MP3 Steganalysis Based on Neural Networks

Abstract

In the history of human communication, the concept and need for secrecy between two parties has always been present. One way of achieving it is to modify the message so that it is readable only by the receiver, as in cryptography, for example. Hiding the message in an innocuous medium is called steganography whereas the counterpart of steganography, that is, discovering whether a message is hidden in a specific medium, is called steganalysis. In this paper, we propose a new model for steganalysis based on Artificial Neural Network (ANN) which is capable of detecting hidden messages within MP3 audio files. A three layer network with each layers interconnected with each other by three different transfer functions is used in this model. Statistical parameters calculated from dominant features extracted from stego MP3 files, are concatenated to form a single vector and subsequently is fed into the neural network. Gradient descent back propagation algorithm with adaptive learning rate is then used for training the network. Experimental results shows that the proposed model is robust to attacks, gives good stability and provides high accuracy in terms of minimum error rate.