CN116527278A - A Blockchain Covert Communication Method Based on Generative Steganographic Network and Image Double Steganography - Google Patents

Abstract

The invention relates to a block chain hidden communication method based on a generation type hidden network and image double hidden, which comprises the following steps: the sending end generates a carrier picture by utilizing a WGAN countermeasure network and a convolution steganalysis network; the sender divides the secret information into two parts and sequentially embeds the secret information into carrier pictures by using an airspace steganography algorithm and a ciphertext domain reversible information hiding algorithm to obtain a second carrier picture, compresses and stores the second carrier picture into an IPFS file system, and returns a hash plaintext abstract; the sender encrypts the hash plaintext abstract based on a certificate-free public key cryptosystem to obtain a hash ciphertext abstract and signature information; the receiving party takes the hash plaintext abstract as input of an IPFS file system, and obtains a second encrypted picture embedded by the sending party; and the receiver decrypts the second encrypted picture by using a space domain steganography algorithm and a ciphertext domain reversible information hiding algorithm according to the encryption process of the sender on the private information to obtain the private information.

Description

A Blockchain Covert Communication Based on Generative Steganographic Network and Image Double Steganography method

technical field

The invention belongs to the technical field of artificial intelligence and data security, and in particular relates to a block chain covert communication method based on generative steganography network and image double steganography.

Background technique

Covert communication technology is a solution for secure transmission of secret information and hiding the communication process between the two parties by protecting the communication channel from eavesdropping. The existing implementation methods of covert communication include anonymous communication, digital watermarking and steganography. As a specific implementation method of covert communication, steganography is widely used to hide information by embedding secret information into carrier information to generate unsuspected secret information.

With the development of deep learning technology, in view of the problem that attackers can illegally obtain secret information by observing or enumerating steganographic algorithms, covert communication based on deep learning can generate carrier information through generative adversarial networks, and the generated carrier information can be embedded with The original carrier is more similar, and then the embedded secret-carrying information is classified by the convolutional neural network. When the secret-carrying information is detected as not embedded, it is transmitted as the final secret-carrying information. Although this method can eventually generate secret information with higher similarity, it must be used in conjunction with a steganographic algorithm with a low embedding rate. When the amount of secret information transmitted is large, even if the convolutional steganalysis network is used, the attacker can It is easy to distinguish carrier information and secret information. When the attacker uses a more accurate steganographic detection network, the carrier generated by generating an adversarial network cannot be used as a suitable and effective carrier, and it is still difficult to resist steganalysis attacks, and this type of scheme is always on a centralized untrusted network It is difficult to guarantee the security of secret information when transmitting information.

Blockchain has the characteristics of decentralization, node anonymization, data immutability and transparency. The covert communication of blockchain based on deep learning ensures the transmission of information in a trusted decentralized network, and at the same time ensures the anonymity of the identities of both parties in the communication. However, with the introduction of blockchain technology, the process of generating secret information still faces the problem of high storage overhead. Some methods use IPFS to compress secret information into strings to reduce the storage pressure of transmitting strings on the blockchain network. However, most solutions use the LSB algorithm to embed strings. This method will generate a large number of account transaction addresses, resulting in a high cost of transmitting strings. In addition, during the block generation process, other non-recipient network node users can also obtain transaction information, resulting in security threats.

Aiming at the problems of steganalysis attack, low embedding rate, insecure information transmission, and high transmission overhead faced in the existing covert communication invention, the present invention proposes a generative steganalysis network and image double steganographic blockchain Covert communication scheme.

Contents of the invention

In order to solve the problems existing in the background technology, the present invention provides a block chain covert communication method based on generative steganography network and image double steganography, the purpose is to ensure the security of covert information transmission and the integrity of covert information ,include:

S1: The sender uses the generator in the WGAN confrontation network to generate an image sample G(z), and embeds random values into the image sample through the spatial steganography algorithm to obtain the generated image Steg(G(z)) as the input of the discriminator. The image data set trains the WGAN confrontation network;

S2: After the WGAN confrontation network training is completed, the sender will generate the picture Steg(G(z)) as the input of the convolutional steganalysis network, and use the real picture data set to train the convolutional steganalysis network. The convolutional steganalysis network screens out the unembedded image from the generated image Steg(G(z)) as the carrier image;

S3: The sender divides the secret information into two parts and uses the space domain steganography algorithm and the ciphertext domain reversible information hiding algorithm to sequentially embed the two parts of the secret information into the carrier image to obtain the second secret image; compress and store the second secret image in IPFS file system, return hash plaintext digest, and set access control permissions;

S4: The sender encrypts the hash plaintext digest based on the certificate-free public key cryptosystem to obtain the hash ciphertext digest and signature information, and encodes the hash ciphertext digest and signature information using Base58 encoding, and the encoded hash password The string "0" is added between the text abstract and the signature information for splicing to obtain the signature information string δ';

