Novel Image Steganographic Algorithms Based On Chaos And LSBMR

With the development of technology, especially the computer and networktechnology, lots of media information spreads in the Internet, which brings someinformation security issues that cannot be ignored. Normally, the path will beexposed if people want to pass secret information. Attackers can take measures tostop the process of passing secret information. Then the technology called‘Information Hiding’ emerges to solve it. This technology can protect the pathpassing information. In other words, it can protect the message imbedded into filesfrom being perceived by attackers. Information Hiding has two major fields,watermarking and steganography. And this thesis conducts research in imagesteganpgraphy. The current digital image steganographic techniques are divided intospatial domain and transform domain steganographic algorithms. Considering thespecific file type, it can be divided into LSB family steganographic algorithms,JPEG image steganographic algorithms, context adaptive steganographic algorithms,palette image steganographic algorithms and other steganographic algorithms. It’sbetter to apply combination of several ideas at the same time. For example, theexcellent steganography named EALSBMR is developed form LSB algorithm andcontext adaptive idea.This thesis provides an improved algorithm in space domain and a novelalgorithm in transform domain. EALSBMR algorithm is an excellent algorithmusing LSBMR scheme and context adaptive idea, which can defense some famoussteganalysis algorithms. But this algorithm introduces one pulse distortion in thehistogram of the absolute value of pixel difference of the embedding unit. Attackerscan detect this pulse distortion using the steganalysis algorithm based on B-spline.This thesis provides one method to weaken this pulse distortion, which reduces theaccuracy of the steganalysis algorithms based on B-spline and keeps good imagequality. Contrast experiment shows that the security of the improved algorithmproposed by this thesis is significantly higher than that of original algorithm. Thisthesis provides a novel algorithm using LSBMR scheme in integer DCT domaincoefficients, so that secure message is embedded into coefficients. Firstly, encryptsecurity information into ciphertext using chaotic system, and divide one image intosub-images to operate integer DCT transform. Then sort these coefficients usingZigzag so that the sorted coefficients are from low frequency to high frequency. Thissteganography embeds ciphertext into high frequency coefficients preferentially anduses last two frequency coefficients as one embedding unit. The ciphertext isembedded into these embedding units; if these embedding units are not enough, embed ciphertext into the third and fourth coefficients from bottom and so on untilembedding all ciphertext. This novel steganography proposed in this thesis is thefirst steganography in the Integer DCT domain. This steganography can defense RSsteganalysis and reduce the accuracy of histogram scheme. The two algorithmsproposed in this topic are both secure, of high embedding rate and of high imagequality.