Using an Ariticial Neural Network to Detect the Presence of Image Steganography
Aron Chandrababu · OhioLink ETD Center (Ohio Library and Information Network) · 2009
The purpose of steganography is to hide the presence of a message.Modern day techniques embed pictures and text inside computer files.Steganalysis is a field devoted to detecting steganography in files and possibly extracting the hidden image or text.This thesis introduces a new idea for steganalysis, that of training an artificial neural network to identify images that have another image embedded in them.Two different types of artificial neural networks, a standard and a shortcut type, are trained for two different types of data sets.One data set contains images with and without hidden images embedded in them.The other data set is derived from calculating the luminance values of the files in the first data set.The experimental results show that the shortcut artificial neural network performs better than the standard trained network, but still does not yield good results.We compare these results to two well known steganalysis tools.To date, no steganalysis technique has shown much promise, but this is highly experimental research.Many questions remain unanswered, and this thesis forms the basis for future experiments with using an artificial neural network as a useful steganalysis tool.