S5: The sender removes the repeated characters in the signature information string δ', re-splicing to obtain an intermediate signature string δ" with different characters, and combines the transaction address of the node in the blockchain network with the intermediate signature string δ" Match, and record the index of the successfully matched character in δ' and the node transaction address, fill in the extraData field of the node transaction in the order of the index; the node transaction will be packaged into a block after being verified, when the block in the After all transactions are verified by other nodes, the block will be transmitted through the blockchain network;

S6: The receiver obtains the index information filled in the extraData field of the node transaction from the block, obtains the hash ciphertext summary and signature information by decoding the index information, verifies the signature information and uses the sender's private key to encrypt the hash summary The text is decrypted into a hash plaintext summary;

S7: The receiver uses the hash plaintext abstract as the input of the IPFS file system, and obtains the second encrypted image embedded in the sender;

S8: According to the encryption process of the sender's private information, the receiver uses the space domain steganography algorithm and the ciphertext domain reversible information hiding algorithm to decrypt the second encrypted picture to obtain the secret information.

Further, the training of the WGAN confrontation network includes:

Calculate the distance between the data distribution of Steg(G(z)) and the data distribution of the real picture, so as to distinguish the real picture from the generated picture, and convert the Wasserstein distance into:

Limit the variation range of the continuous function f _w , that is, the gradient is w, and the value of w is [-c,-c]. The loss function transformed into the WGAN confrontation network is:

in, Indicates the probability that the discriminator distinguishes Steg(G(z)) under the distribution of random noise z～p _z , /> Indicates the probability of discriminating the real image under the distribution of the real data set x is x~p _r ;

The purpose of the generator is to make the Wasserstein distance smaller, and its loss function is The purpose of the discriminator is to make the distance between the real data distribution and the generated data distribution larger, and its loss function is The generator and the discriminator perform gradient descent to update the parameters and optimize iteratively until the loss function converges.

Further, the carrier image includes:

When the SRNet network discriminates the generated Steg(G(z)) as a non-embedded picture, use the Steg(G(z)) picture as the carrier picture; otherwise, continue to train the SRNet network until the Steg(G(z)) ) until it is judged that there is no embedded image.

Further, the step of using the space domain steganography algorithm and the ciphertext domain reversible information hiding algorithm to embed two parts of secret information into the carrier picture in turn to obtain the second secret-carrying picture includes:

S31: Using an airspace steganography algorithm to embed a first secret information into the carrier picture to obtain a first secret-carrying picture;

S32: Using the ciphertext domain reversible information hiding algorithm to embed the second secret information into the first encrypted image to obtain the second encrypted image;

Said embedding the second secret information into the first secret-carrying picture to obtain the second secret-carrying picture comprises:

S321: Use the first row and the first column of the first secret-carrying picture as fixed pixels, and the predicted values of the fixed pixels remain unchanged; other non-fixed pixels are predicted by the median predictor MED to obtain a predicted image;

S322: Compare the binary representations of the pixel values of the non-fixed pixels in the first secret-carrying picture and the predicted image from the highest bit to the lowest bit one by one, and use the number of bits of the same pixel value as the corresponding mark position value, and the mark position of the fixed pixel The value is set to -1, and the mark position value of the non-fixed pixel is converted into an eight-bit binary representation;

S323: Count the probability of occurrence of each mark position value, construct Huffman tree data, calculate the Huffman code corresponding to each mark position value according to the rule of left 0 right 1, and record the position map information; wherein, the position map information It consists of additional information and the binary representation of the Huffman code sequence. The additional information includes the binary representation of the length of the Huffman code corresponding to the mark position value, the correspondence between the Huffman code sequence and the mark position value, and the Huffman code sequence the binary representation of the total length;

S324: Use the preset encryption key _e1 to generate a pseudo-random matrix R of the same size as the first secret-carrying picture, multiply the pseudo-random matrix R and the corresponding pixels of the first secret-carrying picture to obtain an encrypted image, and encrypt the pixel value of the image expressed in binary;

S325: Select the first N rows and the first M columns of pixels in the encrypted image as encrypted fixed pixels, and embed the recorded position map information from left to right and from top to bottom in order to encrypt fixed pixels; *2 Divide into multiple pixel blocks, and embed the Huffman coding sequence corresponding to the marker position value in the four pixels in the pixel block to the pixel block. Binary representation of encrypted fixed pixels from left to right from top to bottom; if the embedding space is insufficient, embed in the next pixel block until the binary representation values of all encrypted fixed pixels are embedded;

S326: Use the preset encryption key e ₂ to encrypt the second secret information. After the encrypted fixed pixels are embedded, the extra embedding space in the pixel block starts to embed the encrypted second secret information. After the embedding is completed, the second secret information is obtained. Upload encrypted pictures.

