Audio Steganalysis Using Fractal Dimension and Convolutional Neural Network (CNN) Model

Abstract

The rate at which secret messages are being transmitted through various digital signal media is alarming; these operations are done in an unsuspicious manner and users transmit these messages without knowledge of the embedded secret messages. Audio steganalysis deals with detecting the presence of secret messages in audio messages. Some of the existing steganalysis methods are laden with having prior knowledge of the steganography methods adopted in embedding the secret message in an audio signal, which reduces the detection efficiency. Consequently, this research developed a Higuchi-based audio steganalysis method that detects secret messages without having prior knowledge of the embedding techniques used. The algorithm reduces the fractal dimension of the audio signal to extract relevant features, while convolutional neural network was used as classifier. The research records high accuracy (96%) when compared with previous research. The accuracy of the developed system shows its effectiveness in detecting embedded messages without prior knowledge of the deployed steganography method.