基于噪声注意力的伪造人脸检测方法  

Noise-attention-based forgery face detection method

作　　者：张博林 朱春陶 殷琪林 付婧巧 刘凌毅 刘佳睿 刘红梅[1,2,3] 卢伟[1,2,3] ZHANG Bolin;ZHU Chuntao;YIN Qilin;FU Jingqiao;LIU Lingyi;LIU Jiarui;LIU Hongmei;LU Wei(School of Computer Science and Engineering,Sun Yat-sen University,GuangZhou 510006,China;Guangdong Province Key Laboratory of Information Security Technology,GuangZhou 510006,China;Ministry of Education Key Laboratory of Machine Intelligence and Advanced Computing,GuangZhou 510006,China)

机构地区：[1]中山大学计算机学院,广东广州510006 [2]广东省信息安全技术重点实验室,广东广州510006 [3]机器智能与先进计算教育部重点实验室,广东广州510006

出　　处：《网络与信息安全学报》2023年第4期155-165,共11页Chinese Journal of Network and Information Security

基　　金：国家自然科学基金(U2001202,62072480)。

摘　　要：随着人工智能和深度神经网络的不断发展,图像生成与编辑变得越来越容易,恶意运用图像生成工具进行篡改伪造的现象层出不穷,这对多媒体安全以及社会稳定造成了极大威胁,因此研究伪造人脸的检测方法至关重要。人脸篡改伪造的方式和工具多种多样,在篡改的过程中可能留下不同程度的篡改痕迹,而这在图像噪声中都有一定程度上的反映。从图像噪声的角度出发,通过噪声去除的方式挖掘反映伪造人脸篡改痕迹的噪声成分,进一步生成噪声注意力,指导主干网络进行伪造人脸检测。使用SRM滤波监督噪声去除模块的训练,并将噪声去除模块所得到的噪声再次加入真实人脸图像中,形成一对有监督的训练样本,通过自监督的方式对噪声去除模块进行加强指导,实验结果说明噪声去除模块得到的噪声特征具有较好的区分度。在多个公开数据集上进行了实验,所提方法在Celeb-DF数据集上达到98.32%的准确率,在FaceForensics++数据集上达到94%以上的准确率,在DFDC数据集上达到92.61%的准确率,证明了所提方法的有效性。With the advancement of artificial intelligence and deep neural networks,the ease of image generation and editing has increased significantly.Consequently,the occurrence of malicious tampering and forgery using image generation tools is on the rise,posing a significant threat to multimedia security and social stability.Therefore, it is crucial to research detection methods for forged faces. Face tampering and forgery can occurthrough various means and tools, leaving different levels of forgery traces during the tampering process. Thesetraces can be partly reflected in the image noise. From the perspective of image noise, the noise componentsreflecting tampering traces of forged faces were extracted through a noise removal module. Furthermore, noiseattention was generated to guide the backbone network in the detection of forged faces. The training of the noiseremoval module was supervised using SRM filters. In order to strengthen the guidance of the noise removalmodule, the noise obtained by the noise removal module was added back to the real face image, forming a pair ofsupervised training samples in a self-supervised manner. The experimental results illustrate that the noise featuresobtained by the noise removal module have a good degree of discrimination. Experiments were also conducted onseveral public datasets, and the proposed method achieves an accuracy of 98.32% on the Celeb-DF dataset,92.61% on the DFDC dataset, and more than 94% on the FaceForensics++ dataset, thus proving the effectivenessof the proposed method.

关 键 词：Deepfake检测 图像噪声 注意力机制 篡改痕迹 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]