A dual-stream model based on PRNU and quaternion RGB for detecting fake faces.
The forensic examination of AIGC(Artificial Intelligence Generated Content) faces poses a contemporary challenge within the realm of color image forensics. A myriad of artificially generated faces by AIGC encompasses both global and local manipulations. While there has been noteworthy progress in th...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Public Library of Science (PLoS)
2025-01-01
|
| Series: | PLoS ONE |
| Online Access: | https://doi.org/10.1371/journal.pone.0314041 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832540176876830720 |
|---|---|
| author | Mohan Hua Shuangliang Li Jinwei Wang |
| author_facet | Mohan Hua Shuangliang Li Jinwei Wang |
| author_sort | Mohan Hua |
| collection | DOAJ |
| description | The forensic examination of AIGC(Artificial Intelligence Generated Content) faces poses a contemporary challenge within the realm of color image forensics. A myriad of artificially generated faces by AIGC encompasses both global and local manipulations. While there has been noteworthy progress in the forensic scrutiny of fake faces, current research primarily focuses on the isolated detection of globally and locally manipulated fake faces, thus lacking a universally effective detection methodology. To address this limitation, we propose a sophisticated forensic model that incorporates a dual-stream framework comprising quaternion RGB and PRNU(Photo Response Non-Uniformity). The PRNU stream extracts the "camera fingerprint" feature by discerning the non-uniform response of the image sensor under varying lighting conditions, thereby encapsulating the overall distribution characteristics of globally manipulated faces. The quaternion RGB stream leverages the inherent nonlinear properties of quaternions and their informative representation capabilities to accurately describe changes in image color, background, and spatial structure, facilitating the meticulous capture of nuanced local distinctions between locally manipulated faces and real faces. Ultimately, we integrate the two streams to establish the exchange of feature information between PRNU and quaternion RGB streams. This strategic integration fully exploits the complementarity between two streams to amalgamate local and global features effectively. Experimental results obtained from diverse datasets underscore the advantages of our method in terms of accuracy, achieving a detection accuracy of 96.81%. |
| format | Article |
| id | doaj-art-4c9e88194c894822b041b45bec494143 |
| institution | Kabale University |
| issn | 1932-6203 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Public Library of Science (PLoS) |
| record_format | Article |
| series | PLoS ONE |
| spelling | doaj-art-4c9e88194c894822b041b45bec4941432025-02-05T05:31:59ZengPublic Library of Science (PLoS)PLoS ONE1932-62032025-01-01201e031404110.1371/journal.pone.0314041A dual-stream model based on PRNU and quaternion RGB for detecting fake faces.Mohan HuaShuangliang LiJinwei WangThe forensic examination of AIGC(Artificial Intelligence Generated Content) faces poses a contemporary challenge within the realm of color image forensics. A myriad of artificially generated faces by AIGC encompasses both global and local manipulations. While there has been noteworthy progress in the forensic scrutiny of fake faces, current research primarily focuses on the isolated detection of globally and locally manipulated fake faces, thus lacking a universally effective detection methodology. To address this limitation, we propose a sophisticated forensic model that incorporates a dual-stream framework comprising quaternion RGB and PRNU(Photo Response Non-Uniformity). The PRNU stream extracts the "camera fingerprint" feature by discerning the non-uniform response of the image sensor under varying lighting conditions, thereby encapsulating the overall distribution characteristics of globally manipulated faces. The quaternion RGB stream leverages the inherent nonlinear properties of quaternions and their informative representation capabilities to accurately describe changes in image color, background, and spatial structure, facilitating the meticulous capture of nuanced local distinctions between locally manipulated faces and real faces. Ultimately, we integrate the two streams to establish the exchange of feature information between PRNU and quaternion RGB streams. This strategic integration fully exploits the complementarity between two streams to amalgamate local and global features effectively. Experimental results obtained from diverse datasets underscore the advantages of our method in terms of accuracy, achieving a detection accuracy of 96.81%.https://doi.org/10.1371/journal.pone.0314041 |
| spellingShingle | Mohan Hua Shuangliang Li Jinwei Wang A dual-stream model based on PRNU and quaternion RGB for detecting fake faces. PLoS ONE |
| title | A dual-stream model based on PRNU and quaternion RGB for detecting fake faces. |
| title_full | A dual-stream model based on PRNU and quaternion RGB for detecting fake faces. |
| title_fullStr | A dual-stream model based on PRNU and quaternion RGB for detecting fake faces. |
| title_full_unstemmed | A dual-stream model based on PRNU and quaternion RGB for detecting fake faces. |
| title_short | A dual-stream model based on PRNU and quaternion RGB for detecting fake faces. |
| title_sort | dual stream model based on prnu and quaternion rgb for detecting fake faces |
| url | https://doi.org/10.1371/journal.pone.0314041 |
| work_keys_str_mv | AT mohanhua adualstreammodelbasedonprnuandquaternionrgbfordetectingfakefaces AT shuangliangli adualstreammodelbasedonprnuandquaternionrgbfordetectingfakefaces AT jinweiwang adualstreammodelbasedonprnuandquaternionrgbfordetectingfakefaces AT mohanhua dualstreammodelbasedonprnuandquaternionrgbfordetectingfakefaces AT shuangliangli dualstreammodelbasedonprnuandquaternionrgbfordetectingfakefaces AT jinweiwang dualstreammodelbasedonprnuandquaternionrgbfordetectingfakefaces |