.png){kind=icon}
Return
Image forgery detection and localization using deep learning techniques
.png){kind=icon}
PRE
AI
DOI：10.1111/1556-4029.70351.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Nowadays, manipulating digital images has become very simple due to the availability of sophisticated image editing tools, making the authenticity verification of digital content increasingly challenging. The widespread use of such tools has raised serious concerns regarding the reliability of digital images in security, forensic, and judicial applications. To address these challenges, this paper proposes a unified deep learning-based framework for automatic image forgery detection and precise localization of manipulated regions in digital images. The forgery detection module employs a convolutional neural network (CNN) for hierarchical feature extraction, followed by a support vector machine (SVM) classifier to identify whether an image is pristine or tampered. For localization, the forged regions are segmented using a combination of VGG16 and U-Net architectures to achieve pixel-level manipulation detection. Experimental evaluations are conducted on the COMOFOD v2.0 and Columbia Image Splicing benchmark datasets. The performance of the proposed framework is evaluated using accuracy, loss, and AUC scores as performance metrics. The experimental results demonstrate that the proposed method achieves state-of-the-art performance for both forgery detection and localization tasks, outperforming several existing deep learning-based forensic approaches. The proposed framework provides an effective and reliable solution for detecting and localizing manipulated regions in digital images, thereby strengthening digital image forensic analysis and improving the trustworthiness of visual content in critical applications.
Keywords:
convolutional neural network
deep learning
image forgery detection
image forgery localization
image segmentation
Journal
J
IF:
1.8
Papers:
175
Citations:
0
Organization
Cited Papers
No cited papers available