
Research Article
Image Copy-Move Forgery Detection in the Social Media Based on a Prior Density Clustering and the Point Density
@INPROCEEDINGS{10.1007/978-3-031-73699-5_17, author={Cong Lin and Hai Yang and Ke Huang and Yufeng Wu and Yamin Wen and Yuqiao Deng}, title={Image Copy-Move Forgery Detection in the Social Media Based on a Prior Density Clustering and the Point Density}, proceedings={Security and Privacy in New Computing Environments. 6th International Conference, SPNCE 2023, Guangzhou, China, November 25--26, 2023, Proceedings}, proceedings_a={SPNCE}, year={2025}, month={1}, keywords={Multimedia forensics Image forensics Copy-move forgery Density clustering Point density}, doi={10.1007/978-3-031-73699-5_17} }
- Cong Lin
Hai Yang
Ke Huang
Yufeng Wu
Yamin Wen
Yuqiao Deng
Year: 2025
Image Copy-Move Forgery Detection in the Social Media Based on a Prior Density Clustering and the Point Density
SPNCE
Springer
DOI: 10.1007/978-3-031-73699-5_17
Abstract
Copy-move forgery is one common image manipulation technique. Many images are compressed by the social media, but most of the existing copy-move forgery detection schemes are proposed to deal with the uncompressed version of images. To handle this problem, in this paper, a copy-move forgery detection scheme for social media is proposed based on the point density and a prior density clustering. Firstly, the concept of “point density” is proposed, and it is combined with feature extraction to improve the extraction effect. Secondly, the hierarchical matching is adopted, and the keypoints are grouped according to the pixel value. Thirdly, a prior-based density clustering is proposed, called prior-DBSCAN. In this scheme, the matching pairs are divided into start points and end points for clustering, respectively, and the prior region of each cluster is obtained. Then, the clustering with the new cluster radius is performed. Finally, an iterative localization technique is used to obtain the final localization results. Considering the compression of images during their transmission through the social media, the proposed scheme is made more suitable for real-world scenarios. The experimental results demonstrate that the proposed scheme based on a prior density clustering and the point density, which is better and more robust than the state-of-the-art schemes on publicly available datasets.