Digital forensic investigators are faced with the tedious task of examining the content of
large datasets of photographic images to identify potential evidence. The analysis of the
results can be a lengthy process when the investigator is searching for specific objects in
the extracted media that can establish a fact or a logical link relevant to the investigation.
To help with this problem, computer vision applications can be used that try to automate
Image Analysis. The goal of this thesis is to evaluate whether a semi-automated data
carving process, in conjunction with an Automatic Image Annotation process, could help
make the digital forensics photo examination procedure more efficient, especially in
terms of time. For this purpose, a thorough examination of the relevant procedures that
could reinforce this idea, was performed, followed by the development of a methodology
that includes these processes. The developed methodology uses Photorec for the carving
of the evidence and Google’s Vision API for the extraction of the features that describe
the images. Experiments conducted on a set of forensic image files produced promising
results.
Collections
Show Collections
