Research On Carrierless Steganography Method For Secure Covert Communication

Steganography is an essential branch of information hiding,and it is a "behavioral security" technology.On the other hand,encryption is a "content security" technology.Steganography and encryption form the basis of secure communication technology.Traditional steganography modifies a specific carrier to embed secret information in an "invisible" way.The essence of steganography is to disguise covert communication behavior as normal behavior to avoid the suspicion of attackers.However,with the development of steganalysis methods,the possibility of steganography being detected by steganalysis methods exists as long as the carrier is modified.Coverless steganography was proposed to resist steganalysis.Coverless steganography "acquires or generates" stego-objects directly,driven by the secret information.It is difficult to distinguish the stego-objects used by coverless steganography from similar normal objects on the network.Thus,coverless steganography can effectively resist detection by steganalysis methods and has higher security.It plays a significant role in the covert transmission of sensitive information such as military intelligence and private information.Coverless steganography is becoming one of the hot issues in covert communication.Therefore,the research of coverless steganography-based covert communication methods is of great significance and importance for developing national information security and privacy protection.This paper studies coverless steganography based on image,text,and behavior.The main work and innovation of this paper are elaborated as follows.(1)In order to resist the detection of steganalysis,this paper proposes a coverless steganographic model based on stego-images searching.It establishes the mapping relationship between secret information and image features by statistically analyzing the features of largescale image datasets and then searching natural images from image datasets whose own feature information contains(or is equivalent to)secret information to deliver secret information.Based on this model,this paper firstly proposes a coverless image steganographic method based on the Bag-of-Words(BOW)model.Then,in order to improve the information extraction efficiency,this paper proposes a coverless image steganographic method based on the average value of image pixels.After that,to improve the robustness,this paper proposes a coverless image steganographic method based on Faster RCNN object detection.The experimental results show that the proposed methods can effectively resist steganalysis.In addition,the method based on the average value of image pixels has better real-time information extraction ability,and the method based on Faster RCNN has better robustness.(2)Although the stego-images-searching-based methods can resist steganalysis,it has the problems of low hiding capacity,the need for a large-scale candidate image dataset,and poor correlation among multiple stego-images.To solve the above problems,this paper proposes a coverless steganographic method based on anime image generation.In this paper,the secret information is transmitted in anime image generation.First,the secret information is converted into recognizable anime image attribute labels.Then,the attribute labels are used as constraints of the generative model to generate the stego-anime-images.Finally,the secret information is recovered by extracting the attribute labels from the stego-anime-images.Because this method uses the constraints of the anime generative model to transmit the secret information,it does not need to build the candidate image dataset,which significantly reduces the storage overhead of the covert communication method.Also,the method can transmit any combination of secret information,which further improves the hiding capacity and the adaptability of covert communication.In addition,there are many anime images generated using the generative model on the network.The stego-anime-images in this paper are indistinguishable from those generated normally without secret information,which can still resist the detection of steganalysis algorithms.(3)The coverless steganographic method based on anime image generation is only suitable for anime enthusiasts.If others use it suddenly,it is likely to arouse the suspicion of the enemy.To this end,this paper proposes a coverless steganographic method based on readable long text generation.The proposed method can generate multiple secret sentences that are semantically continuous and logically related to a certain topic based on the identity of the communicating parties and secret information.The stego-text can be delivered through all public web platforms and is applicable to all people.The proposed method can also solve the problems of low hiding capacity and the need for a large size candidate text database of text selection-based methods,and the problem of semantic contextual incoherence or semantic inconsistency when generating multiple stego-sentences by former text generation-based methods.It first determines the topic of the generated text according to the scenarios of the communication parties.Then,the Plug and Play Language Model(PPLM)is explored to generate long readable text that conforms to the topic and is semantically coherent.A given secret message is hidden during text generation by selecting appropriate words in an established pool of embeddable candidate words.Experimental results show that the proposed method significantly increases hiding capacity while maintaining good imperceptibility compared to the existing text steganographic methods.(4)In order to resist the anomalous behavior analysis against the communicating party and further improve the hiding capacity,robustness and security,this paper proposes a coverless steganographic method based on the recommendation of move behavior in the game of gobang.The proposed method first trains a gobang model,then combine the model and the rules of the gobang to build an embeddable candidate drop behavior pool;finally,select the corresponding drop position according to the secret information.The communicating parties can transmit the secret information through online games or post videos of the game process or chess manual.Because the secret game behavior and normal game behavior are indistinguishable,the proposed covert communication method based on game behavior has better security and resistance to steganalysis detection.In addition,the chess manual of gobang or the video recording of the game behavior process will not affect the judgment of the move order even if it is affected by compression or noise during the transmission in the public channel,so the method has better robustness.