Research Of Security Steganography On Analysis Of Image High Dimensional Characteristic

With the rapid development of digital media technology, the research of steganography is becoming more and more extensive. Steganography was born at the end of last century. Until now, there have been a large number of outstanding image steganography algorithms. However, image steganalysis algorithm has been developed rapidly in the fierce confrontation with image steganography algorithm. With the emergence of the image steganalysis algorithms which are based on high dimensional characteristic in recent years, the research of steganography is facing a great challenge. Therefore, this paper will focuses on the research of security steganography on analysis of image high dimensional characteristic according to the application of high dimensional features in the confrontation between image steganography and image steganalysis.(1) HUGO(Highly Undetectable Steganography) can only minimize the distortion of first-order pixel difference features without considering the distortion of high-order pixel difference features. Therefore, the ability of Highly Undetectable Steganography to resist the detection operation of steganalysis algorithms which are based on high-order pixel difference features is weak. To overcome this shortage, an improved steganography algorithm is proposed based on the high dimensional features. Firstly, an improved distortion function is proposed based on the high dimensional quantization MINMAX features. Then, combined with the Gibbs theoretical framework, an improved steganography algorithm is proposed based on the above improved distortion function. Experimental results show that the proposed algorithm can effectively improve the anti-detection ability and appropriately improve the visual concealment, which meanwhile maintain the same embedding capacity.(2) Although the influence of the embedding location on the security of steganography is considered in OPVD(Octonary Pixel-value Differencing), there are still shortcomings. OPVD can only ensure that the secret information is embedded in the edge region. However, the selected embedding locations are not optimal. To overcome this issue, an improved steganography algorithm is proposed based on the high dimensional pixel difference features and the particle swarm optimization algorithm. Firstly, an improved objective function is proposed based on the criterion of the anti-detection ability. Then, the optimal embedding locations are selected by using the particle swarm optimization algorithm which is based on the above improved objective function. Finally, the secret information is embedded in these locations using the embedding mechanism of OPVD. Experiment results show that the anti-detection ability of the proposed method is improved, which meanwhile maintain the image quality.(3) The steganography scheme for JPEG covers based on the parametric distortion function has many advantages such as that it can prevent over-fitting. However, this algorithm can be detected by a steganalysis approach using neighbouring joint density features. To overcome this issue, an improved steganography algorithm is proposed based on the high dimensional joint probability features. Firstly, an improved distortion function is proposed based on the high dimensional joint probability features. Then, combined with the improved distortion function, an improved steganography algorithm is proposed. Experiment results show that both of the image quality and the anti-detection ability of the proposed method is improved under the premise of the maximum embedding capacity unchanged, compared with other similar algorithms.