Deepfake Detection with Deep Learning: Convolutional Neural Networks versus Transformers

被引:3
作者
Thing, Vrizlynn L. L. [1 ]
机构
[1] Singapore Technol Engn, Singapore, Singapore
来源
2023 IEEE INTERNATIONAL CONFERENCE ON CYBER SECURITY AND RESILIENCE, CSR | 2023年
关键词
deepfakes; misinformation; detection; deep learning; convolutional neural networks; transformers; authenticity verification;
D O I
10.1109/CSR57506.2023.10225004
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
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.
引用
收藏
页码:246 / 253
页数:8
相关论文
共 54 条
  • [1] Afchar D, 2018, IEEE INT WORKS INFOR
  • [2] [Anonymous], FakeApp
  • [3] [Anonymous], 2020, Disney's deepfake technology could be used in film and TV
  • [4] Bao Hangbo, 2021, arXiv, DOI [10.48550/arXiv.2106.08254, DOI 10.48550/ARXIV.2106.08254]
  • [5] Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security
    Chesney, Bobby
    Citron, Danielle
    [J]. CALIFORNIA LAW REVIEW, 2019, 107 (06) : 1753 - 1819
  • [6] Xception: Deep Learning with Depthwise Separable Convolutions
    Chollet, Francois
    [J]. 30TH IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2017), 2017, : 1800 - 1807
  • [7] Cizek K., 2019, Collective Wisdom, V1st
  • [8] Dolhansky B, 2020, Arxiv, DOI arXiv:2006.07397
  • [9] Dolhansky B, 2019, Arxiv, DOI arXiv:1910.08854
  • [10] Dosovitskiy A, 2021, Arxiv, DOI [arXiv:2010.11929, DOI 10.48550/ARXIV.2010.11929]