Studies On Key Problems Of Source Cameraldentification Using Sensor Pattern Noise

In recent years, with the wide application of all kinds of digital image acquisitionequipment and increasingly powerful image processing software, digital image can be easilyachieved and edited, this may lead to serious negative impact on the political, military,security, culture, news report, judicial trial and scientific research. How to identify theimaging equipment of the digital image is one of the important research areas of digital imageforensic technology, and can provide technical support for the protection of digital copyright,tracking the source of bad digital images.At present, imaging device identification technology is mainly realized by usingstatistical properties of image, characteristics of imaging device and artifacts of the imagingprocess. This paper focuses on characteristics of imaging device to discuss the problem ofimaging device identification technology. Based on the analysis of the typical camera sourceidentification algorithm used the correlation of pattern noise, some corresponding solutionsare proposed for some critical problems in the algorithm. The main work can be summarizedas follows:1．We simulate the algorithm proposed by Fridrich at al., and point out two questions forthe rest of the work: the identification accuracy is not high for images of small size andpattern noise is vulnerable to potential attack. Later the influence of different pattern noiseextraction method and different reference image set on the performance of the algorithm wasdiscussed.2．For the first problem pointed out above, the algorithm using the mixed correlaionbetween the phase-only component of noise residues and the reference pattern noise wasimproved, this can greatly improve the accuracy for images of small size.3．The typical attack which algorithm may be suffered is analyzed, and the correspondinganti-attack strategies are proposed.