Digital Image Forensics vs. Image Composition: An Indirect Arms Race
It addresses the problem of detecting malicious image alterations for forensic analysts, but is incremental as it primarily surveys and compares existing methods.
This paper analyzes the ongoing arms race between digital image forensics and image composition, proposing a novel scale to classify forensics assessments and testing composition techniques against different forensics traces, showing that forensics has basic tools to detect advanced composition despite unawareness.
The field of image composition is constantly trying to improve the ways in which an image can be altered and enhanced. While this is usually done in the name of aesthetics and practicality, it also provides tools that can be used to maliciously alter images. In this sense, the field of digital image forensics has to be prepared to deal with the influx of new technology, in a constant arms-race. In this paper, the current state of this arms-race is analyzed, surveying the state-of-the-art and providing means to compare both sides. A novel scale to classify image forensics assessments is proposed, and experiments are performed to test composition techniques in regards to different forensics traces. We show that even though research in forensics seems unaware of the advanced forms of image composition, it possesses the basic tools to detect it.