An automatic enhanced flters with frequency‑based copy‑move forgery detection for social media images
Using images in various fle types has become common in the modern digital world, such as social media posts, research reports, and legal documents. The availability of low-cost image manipulation tools has made it easier to change images, potentially leading to undetected image fraud. One such form...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2024
|
| Subjects: | |
| Online Access: | http://eprints.uthm.edu.my/10869/1/J16830_ace79442b2806b9e1a87ff70f440aaad.pdf http://eprints.uthm.edu.my/10869/ https://doi.org/10.1007/s11042-023-15506-7 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Using images in various fle types has become common in the modern digital world, such as social media posts, research reports, and legal documents. The availability of low-cost
image manipulation tools has made it easier to change images, potentially leading to undetected image fraud. One such form of image manipulation is copy-move forgery (CMF),
which is difcult to detect due to the similarities in image features. There have been eforts to detect CMF using copy-move forgery detection (CMFD) methods. However, most
research has focused on CMF images with attacks rather than social media. Social media has contributed signifcantly to the image manipulation phenomenon, and additional postprocessing techniques on social media platforms have afected the efciency of the CMFD methods. Therefore, this research proposes a two-stage pre-processing phase combined with frequency-based CMFD to detect CMF images in diferent social media platforms. The frst stage includes automatic image selection, followed by image enhancement with flters to improve image quality. The experimental results show that the proposed method
achieves the highest detection score compared to existing CMFD methods, with an average score of 90% for CMF-Facebook, 91% for CMF-WhatsApp, and 85% for CMF-Twitter.
This research highlights the importance of developing solutions to detect image forgery in social media and the potential of combining pre-processing with frequency-based methods to improve results. |
|---|