Blind Image Tampering Identification Based on Histogram Features

Nowadays, digital forensics has emerged as an important research field with applications of authenticity/integrality verification for digital data. In this paper, we focus on image forensic techniques and propose a blind scheme for image tampering identification which is capable to determine the tampering type. Since tampering operations bring in changes to neighboring pixels, we suggest employing the difference image (the image composed of differences of adjacent pixels) for forensic analysis. More specifically, we take the histogram of difference image to construct a feature set, and then use these histogram features to train a Support Vector Machine(SVM) classifier. In this way, the novel scheme can efficiently identify usual image operations such as scaling, JPEG compression, linear and nonlinear filtering. Besides, its superiority over other state-of-the-art work is also demonstrated experimentally.