CapsNet-Based Deep Learning Approach for Robust Image Forgery Detection

• Published in: International Conference on Signal Processing and Communication (Online) pp. 308 - 314
• Main Authors: Joshi, Deepak, Kashyap, Abhishek, Arora, Parul
• Format: Conference Proceeding
• Language: English
• Published: IEEE 20.02.2025
• Subjects: CapsNets; Convolutional neural networks; Convolutional Neural Networks (CNNs); Copy-move forgery; Deep learning; Feature extraction; Forgery; Image forensics; Journalism; Law enforcement; Manipulation detection; Noise; Social networking (online); Splicing
• ISSN: 2643-444X
• DOI: 10.1109/ICSC64553.2025.10968321

One common form of image tampering is image manipulation, where a section of an image is duplicated and pasted within the same image to conceal or replicate content. Detecting such manipulations is essential in digital image forensics, especially in critical domains like journalism, law enforcement, and social media. Traditional image manipulation detection methods struggle with identifying tampered regions due to sophisticated forgery techniques and post-processing like scaling, rotation, and noise addition. In this paper, we explore the application of Capsule Networks (CapsNets) in detecting copy-move or splicing image manipulation. CapsNets offer an advantage in preserving spatial hierarchies between features and improving detection performance under transformations. Experimental results demonstrate that CapsNets outperform conventional Convolutional Neural Networks (CNNs) in detecting copy-move or splicing forgery under various conditions, including scaling, rotation, and added noise.