| It’s an urgent problem to solve how to ensure the security of information resources during transmission and storage at present.Steganography is widely used for the transmission of secret messages,and many scholars have conducted deeply research on it.With the continuous progress of steganalysis technology,higher requirements have been put forward for steganography.In order to resist the detection of steganalysis algorithms,researchers have proposed the concept of coverless steganography.Compared with traditional steganography,coverless steganography does not need to modify the carrier,but obtains or directly generates the carrier driven by the secret data,making it difficult to arouse suspicion of third parties.The existing coverless steganography can be divided into two types: carrier selection and carrier construction.The principle of coverless steganography based on carrier image selection is to use the correlation between the attributes of the carrier image and the secret information to hide the secret information.As the length of the secret information increases,the number of steganography images also increases accordingly,which leads to low hiding capacity and poor correlation between the stego-images.However,the complete construction method in the carrier construction requires high computing power,which leads to difficult construction and limited practicality.Aiming at the practical situation that the load capacity is not high and the practicality is not great in curent coverless steganography,based on the characteristics of the semiconstructive method,using the behavior habits of users on social platforms,and combined with the current popular deep learning technology,this thesis proposes a coverless steganography scheme using small icons of Chinese characters as structural elements.And the covert communication system is implemented according to the steganography scheme.The main work content is described as follows:(1)In order to resist the detection of current steganalysis algorithms,a carrier image that is different from the modified steganography is created.Based on the user’s behavior habits on social platforms as the construction guidelines and driven by secret information,the corresponding icons are selected from the constructed icon library to generate a logical and easily transmitted image according to the construction guidelines.The suggestion content part of the construction guideline is obtained by web crawler technology,so that it is not easy to cause the suspicion of the third party when it is transmitted on the public channel,and ensures the security of the image during the transmission process.(2)In order to overcome the problem of limited embedding capacity caused by the difficulty of database creation,a small icons library of Chinese characters is established.Firstly,each small icon of Chinese characters is encoded in a steganography way.Then,using the idea of image stitching,the Chinese character icons corresponding to the segmented secret information is stitched together to improve the steganographic capacity of a single carrier image.Considering that the stego-images may encounter various kinds of interference when transmitted in public channels,a set of text image dataset with different types of interference is artificially synthesized.Finally,the deep learning technology is used to complete the recognition of the characters in the stego-image,and implement the correct extraction of the receiving end.This method has good robustness.(3)A complete secret information hiding and extraction scheme is designed.And a practical covert communication system based coverless steganography is built with the help of programming language.The system has completed the effective hiding and extraction of secret information,and is easy to operate,so it has good practical value.Finally,experiments are conducted to verify the improvement of the proposed method in this thesis in the indicators of security,robustness,capacity,and extraction efficiency.Compared with other steganography schemes,the proposed scheme has good performance. |