Image steganography and steganalysis

In this dissertation we investigate a number of open issues in image steganography and steganalysis. Most of the related work in the field is either proposing new embedding techniques or steganalyzing previously proposed embedding techniques. Thus, throughout the dissertation it has been our goal to take a more macro look at the field and investigate practical aspects of steganography which have not been studied in depth previously. The dissertation begins by reviewing general concepts of steganography, and related terminology. We then discuss the theoretical work done on steganographical security and capacity. We continue our review, by presenting a taxonomy of images based steganography embedding as well as steganalysis techniques.; We continue by benchmarking the performance of the state-of-the-art embedding and steganalysis techniques. The goal of our benchmark is to understand the behavior of such techniques when confronted with large, and heterogenous dataset of images. Next we point out that with the large number of steganalysis techniques available, the steganalyst will have to select one or more techniques which she will employ on a set of suspected stego images. But what should be done when the results produces are contradictory? We propose and investigate information fusion techniques, that combine a number of steganalysis techniques. We illustrate, through case studies, how one is able to obtain performance improvements as well as scalability by employing suitable fusion techniques.; We conclude the dissertation by making note of the fact that the embedder is free to choose the cover image when embedding a message. Therefore we investigate the problem of selecting covers which produce less detectable stego images. We show how a few simple measures could help the embedder increase his or her chances of going undetected. We extent the results obtained to the cover selection problem, by defining and investigating a new practical approach to evaluate the detectability and capacity of stego images. Our approach is purely practical and with it one is able to obtain the capacity rate for a specific image, given a specific embedding and steganalysis techniques. We also discuss how our practical approach compares to previously done theoretical work.