A Practical Scheme For Robust Steganography Via Modifying Sign Of DCT Coefficients

Modern adaptive image steganography can effectively ensure the safe transmission of in-formation in lossless channel.However,more and more social network platforms have adopted lossy channels to transmit images,which results in the limitation of the application of adaptive steganography to covertly transmit information in actual social networks.Therefore,it is imper-ative to study the robust image steganography that can effectively transmit information in lossy channels.In order to improve undetectability and robustness of the steganography algorithm,this thesis proposes a robust image steganography algorithm based on DCT coefficients sign flipping Then,in order to solve the problem that the diversity of channel attacks on social platforms leads to the difference of sign algorithmic robustness,this thesis designs a partially channel-dependent sign steganography for social networksFirstly,in order to deeply understand the robustness of image steganography,this thesis sum-marizes and expands the image steganography.And the current two main robust steganographic ideas and related algorithms have been studied and discussed in depth.This thesis summarizes the problems existing in robust steganography at the present stage,such as the inability to resist JPEG attacks with multiple quality factors.The above work provides the direction and idea for the research of robust steganography in this thesis.Then,in order to solve the problem that steganography algorithm based on embedded do-main has the low correct extraction rate of secret message and the lock of undetectability,a novel image steganographic algorithm is proposed.First,by using the sign of DCT coefficients,that remains unchangeable before and after JPEG compression,we select the candidate coefficients for resisting JPEG compression;Second,the designed distortion function assigns cost for each candidate DCT coefficient;Finally,relying on both error correction code and Syndrome-Trellis Codes,an encoded message is embedded into the cover image with minimum embedding distor-tion.Compared with prior arts,extensive experimental results highlight both undetectability and robustness of our proposed sign algorithm.When the payload is 0.1,the proposed sign algorithm improves the undetectability by more than 7.7%.In addition,under the JPEG attack of different quality factors,the average correct extraction rates of proposed algorithm can reach more than 99%.Finally,in view of the difference of sign algorithmic robustness caused by the diversity of channel attacks on social platforms,this thesis proposes an improved sign steganography algo-rithm based on partially channel-dependent.According to the actual situation that the quality factor of JPEG compression attacks may be different when the image is transmitted by different social platforms and the characteristics of the cover image itself,different candidate coefficients are selected to carry the secret message,so as to achieve better steganographic robust perfor-mance.In addition,the improved algorithm is applied to Facebook and other platforms.Com-pared with sign steganography,the application results show that the improved algorithm adopts different steganographic schemes according to different social platforms and cover images,which can obtain 100%correct extraction rate.