ID3: Identity-Driven Learning Based on 3D Reconstruction and Frame-Level Residual Enhancement for Deepfakes Detection

IF 1.8 4区 计算机科学 Q3 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE
Hui Ma, Jin Zhang, Xihong Chen, Wenhao Chu, Jiabao Guo, Junze Zheng, Liying Yang, Yanyan Liang
{"title":"ID3: Identity-Driven Learning Based on 3D Reconstruction and Frame-Level Residual Enhancement for Deepfakes Detection","authors":"Hui Ma,&nbsp;Jin Zhang,&nbsp;Xihong Chen,&nbsp;Wenhao Chu,&nbsp;Jiabao Guo,&nbsp;Junze Zheng,&nbsp;Liying Yang,&nbsp;Yanyan Liang","doi":"10.1049/bme2/3764746","DOIUrl":null,"url":null,"abstract":"<p>With the rapid advancement of face manipulation technology, various forged videos of celebrities and politicians have appeared and cause pernicious social impact. In this light, forge video detection becomes a research hot spot recently. Most previous detection approaches focus mainly on forgery artifacts caused by the specific generation defects without considering the individual identity information, so that the detection accuracy is not satisfactory. For instance, for a forgery video of a certain celebrity, everyone knows who she/he is, while this important identity clue is not utilized in current detection methods. To address this problem, a novel perspective of face forgery detection via identity-driven learning, named Identity-Driven Deepfakes Detection (ID<sup>3</sup>), is proposed. By the proposed method, the similarity between suspect inputs and the inherent properties (e.g., geometry and appearance) of the same identity is considered and explored. Specifically, by 3D reconstruction, the physical differences between the forged and real videos are captured in the learning process. In addition, with frame level residual enhancement, the detection accuracy can be further improved. The validity of the proposed method is experimentally verified on several benchmark datasets, and our detection performance is better than some state-of-the-art works.</p>","PeriodicalId":48821,"journal":{"name":"IET Biometrics","volume":"2025 1","pages":""},"PeriodicalIF":1.8000,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ietresearch.onlinelibrary.wiley.com/doi/epdf/10.1049/bme2/3764746","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IET Biometrics","FirstCategoryId":"94","ListUrlMain":"https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/bme2/3764746","RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 0

Abstract

With the rapid advancement of face manipulation technology, various forged videos of celebrities and politicians have appeared and cause pernicious social impact. In this light, forge video detection becomes a research hot spot recently. Most previous detection approaches focus mainly on forgery artifacts caused by the specific generation defects without considering the individual identity information, so that the detection accuracy is not satisfactory. For instance, for a forgery video of a certain celebrity, everyone knows who she/he is, while this important identity clue is not utilized in current detection methods. To address this problem, a novel perspective of face forgery detection via identity-driven learning, named Identity-Driven Deepfakes Detection (ID3), is proposed. By the proposed method, the similarity between suspect inputs and the inherent properties (e.g., geometry and appearance) of the same identity is considered and explored. Specifically, by 3D reconstruction, the physical differences between the forged and real videos are captured in the learning process. In addition, with frame level residual enhancement, the detection accuracy can be further improved. The validity of the proposed method is experimentally verified on several benchmark datasets, and our detection performance is better than some state-of-the-art works.

Abstract Image

Abstract Image

Abstract Image

Abstract Image

基于三维重建和帧级残差增强的身份驱动学习深度伪造检测
随着人脸处理技术的飞速发展,出现了各种伪造名人、政客的视频,并造成了恶劣的社会影响。有鉴于此,伪造视频检测成为近年来的研究热点。以往的检测方法大多只关注由于特定的生成缺陷而产生的伪造伪迹,没有考虑到个体的身份信息,导致检测精度不理想。例如,对于某个名人的伪造视频,每个人都知道她/他是谁,而这一重要的身份线索在目前的检测方法中没有被利用。为了解决这个问题,提出了一种通过身份驱动学习进行人脸伪造检测的新视角,称为身份驱动深度伪造检测(ID3)。通过提出的方法,考虑并探索了可疑输入与同一身份的固有属性(例如几何和外观)之间的相似性。具体来说,通过3D重建,在学习过程中捕获了伪造和真实视频之间的物理差异。此外,通过帧级残差增强,可以进一步提高检测精度。在多个基准数据集上进行了实验,验证了该方法的有效性,检测性能优于目前的一些研究成果。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
IET Biometrics
IET Biometrics COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE-
CiteScore
5.90
自引率
0.00%
发文量
46
审稿时长
33 weeks
期刊介绍: The field of biometric recognition - automated recognition of individuals based on their behavioural and biological characteristics - has now reached a level of maturity where viable practical applications are both possible and increasingly available. The biometrics field is characterised especially by its interdisciplinarity since, while focused primarily around a strong technological base, effective system design and implementation often requires a broad range of skills encompassing, for example, human factors, data security and database technologies, psychological and physiological awareness, and so on. Also, the technology focus itself embraces diversity, since the engineering of effective biometric systems requires integration of image analysis, pattern recognition, sensor technology, database engineering, security design and many other strands of understanding. The scope of the journal is intentionally relatively wide. While focusing on core technological issues, it is recognised that these may be inherently diverse and in many cases may cross traditional disciplinary boundaries. The scope of the journal will therefore include any topics where it can be shown that a paper can increase our understanding of biometric systems, signal future developments and applications for biometrics, or promote greater practical uptake for relevant technologies: Development and enhancement of individual biometric modalities including the established and traditional modalities (e.g. face, fingerprint, iris, signature and handwriting recognition) and also newer or emerging modalities (gait, ear-shape, neurological patterns, etc.) Multibiometrics, theoretical and practical issues, implementation of practical systems, multiclassifier and multimodal approaches Soft biometrics and information fusion for identification, verification and trait prediction Human factors and the human-computer interface issues for biometric systems, exception handling strategies Template construction and template management, ageing factors and their impact on biometric systems Usability and user-oriented design, psychological and physiological principles and system integration Sensors and sensor technologies for biometric processing Database technologies to support biometric systems Implementation of biometric systems, security engineering implications, smartcard and associated technologies in implementation, implementation platforms, system design and performance evaluation Trust and privacy issues, security of biometric systems and supporting technological solutions, biometric template protection Biometric cryptosystems, security and biometrics-linked encryption Links with forensic processing and cross-disciplinary commonalities Core underpinning technologies (e.g. image analysis, pattern recognition, computer vision, signal processing, etc.), where the specific relevance to biometric processing can be demonstrated Applications and application-led considerations Position papers on technology or on the industrial context of biometric system development Adoption and promotion of standards in biometrics, improving technology acceptance, deployment and interoperability, avoiding cross-cultural and cross-sector restrictions Relevant ethical and social issues
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信