Research On Information Hiding Warfare Based On The Characteristics Analysis Of Residual-extracting Filters

With the rapid development of multimedia information processing technology,digital images have wide application in all fields of society and have been gradually becoming an important method of communication in people's daily life.Image steganography secures secret communication via hiding secret information into digital images,which possesses very important applications in military intelligence and personal privacy.However,image steganography is a double-edged sword.If criminals and terrorists use steganography to transmit secret information,which will bring in threats to country security and social stability.Therefore,strengthening the research of image steganalysis,discovering the secret information that hidden in images and blocking the spread of illegal information,which plays an important role in national security and social stability.In recent years,image steganography and steganalysis have been an important subject in the field of information security,which draw more and more academic researchers' attentions.Base on this background,focusing on the research of image steganography and steganalysis that base on synchronize-embedding-changes and rich model,we conducts the following two research works:1)Targeting at the steganography that bases on synchronize-embedding-changes,we put forward an algorithm which decimates pixels to construct a brand new steganalytic.The image steganography that bases on synchronize-embedding-changes makes secret information that hidden in images blend into low-frequency contents,which reduces the disturbance of original images and improves the safety of steganography.Targeting at the strategy that synchronize-embedding-changes,we propose an algorithm that decimates the pixel of the original target image to reconstruct a new image,and extract features from both the original target image and the reconstructed image using rich model features extraction algorithm.Theoretical analysis and experiment results show that the algorithm can not only reduce clustering of synchronized embeddings effectively,but also improve the accuracy of estimation probability in selection-channel-aware rich model,which makes the detection performance of our proposed features more better than the original one.2)we put forward image steganography that bases on the frequency-domain analysis of steganalytic filters.We reveal that there are a lot of low energy-spectrum intervals by decomposing the high-pass filters used in rich-model steganalysis.Basing on this findings,we utilize the content-adaptive strategy and put forward a novel method that builds a new distortion cost function.The new distortion cost function divides into two principal parts.The first part is weight function which has a very close link to the energy of spectrum.This weight function adjusts the contribution of different energy-spectrum to the total distortion cost value,and makes the secret information more likely to be embedded into lower energy-spectrum.The second part is the distortion cost functions that bases on content-adaptive strategy,which adjusts the number of secret information that hidden in high-frequency area of images.Although the new distortion cost function does not achieve great improvement compared with the classical algorithm,we find that our proposed weight model that consider both energy value and frequency still has a great performance compare with other models in the experiment process.At the same time,we also find out the reason why the new distortion cost function can not improve the performance greatlyTo sum up in conclusion,this paper presents a method that decimates pixels to reconstruct the feature of steganalysis which can improve the performance of steganalysis,and an image steganographic algorithm that based on the frequency-domain analysis of filters and a new design idea of constructing distortion cost function is proposed,which indicates our future research direction.