Research On Source Camera Identification And Analysis Of Attack And Counter-Attack

Digital media plays an increasingly important role in daily life, and often takes the formof digital evidence in court. However, powerful image editing software enables attackers toeasily make forgery, which causes people to doubt the reliability of digital images. Under thissituation, digital image forensics becomes an important technology to ensure the reliability ofdigital images. Digital image source identification is a very important branch of digital imageforensics.This paper researches the source camera identification algorithms based on camerasensor characteristics, and gives detailed analysis of fingerprint-copy attack andcounter-attack problem. Moreover，an improved algorithm based on enhanced pattern noise ispresented to improve performance. The contributions of this thesis can be described asfollows:1. A new idea to remove the CFA interpolation noise is proposed, which improves theperformance of camera source identification algorithm. By estimating the CFA model oftested camera, the sensor pattern noise based on original pixels can be extracted from theexisting pattern noise to achieve better correlation detection. The experimental results showthat the improved method has a better recognition performance.2. The key factors affecting counter-attack algorithms have been summarized and theoptimal attack combination is presented. If attackers use the optimal attack combination,higher probability of successful attacks can be achieved.3. The triangular test method has been simplified, and enhanced pattern noise extractionmethod has been used to deactivate counter-attack algorithm. When attackers use sufficientnumber of images with enhanced pattern noise extraction methods to acquire pattern noise,recognition performance of counter-attack algorithms drop significantly.