Research On Passive Blind Digital Inage Forensics

Blind digital media forensics refers to the techniques that passively identify the authenticity, integrity and origination of the digital media without the help of embedded information. Digital media includes digital images, videos, audios and other digital signals. The dissertation focuses on the passive blind image forensics, which is an interdisciplinary frontier research field that is still in its infancy. Passive blind image forensics is a chanllenging new hotspot of multimedia information security with broad prospect.Although tampered images often leave no visual clues of forgery, the tampering operations leave special forgery traces. We propose three techniques that quantify and detect traces left by three kinds of tampering operations. The main work and contributions of the dissertation are outlined as follows:1. The main branches and existing techniques of blind digital media forensics are summarized. Then the robustness and security of the passive blind forensics are discussed.2. The existing Copy-move forgery is generalized to Copy-transform-move forgery and a detection method based on SIFT points is proposed. The matching SIFT points are found to locate the forged regions initially. Then the forged regions grow up stepwise based on an appropriate region growing strategy with matching SIFT points as"seed-points". To deal with down sampled images for detection, a method is proposed to fuse the multi-channel image information into a single gray-level image and the fused image is used for detection. This method can also reduce the computation load by downsamling high resolution images beforehand. With the help of the previous methods, a detection method for Photoshop Healing Brush and Patch Tool is proposed. Experimental results of doubtful news pictures show the effectiveness of these methods.3. The image forgeries which use JPEG images as the background images are analyzed and a novel technique is proposed to automatically locate the tampered regions. The proposed method is based on inconsistency of blocking artifacts and the extracted image noise is used to measure the JPEG blocking artifact to improve the SNR. The idea that improves the SNR using image noise is also used to estimate the JPEG grid location and the size of minimum coded unit. Experimental results show that the idea achieves better results than existing methods.4. Based on the fact that image patches with similar depth have similar blur kernel sizes, a detection technique based on the inconsistency of defocus blur is proposed. The proposed technique chooses different image patches with similar depth and uses Elder-Zucker method to estimate the defocus blur. The inconsistency of blur exposes the forgery. Experimental results of tampered images from law cases show the effectiveness of this technique.