High Capacity Steganography Algorithm In Digital Images

Digital image steganography, one of the important branches of information hiding, isapplied to more and more fields in modern times. But some special fields require higherpracticality of this technology, such as military, diplomatism and commerce. On one hand, moststeganography algorithms in spatial domain cannot resist intense steganalysis, and improvedalgorithms usually get higher anti-steganalysis capability at the cost of hiding capacity. On theother hand, algorithms in transform domain have good robustness, but their hiding capacity istoo low to satisfy the practical applications very well.The thesis mainly focuses on improving the hiding capacity of image steganographyalgorithms and making them more practical. The main work and innovation are as follows:(1) An improved BPCS (Bit-Plane Complexity Segmentation) algorithm based onpreprocessing is proposed. BPCS has large hiding capacity and good imperceptibility, but it isvulnerable to the complexity histogram attacks. Because the improved BPCS algorithm makesreverse preprocessing compensation to the changes of complexity histogram, it bears goodimperceptibility and resists complexity histogram attacks very well. While satisfying the samesecurity, the improved BPCS algorithm has larger hiding capacity than the similar one, for thecompensation process is taken before hiding secret information.(2) A high capacity image steganography algorithm in the DCT domain is proposed. Thesteganography algorithms in the DCT domain cannot satisfy practical applications because oflow hiding capacity. On one hand, the algorithm enhances the security of secret information andincreases hiding capacity by preprocessing the secret information and cover-image. On the otherhand, the determination of embedding position, intensity, time and method is on the principle ofincreasing hiding capacity while guaranteeing the steganography in security. The experimentalresults show that the algorithm has higher hiding capacity than traditional ones in the DCTdomain, and bears good imperceptibility and robustness.(3) An improved Outguess steganography algorithm based on dynamic compensation isproposed. Outguess algorithm corrects the statistics after embedding secret information, but itwastes hiding capacity to modificate and compensate the first-order statistics. The improvedalgorithm adopts the MC (Markov Chain) model to compensate the second-order statisticsdynamically while embedding secret information. Without reserving specific redundant bits tocorrect the statistics, hiding capacity of the improved algorithm is increased obviously and itsimperceptibility is further enhanced.