Research On Steganography Algorithm And Protocol Based On Audio Synthesis Model

Encryption is the main technique to achieve secure communication.However,transmitting data in cipher text form exposes the existence of secret communication behavior.Steganography is a technique to hide secret information into a cover medium and obtain a stego medium similar to it.Transmitting data in cover media form both protects the secret message and hides the act of secret communication;therefore,steganography provides a higher level of security for secret communication.Steganalysis is a technique for detecting whether a media file harbors a secret message.In recent years,researchers have proposed numerous steganalysis algorithms with superior performance,posing a greater challenge to steganography.The existing adaptive steganography methods have been difficult to resist steganalysis detection well.In this paper,we conduct research on audio steganography to further improve the security of audio steganography using speech synthesis technology:to address the problem that steganographic coding,e.g.STCs cannot perfectly achieve minimum distortion,we propose a cover-recovery audio steganography scheme based on a text-to-speech model;to further approach steganographic theoretical security,we propose a steganographic scheme based on an explicit probabilistic audio generation model that uses arithmetic coding to achieve computationally provable security and related security protocols.The main research work and innovations of this dissertation can be summarized as follows:1.proposed a cover reproducible audio steganography method based on audio synthesis In the traditional adaptive steganography scheme,since the receiver does not have access to the original cover and modification information,it needs to complete the information embedding and extraction with the help of steganographic coding(channel codes).The performance of current steganographic coding can only approximate the rate distortion bound,and the computational complexity is high.In this paper,we design an adaptive speech steganography algorithm based on source coding based on the text-to-speech transformation generation model,using the robustness of speech semantics to steganographic modifications and the recoverability of cover speech after sharing the generation model.The algorithm can be embedded according to the optimal probability distribution.Experimental results show that the performance of the cover-reproducible speech steganography-based algorithm is better than the existing STC coding-based speech steganography scheme.2.proposed a secure steganography algorithm and protocol based on audio synthesis under computationally provable security The theoretical security of steganography has been a goal pursued by researchers,and the emergence of computergenerated media with explicit probabilities provides a practical application basis for provably secure steganography theory.In this paper,we propose a computationally provably secure steganography scheme based on the audio synthesis model WaveNet,and verify its security from both theoretical and experimental perspectives.The experimental results of steganography analysis show that an attacker cannot distinguish between the cover and the stego audio.On the other hand,deep learning enables attackers to better analyze the abnormal behavior of both sides of the communication,so it is especially important to further improve the behavioral security of both sides of the communication.In this paper,based on the secure steganography algorithm,we propose a secure steganography protocol,which implements the whole process of steganographic communication through a steganographic channel,including three parts:symmetric key steganography,public key steganography and key exchange process.