Research On The Digital Watermarking And Steganalysis Algorithms Of Audio Information Hiding

With the popularity of Internet applications, the transmission opportunities of multimedia data become increasing. How to manage and protect these multimedia data has attracted more attentions in the field of information security. Lots of problems and application requirements raise when the digital content distributes on networks, which generally can be divided into two parts:the copyright protection and camouflage confidential communication.Initially, the researchers want to use traditional cryptography to solve the security problems of multimedia content. But the traditional encryption methods can not meet its requirements. So the researchers begin to look for new solutions as a complement for the traditional cryptography. Information hiding is a new solution for the security of digital media content, which embeds secret information into digital carrier imperceptibility. Digital watermarking and steganography are two important branches of inforamtaion hiding. The carrier can be digital graph, audio, video, text and network protocol.Audio digital watermarking embeds the secret message which related with the audio into the audio to protect the audio file. Steganography embeds the secret message to hide the existence of the hidden message for confidential communication. According to the research achievements at present, audio information hiding can be roughly divided into four aspects: digital watermarking, watermarking attack, steganography and steganalysis. Audio watermarking algorithm focuses on kinds of audio watermark embedding and extracting watermarking algorithms. Audio watermarking attack mainly research on various attack methods to audio watermarking, which target is that obstacle the watermark information extracted smoothly. Audio steganographic methods mainly research on various algorithms used to hide the secret information in audio carrier used for confidential communication. Steganalysis is the opposite aspect of Steganography, which mainly studies various technical analysis algorithms. The algorithms are used to judge the audio is the stego audio or not.This dissertation focuses on audio digital watermarking and steganalysis, and proposes two novel audio watermarking algorithms and two novel specific audio steganalysis algorithms. The main content of this dissertation is summarized as follows:1. A novel DWT-DCT-SVD based audio blind watermarking algorithm is proposed. In this algorithm, the audio is split into non-overlapping frames, and each frame is decomposed on one-dimensional discrete wavelet transform (DWT), then the approximate sub-band coefficients are decomposed on discrete cosine transform (DCT) and partitioned into2-D matrix A for the first quarter of the DCT coefficients, then A decomposed on singular value decomposition (SVD) and get a diagonal matrix S. the watermarking information is embedded into S(1,1) according the ratio of S(1,1) and S(2,2). The change of S-matrix does not affect the value of matrix A, the algorithm has good transparency. The algorithm has better robustness because the watermark information is embedded in the approximate sub-band coefficients. S-matrix has a strong stability, and the change of matrix A does not greatly affect the value of S-matrix, so the algorithm can resist all kinds of common audio signal processing attacks. The experiments show that the algorithm has strong robustness against the common audio signal processing such as additive white Gaussian noise, MP3compression, resampling, low-pass filtering, requantization, cropping and denoising.2. A new DWT-based semi-fragile audio watermarking algorithm used for content authentication is proposed. In this algorithm, the original audio is decomposed on three layers DWT transform to get DWT approximation coefficients. Then the values of all approximate coefficients are computed and as the quantization step. Then the approximate coefficients are split into non-overlapping blocks, and the mean value of each approximate component block is quantized and embedded watermark information. The transparency of this algorithm is excellent and can tolerate MP3compression, Re-sampling, low-pass filter and other conventional signal processing. As a content authentication algorithm, this algorithm can accurately detect the replacing positions from mute replaced, or using natural audio instead of the Stego audio, or even two methods with combination. The accuracy of this algorithm is6%3. A novel specific steganalysic algorithm based on DCT is proposed. The analysis of the Steganography changed the statistical characteristics of DCT coefficient symbols. Based on this, a new difference ratio steganalysis algorithm is proposed. The detection method is to make all test carriers have second Steganography. Since the DR value proportion of the natural audio is significantly higher than that of Stego audio before and after the second Steganography, then the threshold testing could judge whether the carrier is the Stego audio. The experiment results show that the accuracy of this algorithm is80.6%.4. A novel special steganalysis algorithm based on echo hiding is proposed. Based on the specialty that echo hiding alter nature language epstrum distribution, this algorithm constructs cepstrum and difference variance of VDSC (Variants of Sum of Cepstrum) to reflect the influence of steganalysis algorithm to natural audio, which can be used to judge whether the test audio is stego audio. Analyzing process uses different migration conditions to calculate VDSC value of the testing carrier. If there is no significant change of the VDSC values and it is below the value of relative judgment domain, the test audio is the original audio. If there is significant change of the VDSC values and it is above the value of relative judgment domain, then the test audio is the stego audio. The experiment shows that the accuracy of this algorithm is88.2%.