Digital Image Forensic Algorithms Based On Imaging Sensor Features

The rapid increase of imaging device and the popularity of image editing software bring tremendous convenience to people's lives, meanwhile, create lots of new information security challenge. Digital image forensics has been considered as one of the key technology to solve these problems. This technology now has been gradually recognized and applied to source camera identification, forgery detection and integrity verification. Therefore, the research of this technology has important significance for ensuring social trust, crime-controlling and defending judicial justice.This dissertation mainly studies digital image forensics methods based on imaging sensor features for source camera and scanner identification. We first analyze the advantages and disadvantages of representative methods for source device identification, and then propose our algorithms and solutions. The main work can be summarized as follows:1. Based on imaging sensor pattern noise, the classical source camera identification algorithm proposed by Lukas is simulated, and then its deficiencies are reviewed. We emphatically analyze the source camera identification technique using large components of imaging sensor pattern noise proposed by us before, the influence of the de-noising filter and the de-CFA interpolation operation that can improve the quality of pattern noise is discussed. Afterward, we focus on discussing why and how the use of large components of the sensor pattern noise can improve the performance of the correlation-based detector, and then point out the advantages and disadvantages.2. Three new schemes of global information from color images for source camera identification are presented. Current methods only utilize forensic information derived from one single-channel image, and ignore information between the color channels. We investigate the characteristics of imaging process of a camera sensor, and then extend our previous camera identification method for single-channel images to color images. Our new methods utilize information from all of the three channel images and hence can reflect more characteristics of the imaging sensor. 3. Two methods for source scanner identification based on imaging sensor features are proposed. We analyze the structure of scanner and its imaging-forming principle, then two algorithms of array pattern and line pattern are proposed for identifying the source scanner. The simulation results show the validity of the proposed algorithm.