Adaptive feature selection for digital camera source identification

Digital forensics which identifies the characteristics and the originality of the digital devices has lately become one of the very important applications. If it wants to serve as the evidence to the court like its traditional counterparts, the digital forensics issues like verifying the authenticity of digital image, detecting the forged regions or identifying digital camera source of an image will need to be addressed. This study has focused on analyzing the relationship between digital cameras and the photographs by using the support vector machine (SVM). In this paper, several feature selection algorithms will be implemented with SVM-based classifier to increase the predicting accuracy rate of identifying digital camera source of an image. The experiment results demonstrate that our adaptive feature selection scheme can achieve higher identification rate for unbiased camera sources than the results without using feature selection approaches.