Deepfake detection of occluded images using a patch-based approach

被引:1
作者
Soleimani, Mahsa [1 ]
Nazari, Ali [1 ]
Moghaddam, Mohsen Ebrahimi [1 ]
机构
[1] Shahid Beheshti Univ, Fac Comp Sci & Engn, Tehran 1983969411, Iran
关键词
DeepFake; Deep learning; Generative adversarial networks;
D O I
10.1007/s00530-023-01140-8
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
DeepFake involves the use of deep learning and artificial intelligence techniques to produce or change video and image contents typically generated by GANs. Moreover, it can be misused and leads to fictitious news, ethical and financial crimes, and also affects the performance of facial recognition systems. Thus, detection of real or fake images is significant specially to authenticate originality of people's images or videos. One of the most important challenges in this topic is obstruction that decreases the system precision. In this study, we present a deep learning approach using the entire face and face patches to distinguish real/fake images in the presence of limitations of blurring, compression, scaling and especially obstruction with a three-path decision: first entire-face reasoning, second a decision based on the concatenation of feature vectors of face patches, and third a majority vote decision based on these features. To test our approach, new data sets including real and fake images are created. For producing fake images, StyleGAN and StyleGAN2 are trained by FFHQ images and also StarGAN and PGGAN are trained by CelebA images. The CelebA and FFHQ data sets are used as real images. The proposed approach reaches higher results in early epochs than other methods and increases the SoTA results by 0.4%-7.9% in the different built data sets. In addition, we have shown in experimental results that weighing the patches may improve accuracy.
引用
收藏
页码:2669 / 2687
页数:19
相关论文
共 50 条
[31]   Audio Deepfake Detection Using Deep Learning [J].
Shaaban, Ousama A. ;
Yildirim, Remzi .
ENGINEERING REPORTS, 2025, 7 (03)
[32]   Enhancing Deepfake Detection With Diversified Self-Blending Images and Residuals [J].
Liu, Qingtong ;
Xue, Ziyu ;
Liu, Haitao ;
Liu, Jing .
IEEE ACCESS, 2024, 12 :46109-46117
[33]   Security Strengthen and Detection of Deepfake Videos and Images Using Deep Learning Techniques [J].
Talreja, Sumran ;
Bindle, Abhay ;
Kumar, Vimal ;
Budhiraja, Ishan ;
Bhattacharya, Pronaya .
2024 IEEE INTERNATIONAL CONFERENCE ON COMMUNICATIONS WORKSHOPS, ICC WORKSHOPS 2024, 2024, :1834-1839
[34]   Patch-Based Adversarial Training for Error-Aware Circuit Annotation of Delayered IC Images [J].
Tee, Yee-Yang ;
Hong, Xuenong ;
Cheng, Deruo ;
Chee, Chye-Soon ;
Shi, Yiqiong ;
Lin, Tong ;
Gwee, Bah-Hwee .
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II-EXPRESS BRIEFS, 2023, 70 (09) :3694-3698
[35]   Deepfake detection using deep feature stacking and meta-learning [J].
Naskar, Gourab ;
Mohiuddin, Sk ;
Malakar, Samir ;
Cuevas, Erik ;
Sarkar, Ram .
HELIYON, 2024, 10 (04)
[36]   A new deep learning approach for the retinal hard exudates detection based on superpixel multi-feature extraction and patch-based CNN [J].
Huang, Chenxi ;
Zong, Yongshuo ;
Ding, Yimin ;
Luo, Xin ;
Clawson, Kathy ;
Peng, Yonghong .
NEUROCOMPUTING, 2021, 452 :521-533
[37]   Deep Patch-based Global Normal Orientation [J].
Wang, Shiyao ;
Liu, Xiuping ;
Liu, Jian ;
Li, Shuhua ;
Cao, Junjie .
COMPUTER-AIDED DESIGN, 2022, 150
[38]   Detection of Partially Occluded Area in Images Using Image Segmentation Technique [J].
Cherapanamjeri, Jyothsna ;
Rao, B. Narendra Kumar .
FOURTH CONGRESS ON INTELLIGENT SYSTEMS, VOL 1, CIS 2023, 2024, 868 :217-227
[39]   Data augmentation for patch-based OCT chorio-retinal segmentation using generative adversarial networks [J].
Kugelman, Jason ;
Alonso-Caneiro, David ;
Read, Scott A. ;
Vincent, Stephen J. ;
Chen, Fred K. ;
Collins, Michael J. .
NEURAL COMPUTING & APPLICATIONS, 2021, 33 (13) :7393-7408
[40]   Data augmentation for patch-based OCT chorio-retinal segmentation using generative adversarial networks [J].
Jason Kugelman ;
David Alonso-Caneiro ;
Scott A. Read ;
Stephen J. Vincent ;
Fred K. Chen ;
Michael J. Collins .
Neural Computing and Applications, 2021, 33 :7393-7408