Research On Coverless Steganography Methods Based On Texts

With the rapid development of network and digital technology,information security is facing increasingly severe and complex security tests.As an important supplement to encryption technology,information hiding can effectively reduce the perceptibility of secret information and use the normal appearance of stego carriers to hide communication behaviors.However,traditional methods need to modify the carrier to embed secret information,which is easy to cause abnormal changes in the characteristics of the carrier and cannot resist the increasingly mature steganographic detection attacks.How to reduce the impact of information hiding on the stego carriers,avoid leaving traces of steganography and ensure the security of data transmission under open channel conditions,has gradually become an important issue in the related research field.In recent years,researches on coverless information hiding have been carried out gradually,providing effective solutions to the problems above.Different from the existing methods,coverless information hiding methods retrieve stego carriers from the text big data set,or generate the corresponding carriers directly.This type of method makes no modification to the carrier,so it has ideal detection resistance and safety.Due to the late start of relevant research,the existing coverless information hiding methods still have some limitations.This paper mainly explores the potential ways to improve the performance of the coverless information hiding methods.The main work and innovation points are as follows:(1)Aiming at the limitation that the search-based coverless information hiding method relies too much on large text data sets,a coverless information hiding method based on feature mapping was proposed in this paper.The method proposed implements text representation based on tone features of Chinese characters,which enables texts to represent binary information statically,and then the mapping between texts and secret information is established by feature matching and specified matching algorithm.Experimental results show that the success rate of the proposed method can reach 100% under small text set conditions.Under the condition of the same text big data set,the increase of hiding capacity reached 38.5%.(2)In order to solve the problem that the hiding capacity of search-based method is generally small,this paper proposed a coverless information hiding method based on word association characteristics.This method constructs the word node tree based on the word association feature,encodes the transfer path of the word node tree to embed the secret information,and establishes the mapping relationship between the transfer path of the word node and the text according to the isomorphic text set.The experimental results show that the proposed method has a stable hiding success rate,can effectively resist the existing steganographic detection methods,and its hiding capacity is significantly improved compared with similar algorithms.(3)Existing generative methods are difficult to strike a good balance between hiding capacity and concealment.Based on this,this paper proposes a generative coverless method based on adaptive steganography coding strategy.This method uses the GPT pre-training model to model the language,selects and encodes the candidate words according to the conditional probability distribution,outputs the corresponding candidate words according to the secret information bitstream,and thus realizes the stego-text generation driven by the secret information.Compared with the similar methods,the adaptive steganographic coding strategy can better filter out the candidate words with unreasonable conditional probability,so as to ensure the ciphertext generated is smooth and natural enough and effectively improve the concealment of the algorithm.