Research On Adaptive Steganographic Algorithm Based On Distortion Function Design

Steganography is an information hiding technique in which secret message can be embedded in the cover image and then transmitted in a public channel.The purpose of steganography is to realize secure communication between secret message sender and receiver.At present,research of steganography is focus on adaptive steganography.Adaptive steganography can embed secret messages into the texture-rich regions of the cover image,thus it has a good ability when detected by steganalysis.Minimizing the additive distortion model is a common framework for designing an adaptive steganographic algorithm.Under this framework,this dissertation has been studied how to design distortion function for adaptive steganographic algorithm.The main works are as follows:In order to obtain the texture regions of cover image and improve the coherence of different pixels in steganography,an adaptive steganography algorithm based on non-directional filter is proposed.In the distortion function of this algorithm,a non-directional filter is used to extract residuals of cover images,in order to solve the problem that the correlations between image pixels are not considered in additive distortion model,a Gaussian low-pass filter is used to smooth the cover image residuals,thus the pixel residual weights spread to its neighboring pixels.Finally,after distortion function is defined,secret messages are encoded by syndrome trellis codes.Compared with S-UNIWARD algorithm,the proposed algorithm achieves a better performance on resist the state-of-the-art steganalysis under a large information embedding rate and gets a lower algorithm time complexity.Aiming at improving the performance of obtaining texture regions by directional filters and achieving better steganography security,an adaptive steganography algorithm based on directional filter is proposed.In the distortion function of this algorithm,an eight-direction filter bank is defined by Gabor wavelet,and then the cover image is decomposed into eight directions by this directional filter bank.Based on WOW algorithm,the distortion function is established with a similar way.Finally,according to the distortion function,secret messages are embedded by syndrome trellis codes.Experimental results demonstrate that the proposed algorithm achieves a better anti-detection performance than WOW algorithm.In addition,by introducing the strategy of changing focus position,the proposed algorithm is further improved.The improved algorithm reduces the possibility of embedding secret messages in isolated pixels so it achieves a better performance when attacked by the state-of-the-art steganalysis.