The impact of exposure settings in the digital image forensics

The digital image forensics academic community is facing a growing challenge. The volume of images presented to digital image forensic practitioners increases every day, and with it, more variety of possible outcomes in image analysis. When an academic forensic tool is applied to a realistic case, the effects of imaging factors, including the noise level, should be seriously considered. Although there have been conjectures that shot noise would affect the empirical accuracy of a forensic analyzer, it has not yet received enough experimental support. In this paper, instead of estimating the noise, we inspect two measurable factors of exposure settings, ISO speed and exposure time, and present a set of experiments using mobile phone data to demonstrate the effect of exposure settings on steganalysis and PRNU-based camera device identification. Our results show that more investigations into the characteristics of image data with respect to exposure settings is required to fully understand identification or classification that is concerned with low-magnitude noise measurements, such as PRNU or stegoembedded messages.