Detecting Information Hidden in JPEG Images: An investigation of machine learning methods for image steganalysis

This book describes using artificial neural networks as a potential approach for the steganalysis of JPEG images. The features that would be used by the detection algorithm have been previously suggested in various literature. These have given the best published results. The features are discussed in detail in the book. However, most earlier approaches have used SVMs for the detection process, while the proposed approach is based on using neural networks instead. Several experiments were conducted while researching for this book to see if the number of features required for adequate detection accuracy can be reduced without impacting the accuracy. The results from reduced feature sets are benchmarked according to the detection accuracy that was achieved as well as the time taken for the extraction of those particular features. A point to note is that the discusses the problem of detecting messages hidden in JPEG images, rather than actually recovering the hidden messages.