An Evaluation of Convolutional Neural Network (CNN) Model for Copy-Move and Splicing Forgery Detection

Thiiban Muniappan, Thiiban Muniappan and Abd Warif, Nor Bakiah and Ismail, Ahsiah and Mat Abir, Noor Atikah (2023) An Evaluation of Convolutional Neural Network (CNN) Model for Copy-Move and Splicing Forgery Detection. INTELLIGENT SYSTEMS AND APPLICATIONS IN ENGINEERING, 11 (2). pp. 730-740. ISSN 2147-6799

[img] Text
J16045_03bd2e85e9c5ccb6bf013a56ac2dba37.pdf
Restricted to Registered users only

Download (597kB) | Request a copy

Abstract

Image forgeries such as copy-move and splicing are very common due to the availability of the advancement in software editing techniques. However, most of the existing methods for forgery detection consider only one type of image forgery due to the reason that both forgeries have different traits. In this paper, a Convolutional Neural Network (CNN) model which is one of the deep learning approaches is simulated and analyzed to detect any forged image without knowing their types of forgeries. In the model, three phases are involved: Data Preprocessing, Feature Extraction, and Classification. The model learns to extract features from convolutional, pooling, and Rectified Linear unit layer, and classified the image whether it is original or forged using fully connected layer. For the experimental works, three datasets namely MICC-F2000 (2000 images), CASIA 1 (1721 images), and CASIA 2 (12615 images) are tested and compared with existing deep learning-based methods. The results show that the CNN model achieved the highest performance with accuracy of 79% for CASIA 1 and 89% for CASIA 2

Item Type: Article
Uncontrolled Keywords: Convolutional Neural Network, Deep Learning, Image Forgery Detection
Subjects: T Technology > T Technology (General)
Divisions: Faculty of Computer Science and Information Technology > Department of Multimedia
Depositing User: Mr. Mohamad Zulkhibri Rahmad
Date Deposited: 25 Sep 2023 01:45
Last Modified: 25 Sep 2023 01:45
URI: http://eprints.uthm.edu.my/id/eprint/10004

Actions (login required)

View Item View Item