Identification of image fragments for file carving

Recovering images intact is an important process in digital forensics, as they may represent primary evidences in crime cases such as child pornography. Due to file syetems´ fragmentation mechanisms, images may be split into several fragments on a physical storage. As such, recovering images fragments and reconstructing the original images embody challenges for file carving tools particularly when the filesystem metadata are not available. In this paper, we propose a method for image fragment identification using a machine learning approach. Our method exploits features in unknown images fragments, and applies various machine learning algorithms to reconstruct the original images by identifying to which particular image a fragment belongs. We provide the details of our methods as well as a validation of its effectiveness.