Camera model identification using a deep network and a reduced edge dataset
- PDF / 637,195 Bytes
- 8 Pages / 595.276 x 790.866 pts Page_size
- 100 Downloads / 176 Views
(0123456789().,-volV)(0123456789(). ,- volV)
GREEN AND HUMAN INFORMATION TECHNOLOGY 2019
Camera model identification using a deep network and a reduced edge dataset Changhee Kang1 • Sang-ug Kang2 Received: 6 May 2019 / Accepted: 22 November 2019 Ó Springer-Verlag London Ltd., part of Springer Nature 2019
Abstract Today, the importance of digital images as a medium for social communication is growing rapidly. Sometimes, an image needs to be authenticated by verifying its source camera model or device. Recently, deep networks have become very successful at visual pattern recognition. With this motivation, several investigators have explored the possibility of using convolutional neural networks (CNNs) for camera source identification. In this paper, we use selective preprocessing, instead of a indiscriminate one, in order not to hinder the CNN’s strong ability to learn useful features for this kind of forensic task. To generate a consistent and balanced dataset, we limit the maximum number of original images to 200 per camera model, and we discard vertically taken images. Using a relatively simple deep network structure, the proposed method achieved a better prediction accuracy—95.0%—than GoogleNet and other existing methods. Also, challenging camera models such as the Sony DSC H50 and W170 can be classified with the quite high prediction accuracies of 87.9% and 83.1%, respectively. Keywords Camera source identification Deep learning Balanced dataset Edge detection
1 Introduction Digital images are increasingly being used as evidence to prove crimes and to help make judgments in court. A digital image is a type of data like a document, a file, or even physical evidence. Such data must be authenticated in order to be used for forensic purposes. Two important aspects of data authentication are (1) to verify that the data have not been altered and (2) that the data have actually come from their alleged sources. Accordingly, methods for detecting digital image modification have been the subject of active research, because images can easily be counterfeited or tampered with. On the other hand, source identification techniques have also been studied in recent years & Sang-ug Kang [email protected] Changhee Kang [email protected] 1
Department of Computer Science, Sangmyung University, 20 Hongjimun2-Gil, Jongno-Gu, Seoul, Korea
2
Faculty of Computer Science, Sangmyung University, 20 Hongjimun2-Gil, Jongno-Gu, Seoul, Korea
because of the social need to identify copyright holders of photographic artwork and because of the increasing need to control photographs taken by hidden cameras. While camera model identification techniques cannot completely resolve the source identification aspect of authentication, they can at least narrow the scope of source investigation. For example, if someone states that he did not take this photograph with his own camera, then he can fully prove that repudiation when the photograph is actually taken with a camera of different model. Model identification is considered as a first s
