Steganography And Steganalysis Of Compressed Domain Audio

As an important technology of information security, data hiding has attracted greatinterests and more attention. Steganography and steganalysis are two branches of data hiding.Steganography refers to the problem of sending messages hidden in “innocent looking”communication over a public channel so that an adversary eavesdropper cannot even detectthe presence of the hidden message. The countermeasure to steganography is steganalysis,and its aim is to reveal the hidden message by statistical analysis.This dissertation mainly focuses on the steganography and steganalysis of compresseddomain audio. First, to improve the performance of the stegographic algorithm in compresseddomain audio, some novel algorithms have been proposed. Then, to attack two specificcompressed domain steganographic tools, a number of steganalytic algorithms have beenproposed. Finally, a conclusuion with a discussion of the direction for future research is given.The dissertation includes three main parts:In the first part, review for steganography and steganalysis of compressed domain audiois given. With respect to the application background, the practical and theoretical values ofthe dissertation are described. The basic concepts, models of steganography and steganalysisare introduced briefly. The research progress of these two fields is surveyed in detail.In the second part, novel steganographic algorithms for compressed domain audio areproposed. Based on the parameters of the encoder, three compressed domainaudio steganographic algorithms are proposed. In the first algorithm, the window type ismapped with the parity of the secret bit. The secret message is embedded by adjustingthe window switching mechanism. The second algorithm successfully solves the defect of theMP3Stego algorithm. The flexibility of the choice of Huffman table is exploited in the thirdmethod. The experimental results show that the proposed algorithms are good atimperceptibility and undetectability. In addition to the encoder parameters, the audio data isused to achieve high capacity steganography. Through modifying the QMDCT coefficients,the secret message is embedded. The distortion of the stegnography can be always maintainedbelow the masking threshold with the psychoacoustic model of the encoder. The finalalgorithm expands the space of the Huffman codeword. Combined with the multibasesystem, a high capacity steganography is proposed. In the third part, specific steganalysis of compressed domain audio is studied. For twotypical compressed domain steganographic tools, passive and active steganalysis are carriedout. First, as for MP3Stego which is a typical MP3stego tool, the statsistics of quantizationstep and bit reservoir is extracted as the steganlytic features. The experimental results showthat the differential operation can improve the detection accuracy. Finally, the estimation ofembedding ratio of MP3Stego is studied. Recompression calibration is used to estimate thecover audio and change point detection methods are applied to locate the embedding position.For UnderMP3Cover, RS detection method is applied to estimate the embedding ratio.The experimental results show that the proposed algorithm can effectively attack theUnderMP3Cover. Meanwhile, the relationship between the parameters of RS method and theperformance of the proposed algorithm is verified.