Matrix Embedding For Large Payloads And Its Applications In Steganography

The purpose of steganography is covert communication by embedding secret messages intomultimedia data, and an essential problem is to how to embed information as much as possibleby less modification on the cover, that is, how to improve the embedding efficiency. Thisproblem can be solved by steganographic codes (stego codes). According to embeddingdistortion for each pixel, stego codes models can be divided into three models: basic model, wetpaper model and wet-scale model, which correspond constant distortion, two level distortion andmultilevel distortion respectively. Studies have shown that the codes of three models can beboiled down to the codes for large embedding rates by ZZW(Zhang Zhang Wang), Paper Foldingor Wet ZZW(WET Zhang Zhang Wang), and reconfigurable distortion respectively. Therefore,the study of large embedding rates codes for stego codes and steganography algorithm have theimportant meaning.This thesis mainly study the construction of fast codes for large embedding rates and how toimprove security of steganography using large embedding rates codes, and obtain the followingresults:1. Constructing the fast codes for large embedding rates. Two fast matrix embeddingschemes are proposed. we use bit-control and matrix extending base on pre-existing matrixembedding, to expand solution space of optimal solution exponentially with increasing thecomputational complexity linearly. The novel methods can exponentially decrease computationalcomplexity while keeping high embedding efficiency, and can flexibly make a trade-off betweenthe embedding efficiency and embedding speed.2. Proposing a fast±2stego codes and adaptive±2steganography. By using fast matrixembedding scheme in LSB and second LSB, the number of pixel with±2embedding can bedecreased via revising objective function of matrix embedding. The experiments show that thenovel method can increase embedding speed and embedding efficiency than pre-existing±2stego codes. Furthermore, based on±2stego codes,±2adaptive steganography is proposed.Byusing high five bits of pixels to calculate noisy level, the recipient can identify the embeddingregion after±2embedding. The experiments show that the proposed method can achievestronger security3. A novel adaptive±1steganography for spatial image is proposed. By extending thebounder of noisy region in NRE(Noisy Region Embedding), the "DLE"(double layeredembedding) is exploited to reduce the number of±1modifications, in which fast matrixembedding and WPC (wet paper codes) are applied to the LSB(Least Significant Bit) plane andthe second LSB plane respectively, therefore the message is embedded into the noisiest areas andthe recipient can identify the embedding region. The experiments show that the proposed methodcan achieve stronger security and faster embedding speed.