CNNs in Image Forensics: A Systematic Literature Review of Copy-Move, Splicing, Noise Detection, and Data Poisoning Detection Methods

Authors

Mohammed R. Subhi, Department of Petroleum Systems Control Engineering. College of Petroleum Process Engineering, 34001 Tikrit, Tikrit University, IraqFollow
Salman Yussof, Institute of Informatics and Computing in Energy, Universiti Tenaga Nasional, 43000 Kajang, Malaysia
Liyana Adilla binti Burhanuddin Adilla binti Burhanuddin, Institute of Informatics and Computing in Energy, Universiti Tenaga Nasional, 43000 Kajang, Malaysia
Firas Layth Khaleel, Department of Petroleum Systems Control Engineering. College of Petroleum Process Engineering, 34001 Tikrit, Tikrit University, Iraq
Firas Layth Khaleel, Department of Petroleum Systems Control Engineering. College of Petroleum Process Engineering, 34001 Tikrit, Tikrit University, Iraq

Document Type

Article

Keywords

Copy-move forgery detection (CMFD), Splicing images, Convolutional Neural Network, Data poisoning detection

Abstract

Image forgery, such as copy-move and splicing, poses significant challenges to the authenticity of digital images, and this challenge is exacerbated by the rapid development of image manipulation tools. Convolutional neural networks (CNNs) have shown promise in detecting such forgeries, but limitations remain, especially in detecting small duplicate regions and low-contrast regions, as well as in dealing with optical artefacts such as noise and blur. This systematic literature review examines CNN-based approaches to detect image forgery and explores strategies to mitigate data poisoning attacks, which can compromise the integrity of machine learning models. To our knowledge, there are currently no studies that comprehensively address the integration of forgery detection and splicing techniques with data poisoning detection. Our results reveal that while CNNs are effective in detecting manipulated images, challenges remain in dealing with complex manipulations and adversarial attacks. This review highlights the need for more robust detection methods and defence mechanisms against data poisoning, as current strategies are inadequate to address these issues comprehensively. We propose future research directions that focus on improving model generalizability, incorporating data poisoning defences, and enhancing the interpretability and flexibility of detection systems.

How to Cite This Article

Subhi, Mohammed R.; Yussof, Salman; Burhanuddin, Liyana Adilla binti Burhanuddin Adilla binti; Khaleel, Firas Layth; and Khaleel, Firas Layth (2025) "CNNs in Image Forensics: A Systematic Literature Review of Copy-Move, Splicing, Noise Detection, and Data Poisoning Detection Methods," Mesopotamian Journal of CyberSecurity: Vol. 5: Iss. 2, Article 19.
DOI: https://doi.org/10.58496/MJCS/2025/038
Available at: https://map.researchcommons.org/mjcs/vol5/iss2/19