A Curve Fitting Thresholding Approach for Forensic Source Identification of JPEG Compressed Images
Joybrata Sarkar, Ruchira Naskar · 2020
One of the major threats of today's digital era is detection of cyber-crime pertaining to safety and security of multimedia data. Given the huge volumes of digital data circulated online by every common man on a day-to-day basis, this threat becomes more prominent. The risk of online fraudulences is at its peak today. In relation to cyber-crime detection, source camera identification is a much-researched forensic problem, which is mapping a contentious image back to its source device. This is mainly done by exploiting the camera sensor specific artifacts left behind in an image. However, the high degrees of compression, brought about (in an image) by online networking sites, cause the statistical image properties to get destroyed along with the camera sensor specific artifacts. This inhibits the image-to-source mapping in a forensic analysis, and hugely impacts the source detection accuracy. In this work, we present a detailed analysis of traditional forensic source detection techniques and evaluate their performance in presence of JPEG compression. In this paper, we propose a curve fitting based adaptive thresholding approach for image to source mapping. The proposed approach outperforms the current state-of-the-art in terms of source detection accuracy of digital images.