Blind Cell-Phone Image Forensics Based On Multi-Features And Soft-Imaging

Human society has entered digital times, with the popularity of digital camera and cell-phone, ditigal images have become one of the major media for recording scene in daliy life. Meanwhile many image processing softwares being general, forging images become more and more easily. Although most of the tampering is not malicious, disputeses causing by tampering images happened everywhere in politics and economy fields, especially in newspaper. People worry about how to ensure the authenticity of digital media. And then the digital image forensics for image authenticity and integrality, in recent year, especially blind digital image forensics based on statistical properties becomes a hotsport. In this paper, we will research the blind digital image forensics technology, and main contributions are summarized as follows.Proposing a source cell-phone identification method based on multi-feature. Currently, there are fewer literatures about source cell-phone identification at home and abroad. Most of them are following source camera idengtification and do not consider the differences between camera and cell-phone. This will make a low identification rate, especially in the same brand between different models. To solve this problem, we analyze cell-phone imaging traits and extract effecitive features and remove the redundancy and interference among the features. Using the method, the average classification accuracy rate of 8 cell-phone iamges reaches to 95.75%, in particular for the same brand ones.Proposing a source identification and tampered detection algorithm for cell-phone images based on SVD features.Most of the source identification algorithm can only be effective on the whole image, but in this paper we can also be effective on image block through extracting SVD micro features.Proposing a tampering detection algorithm for cell-phone images using CFA soft-imaging technology. Since the false detection rate of the method based on image block can't satisfy, a method being able to locate the tampered region will be more direct and effective. So, we put forward a tampering detection algorithm based on CFA soft-imaging. We can detect various tampered images from different cell-phones, even from a same one with this method. Simultaneously, this method can resist JPEG re-compression attack, and also apply to images tampered by Adobe Photoshop tools. Proposing a model for detecting copy-past tampering based on CFA soft-imaging and FMT. Because traditional detecting method for copy-past tempering can't locate the real tempering region, but CFA soft-imaging can do it, we can combine them together. First, we bring forward a method based on FMT, which can resist rotation and JPEG re-compression attack. Then, combining with CFA soft-imaging, we advance the new model.