive balance factor to avoid the imbalance category distribution caused by the generation of negative instances. Compared with other state-of-the-art methods, our model has not only achieved better performances in cross-domain and intra-domain detection, but also obtained faster detection efficiency. The code for our model has been open-sourced at: https://github.com/LNNU-computer-research-526/LFA-CDFD