STCs Minimizing Distortion Embedding Steganography Based On Optional Distortion Measurement And Combination Of Sub-block

As an active and critical component of network information security, digitalsteganographic techniques have been sudied and developmented in recent years. This thesis ismainly focused on blocked embedding, where JPEG a widely used image format on Internetwould act as a typical host. However, traditional JPEG steganographic algorithms even thehigh security algorithms with minimum embedding distortion, such as STCs (syndrome-trelliscodes),could be detected by a cutting based and self-calibration blind steganalysis. Aiming toimprove the security performance of a JPEG based steganography algorithm, our main worksare as follows:1. Focusing on minimizing distortion embedding steganography, especially STCs,distortion measurement could not reasonably response the actual coefficient modificationdistortion, a multi-distortion measurement was proposed. According to embed rate, it choosespower law function based on the relationship of DCT coefficients or HVS contrast function asthe distortion measurement. Then， the secret message bits are embedded by STCs.Steganography which embedding rate is high, it’s secret messages are hidden in the texturearea of the carrier, the distortion measurement function using HVS contrast function is morereasonable; Steganography which embedding rate is low, it’s secret messages are hiddendistributed in the carrier, they should be embedded in the bigger coefficient, the distortionmeasurement using a power law distribution based on distribution of DCT coefficients ismore reasonable. Experiments show that the proposed distortion measure function can morereasonably response the actual coefficient modification distortion than others，algorithm hashigher security both in spatial and frequency domain. Compared to MB, nsF5and STC-Constalgorithms, its detect probability is significantly reduced without reducing the capacity.2. In order to improve the insufficient randomness of STCs algorithm, we proposedY-STCs algorithm based on random selection of sub-block. We divided the image into B*B(B>8) block(B-block), and randomly selected sub-block of8*8(S-block) from the B-blockfor hiding information. According to the B-block’s position in the image, the selected S-blockreassembled into a large matrix M. the secret message bits are embedded/extracted to M by STCs. After the message is embedded, the matrix M will be divided into the same size asS-block. Finally, these S-blocks return to the original image. Experiments demonstrate that theproposed Y-STCs can effectively resist observation domain features steganalysis, and hashigher security both in spatial and frequency domain.