StegoHunter: Passive audio steganalysis using Audio Quality Metrics and its realization through genetic search and X-means approach

Steganography is used to hide the occurrence of communication. This creates a potential problem when this technology is misused for planning criminal activities. Differentiating anomalous audio document (stego audio) from pure audio document (cover audio) is difficult and tedious. This paper investigates the use of a Genetic-X-means classifier, which distinguishes a pure audio document from the adulterated one. The basic idea is that, the various audio quality metrics (AQMs) calculated on cover audio signals and on stego-audio signals vis-a-vis their denoised versions, are statistically different. Our model employs these AQMs to steganalyse the audio data. Genetic paradigm is exploited to select the AQMs that are sensitive to various embedding techniques. The classifier between cover and stego-files is built using X-means clustering on the selected feature set. The presented method can not only detect the presence of hidden message but also identify the hiding domains. The experimental results show that the combination strategy (Genetic-X-means) can improve the classification precision even with lesser payload compared to the traditional ANN (Back Propagation Network).