Counter Forensics: A New PRNU Based Method for Image Source Anonymization

Abstract

Digital Image Forensics deals with the authenticity or originality and/or the detection of image forgery. Even though there are different forensic techniques for the analysis of digital images, hackers are advancing their methods to manipulate images. There arises the need of counter forensic techniques to resist the robust attacks. The Photo Response Non Uniformity (PRNU) of the source camera can be considered as the unique identity of the device, just like the fingerprint of a human being. This paper introduces a new counter forensic method to anonymize the identity of the image from attackers. The PRNU (fingerprint) of the image is removed and the variance of the PRNU of another camera is introduced to anonymize the identity. Also the image quality of the forged image is analyzed using Image Quality Metrics (IQM) and PSNR (Peak Signal to Noise Ratio) values. The results show that even after source anonymization, the image has better visual quality with average Mean Absolute Error (MAE) and Mean Square Error (MSE) values of 0.1278 and 0.1111 respectively. The noise residual of the forged image is correlated with the PRNU of the original acquisition device, and verified that they are uncorrelated as the values ranging from -0.0025 to 0.0117. This indicates that the identity of the image is forged or the forged image has little traces of the original acquisition device. The method was tested with images of the same scene captured by five different cameras.