Digital image forensics is a field encompassing camera identication, forgery detection and steganalysis. Statistical modeling and machine learning have been successfully applied in the academic community of this maturing field. Still, large gaps exist between academic results and applications used by practicing forensic analysts, especially when the target samples are drawn from a different population than the data in a reference database.
This thesis contains four published papers aiming at narrowing this gap in three different fields: mobile stego app detection, digital image steganalysis and camera identification. It is the first work to explore a way of extending the academic methods to real world images created by apps. New ideas and methods are developed for target images with very rich flexibility in the embedding rates, embedding algorithms, exposure settings and camera sources. The experimental results proved that the proposed methods work very well, even for the devices which are not included in the reference database.