Digital Image Source Identification Based On Sensor Pattern Noise

With the rapid popularity of the digital camera, and the extensive development ofnetwork technique, the digital image has been widely used in people’s daily work andlife, and at the same time, the proportion of the digital image in the networkinformation transfer is more and more big. When the progress of science bringsconvenient to our life, we must face up to an overwhelming challenge: illegal forgedmanipulation and spread of digital image. It will cause great social problems if theimages involving copyright dispute, image forged events, digital forensic and somecriminal cases, the judiciary be an urgent need for images tracking and identificationof sources.This paper will study the sensor pattern noise for digital image sourceidentification. To save storage space and transmission time, the digital image iscompressed, and JPEG compression is more commonly used to compress the image.JPEG compression can compress a digital image in a small storage space, certaininformation of the image will be lost, and so it will be more difficult to identify thedigital image source. After deeply analyzing the image forensics technology based onsensor pattern noise and JPEG encoding systems, we proposed a weighted correlationdetection algorithm. It richly applies quantization noise which appears in the processof JPEG encoding, to calculate the reliability factor for each pixel in sensor patternnoise, and import reliability factors to improve traditional correlation detectionalgorithm. This improved algorithm can further increase the precision rate of JPEGcompression images source identification. The experimental results verify that thealgorithm in this paper has stronger robustness and better stability for JPEGcompression.