In recent years,with the rapid growth of generative adversarial networks(GANs),a photo-realistic face can be easily generated from a random vector.Moreover,the faces generated by advanced GANs are very realistic.It is...In recent years,with the rapid growth of generative adversarial networks(GANs),a photo-realistic face can be easily generated from a random vector.Moreover,the faces generated by advanced GANs are very realistic.It is reasonable to acknowledge that even a well-trained viewer has difficulties to distinguish artificial from real faces.Therefore,detecting the face generated by GANs is a necessary work.This paper mainly introduces some methods to detect GAN-generated fake faces,and analyzes the advantages and disadvantages of these models based on the network structure and evaluation indexes,and the results obtained in the respective data sets.On this basis,the challenges faced in this field and future research directions are discussed.展开更多
Due to the power of editing tools,new types of fake faces are being created and synthesized,which has attracted great attention on social media.It is reasonable to acknowledge that one human cannot distinguish whether...Due to the power of editing tools,new types of fake faces are being created and synthesized,which has attracted great attention on social media.It is reasonable to acknowledge that one human cannot distinguish whether the face is manipulated from the real faces.Therefore,the detection of face manipulation becomes a critical issue in digital media forensics.This paper provides an overview of recent deep learning detection models for face manipulation.Some public dataset used for face manipulation detection is introduced.On this basis,the challenges for the research and the potential future directions are analyzed and discussed.展开更多
目的随着深度伪造技术的快速发展,人脸伪造图像越来越难以鉴别,对人们的日常生活和社会稳定造成了潜在的安全威胁。尽管当前很多方法在域内测试中取得了令人满意的性能表现,但在检测未知伪造类型时效果不佳。鉴于伪造人脸图像的伪造区...目的随着深度伪造技术的快速发展,人脸伪造图像越来越难以鉴别,对人们的日常生活和社会稳定造成了潜在的安全威胁。尽管当前很多方法在域内测试中取得了令人满意的性能表现,但在检测未知伪造类型时效果不佳。鉴于伪造人脸图像的伪造区域和非伪造区域具有不一致的源域特征,提出一种基于多级特征全局一致性的人脸深度伪造检测方法。方法使用人脸结构破除模块加强模型对局部细节和轻微异常信息的关注。采用多级特征融合模块使主干网络不同层级的特征进行交互学习,充分挖掘每个层级特征蕴含的伪造信息。使用全局一致性模块引导模型更好地提取伪造区域的特征表示,最终实现对人脸图像的精确分类。结果在两个数据集上进行实验。在域内实验中,本文方法的各项指标均优于目前先进的检测方法,在高质量和低质量FaceForensics++数据集上,AUC(area under the curve)分别达到99.02%和90.06%。在泛化实验中,本文的多项评价指标相比目前主流的伪造检测方法均占优。此外,消融实验进一步验证了模型的每个模块的有效性。结论本文方法可以较准确地对深度伪造人脸进行检测,具有优越的泛化性能,能够作为应对当前人脸伪造威胁的一种有效检测手段。展开更多
基金supported by National Natural Science Foundation of China(62072251).
文摘In recent years,with the rapid growth of generative adversarial networks(GANs),a photo-realistic face can be easily generated from a random vector.Moreover,the faces generated by advanced GANs are very realistic.It is reasonable to acknowledge that even a well-trained viewer has difficulties to distinguish artificial from real faces.Therefore,detecting the face generated by GANs is a necessary work.This paper mainly introduces some methods to detect GAN-generated fake faces,and analyzes the advantages and disadvantages of these models based on the network structure and evaluation indexes,and the results obtained in the respective data sets.On this basis,the challenges faced in this field and future research directions are discussed.
基金This work is supported by National Natural Science Foundation of China(62072251).
文摘Due to the power of editing tools,new types of fake faces are being created and synthesized,which has attracted great attention on social media.It is reasonable to acknowledge that one human cannot distinguish whether the face is manipulated from the real faces.Therefore,the detection of face manipulation becomes a critical issue in digital media forensics.This paper provides an overview of recent deep learning detection models for face manipulation.Some public dataset used for face manipulation detection is introduced.On this basis,the challenges for the research and the potential future directions are analyzed and discussed.
文摘目的随着深度伪造技术的快速发展,人脸伪造图像越来越难以鉴别,对人们的日常生活和社会稳定造成了潜在的安全威胁。尽管当前很多方法在域内测试中取得了令人满意的性能表现,但在检测未知伪造类型时效果不佳。鉴于伪造人脸图像的伪造区域和非伪造区域具有不一致的源域特征,提出一种基于多级特征全局一致性的人脸深度伪造检测方法。方法使用人脸结构破除模块加强模型对局部细节和轻微异常信息的关注。采用多级特征融合模块使主干网络不同层级的特征进行交互学习,充分挖掘每个层级特征蕴含的伪造信息。使用全局一致性模块引导模型更好地提取伪造区域的特征表示,最终实现对人脸图像的精确分类。结果在两个数据集上进行实验。在域内实验中,本文方法的各项指标均优于目前先进的检测方法,在高质量和低质量FaceForensics++数据集上,AUC(area under the curve)分别达到99.02%和90.06%。在泛化实验中,本文的多项评价指标相比目前主流的伪造检测方法均占优。此外,消融实验进一步验证了模型的每个模块的有效性。结论本文方法可以较准确地对深度伪造人脸进行检测,具有优越的泛化性能,能够作为应对当前人脸伪造威胁的一种有效检测手段。