The rapid evolvement of deepfake creation technologies is seriously threating media information trustworthiness. The consequences impacting targeted individuals and institutions can be dire. In this work, we study the evolutions of deep learning architectures, particularly CNNs and Transformers. We identified eight promising deep learning architectures, designed and developed our deepfake detection models and conducted experiments over well-established deepfake datasets. These datasets included the latest second and third generation deepfake datasets. We evaluated the effectiveness of our developed single model detectors in deepfake detection and cross datasets evaluations. We achieved 88.74%, 99.53%, 97.68%, 99.73% and 92.02% accuracy and 99.95%, 100%, 99.88%, 99.99% and 97.61% AUC, in the detection of FF++ 2020, Google DFD, Celeb-DF, Deeper Forensics and DFDC deepfakes, respectively. We also identified and showed the unique strengths of CNNs and Transformers models and analysed the observed relationships among the different deepfake datasets, to aid future developments in this area.

翻译：深度伪造生成技术的快速演进严重威胁着媒体信息的可信度，对目标个人及机构可能造成灾难性后果。本研究系统探究了深度学习架构的演进历程，重点聚焦卷积神经网络与Transformer两类架构。我们遴选出八种具有前景的深度学习架构，设计并开发了相应的深度伪造检测模型，并在多个权威深度伪造数据集上开展实验——这些数据集囊括了最新的第二代和第三代深度伪造数据。通过单模型检测器的效能评估与跨数据集验证，我们在FF++ 2020、Google DFD、Celeb-DF、Deeper Forensics及DFDC数据集上分别取得88.74%、99.53%、97.68%、99.73%、92.02%的准确率，以及99.95%、100%、99.88%、99.99%、97.61%的AUC值。本研究还揭示了卷积神经网络与Transformer模型的独特优势，并分析了不同深度伪造数据集间的关联规律，以期为该领域的未来发展提供助益。