Research On Image Steganography And Steganalysis Based On Embedding Changes

As two techniques of mutual confrontation in the technology of covert communications,steganography and steganalysis are developing rapidly in recent years because of requirements of today's network information security.By using of image steganography,a variety of Internet information security issues could be solved,such as covert communication,protecting privacy and maintaining copyright.In this paper,the digital image is used as the carrier of the steganography research.The main contents in this thesis are how embedding changes will affect the security of steganography and the detection accuracy of steganalysis method.The research contents are as follows.1)A new adaptive steganography method is proposed from the perspective of digital steganography.The purpose of this method is to aim at scrambling the embedding changes when the message is embedded by using the distortion,function,and this method can improve the steganographic security.WOW algorithm is one of the most representative method steganography,and it can be help to embedded message into these domains of complicated textures of the cover image based on the embedding cost of each pixel.In this paper,we will design an improved WOW adaptive image steganography method to further improve the security of steganography.Firstly,we use a binary random matrix to preprocess the vector image to generate a new image.Next,three direction filters are used to measure the embedding cost of each pixel of the newly generated image.Finally,the secret information is embedded into the original cover image by using STC codes to achieve minimized distortion steganography.The experimental results show that this method has better performance than the previous WOW algorithm.2)An adaptive steganalysis method for color images is designed from the perspective of steganalysis.This method is able to location of the embedding changes as more as possible to reduce unchanged pixels disturbance to detection veracity rate of steganalysis.In this paper,we propose a content-adaptive steganalysis strategy for color images.The new strategy aims to just extract color rich model features from the pixels that may have been modified.Firstly,an embedding cost function is used to calculate the embedding costs of all the pixels in each channel.Then,a subset of the pixels of each channel with a smaller embedding cost is selected according to a set percentage.Finally,the intersection of the pixels of the three channels obtained by the intersection and the union of each of the three channels are calculated to obtain the final two different sets of pixels.In the process of extracting the CRM features,only the corresponding features are extracted from the subset of pixels value.Experimental results show that the proposed strategy performs better than the state-of-the-art color image steganalysis method.