Further, embedding the Huffman coding sequence corresponding to the marker position value in the four non-fixed pixels in the pixel block into the pixel block includes:

Among them, t represents the marked position value of the non-fixed pixel, and b _s represents the Huffman coding sequence corresponding to the marked position value. When the marked position value t of the non-fixed pixel in the pixel block is less than or equal to 7, the embedding of the pixel in the pixel block The size of the space is t+1 bits.

Further, said encrypting the hash plaintext digest includes:

S41: KGC generates a random number λ, uses the random number λ as the system private key s, multiplies the system private key s by the base point G on the elliptic curve, and obtains the system public key P _pub =sG;

S42: The sender's user ID is ID, and KGC generates a random number r _ID to calculate R _ID = r _ID G, where G represents a fixed base point on the elliptic curve, and generates a partial private key D _ID = r _ID + s H (ID, R _ID ), and return the R _ID and part of the private key D _ID to the sender, and H represents the hash function;

S43: The sender verifies whether D _ID ·G=R _ID +P _pub ·H(ID,R _ID ) is established, if established, D _ID is valid, otherwise invalid;

S44: The sender randomly selects the secret number x _ID , calculates X _ID = x _ID G, publishes the sender's public key PK _ID = (R _ID , X _ID ), and calculates the private key as SK _ID = (D _ID , x _ID );

S45: The sender encrypts and transmits the private key SK _ID to the receiver through the negotiated session key, and uses the public key PKID=(R _ID , X _ID ) to encrypt the hash plaintext digest m to obtain the hash ciphertext digest m';

S46: The sender selects a random number t, and calculates respectively T=t·G, v=H(m,T,R _ID ,X _ID ) and u=t=x _ID (H(ID,R _ID )+v)+ D _ID generates signature information δ=(u, v).

Further, said decoding the index information to obtain hash ciphertext digest and signature information includes:

S61: The receiver checks all transaction information related to the sender from the block, filters out the index information contained in the extraData field of all transaction information; reassembles and restores the signature information string δ according to the index information contained in the extraData field of all transaction information ';

S62: The receiver separates the restored signature information string δ' through the string "0" to separate the Base58-encoded signature information and hash ciphertext digest; use Base58 decoding to obtain the sender's signature information and hash ciphertext digest ;

S63: The receiver uses the sender's public key to verify the signature information, and if the verification is passed; uses the sender's private key to decrypt the hashed ciphertext digest to obtain the hashed plaintext digest.

Further, the receiver using the sender's public key to verify the signature information includes:

h=H(ID,R _ID ),T'=u·GX _ID *(h+v)-R _id -P _pub *h,v'=H(m,T',R _ID ,X _ID ), if v=v', then the verification is passed.

The present invention has at least the following beneficial effects

(1) WGAN is used to generate confrontation network and SRNet steganalysis network, which solves the training problem of traditional GAN and the detection accuracy of whether it contains secrets, and uses the airspace steganography algorithm to process the generated samples during the WGAN training process to ensure the final generated The carrier image is not distorted and has the ability to resist steganalysis attacks.

(2) Divide the carrier information into two parts and use the spatial domain steganography algorithm and the ciphertext domain reversible information hiding algorithm to embed in turn, which solves the problem of neural network training algorithms with high embedding rates and prevents attackers from enumerating the spatial domain steganographic algorithm. Information is extracted for increased security.

(3) By storing the string index, the storage overhead is reduced, access control permissions are set to prevent malicious internal nodes from obtaining secret information, and the security of the decentralized network is ensured.

Description of drawings

Fig. 1 is the overall communication flowchart of the embodiment of the present invention;

FIG. 2 is a flowchart of a generative steganographic network according to an embodiment of the present invention;

FIG. 3 is a flow chart of image steganography embedding according to an embodiment of the present invention;

Fig. 4 is a flow chart of blockchain data transmission according to an embodiment of the present invention.

Detailed ways

Embodiments of the present invention are described below through specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific implementation modes, and various modifications or changes can be made to the details in this specification based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the diagrams provided in the following embodiments are only schematically illustrating the basic concept of the present invention, and the following embodiments and the features in the embodiments can be combined with each other in the case of no conflict.

Wherein, the accompanying drawings are for illustrative purposes only, and represent only schematic diagrams, rather than physical drawings, and should not be construed as limiting the present invention; in order to better illustrate the embodiments of the present invention, some parts of the accompanying drawings may be omitted, Enlargement or reduction does not represent the size of the actual product; for those skilled in the art, it is understandable that certain known structures and their descriptions in the drawings may be omitted.

In the drawings of the embodiments of the present invention, the same or similar symbols correspond to the same or similar components; , "front", "rear" and other indicated orientations or positional relationships are based on the orientations or positional relationships shown in the drawings, which are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred devices or elements must It has a specific orientation, is constructed and operated in a specific orientation, so the terms describing the positional relationship in the drawings are for illustrative purposes only, and should not be construed as limiting the present invention. For those of ordinary skill in the art, the understanding of the specific meaning of the above terms.

