Steganalysis On Audio Data Through Probabilistic Neural Network Using An Ideal Bias

Abstract

In this study, we have developed a new steganalysis method that combines both Chi-square and Probability Neural Network (PNN) techniques. We have tested our steganalysis method on digital audio files. The stego files under test have been assumed to contain a secret data file embedded by an unknown steganography method, which is Least Significant Bit (LSB) method in our test environment. The method developed first analyzes the LSB bits in cover audio files then obtained intermediate data are applied to PNN to achieve more realistic results. While most steganalysis methods focuses on only detection of secret data, our method can also determine the rate of secret data. The obtained results have been further optimized as real as possible by the help of a PNN supported by an ideal bias selection mechanism.