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:随着Deepfake技术的快速演进,媒体信息的可信度受到了严重的威胁。对受影响的个人和机构来说,影响可能是灾难性的。在这项工作中,我们研究了深度学习架构的演变,特别是CNN和Transformer。我们确定了八种有前途的深度学习架构,设计和开发了我们的Deepfake检测模型,并在深受欢迎的Deepfake数据集上进行了实验。这些数据集包括最新的第二和第三代Deepfake数据集。我们评估了我们开发的单一模型检测器在Deepfake检测和数据集交叉评估方面的有效性。在FF++2020、Google DFD、Celeb-DF、Deeper Forensics和DFDC Deepfake检测方面,我们分别达到了88.74%、99.53%、97.68%、99.73%和92.02%的准确率以及99.95%、100%、99.88%、99.99%和97.61%的AUC。我们还确定并展示了CNN和Transformer模型的独特优势,并分析了不同的Deepfake数据集之间观察到的关系,以促进未来的发展。