Please refer to Figure 1, the present invention provides a block chain covert communication method based on generative steganography network and image double steganography, including:

Please refer to Figure 2, S1: The sender uses the generator in the WGAN confrontation network to generate an image sample G(z), and embeds random values into the image sample through the spatial domain steganography algorithm to obtain the generated image Steg(G(z)) as a discriminator The input of the WGAN confrontation network is trained using the real picture data set;

The generation model G of WGAN is a multi-layer perceptron, which is mainly used to generate fake picture samples. By inputting random noise z (z is a random variable subject to Gaussian distribution), picture samples G(z) can be generated;

After the image sample G(z) is generated, the appropriate amount of information, that is, random values, is embedded in the image sample through the spatial adaptive steganography algorithm to obtain the encrypted image sample Steg(G(z)) as the input of the discriminator.

The obtained Steg(G(z)) is input to the discriminant model D of WGAN, and the other part of the discriminant model is the real picture data set.

Calculate the distance between the data distribution of Steg(G(z)) and the data distribution of the real picture, so as to distinguish the real picture from the generated picture, and convert the Wasserstein distance into:

Limit the variation range of the continuous function f _w , that is, the gradient is w, and the value of w is [-c,-c]. The loss function transformed into the WGAN confrontation network is:

in, Indicates the probability that the discriminator distinguishes Steg(G(z)) under the distribution of random noise z～p _z , /> Indicates the probability of discriminating the real image under the distribution of the real data set x is x~p _r ;

The purpose of the generator is to make the Wasserstein distance smaller, and its loss function is The purpose of the discriminator is to make the distance between the real data distribution and the generated data distribution larger, and its loss function is The generator and the discriminator perform gradient descent to update the parameters and optimize iteratively until the loss function converges.

S2: After the WGAN confrontation network training is completed, the sender will generate the picture Steg(G(z)) as the input of the convolutional steganalysis network, and use the real picture data set to train the convolutional steganalysis network. The convolutional steganalysis network screens out the unembedded image from the generated image Steg(G(z)) as the carrier image;

Preferably, the training of the convolutional steganalysis network SRNet includes:

The input of the convolutional steganalysis network SRNet is Steg(G(z)), and the other part of the input is the real picture data set. The first 7 layers of the SRNet network extract the Steg(G(z)) steganographic embedded data through the convolution operation. High-frequency noise, followed by 8-12 layers of residual-connected convolutional layers, and average pooling operations are used to reduce the dimension of the feature map, and finally the fully-connected layer is used as a classifier to classify pictures; the perceptual loss selected by SRNet The function form is as follows:

PerceptualLoss＝λ*ContentLoss+γ*StyleLossss

Among them, λ and γ are parameters that weigh the content loss and style loss, and control the weight of the two losses in the total loss; Content Loss is used to measure the similarity between the generated image and the real image at the feature level, and Style Loss is used for The measure of the style difference between the generated image and the real image is the class difference.

The training process of the convolutional steganalysis network SRNet is as follows:

(a) Input the low-resolution image into SRNet, and calculate the output image through forward propagation. Then, the output image is compared with the corresponding high-resolution image to calculate the value of the loss function. Next, the network parameters are updated using the stochastic gradient descent backpropagation algorithm, reducing the value of the loss function.

(b) During training, hyperparameters such as learning rate, regularization term, and batch size are adjusted to obtain better model performance.

(c) Iterative training: Repeat steps a and b until the predetermined number of training rounds is reached or the model performance meets the requirements.

(2) Obtain the carrier image: When the SRNet network judges the generated Steg(G(z)) as a non-embedded image, use the Steg(G(z)) image as the carrier image; otherwise, continue to train the SRNet network until Steg(G(z)) is judged as a picture without embedding.

As shown in Figure 3, S3: the sender divides the secret information into two parts and uses the space domain steganography algorithm and the ciphertext domain reversible information hiding algorithm to embed the two parts of the secret information into the carrier picture to obtain the second secret picture; The second encrypted image is compressed and stored in the IPFS file system, and the hash plaintext summary is returned, and the access control permission is set;

Preferably, said using the space domain steganography algorithm and the ciphertext domain reversible information hiding algorithm to sequentially embed two parts of secret information into the carrier image to obtain the second secret image includes:

S31: Using an airspace steganography algorithm to embed a first secret information into the carrier picture to obtain a first secret-carrying picture;

S32: Using the ciphertext domain reversible information hiding algorithm to embed the second secret information into the first encrypted image to obtain the second encrypted image;

Said embedding the second secret information into the first secret-carrying picture (secret-carrying picture 1) to obtain the second secret-carrying picture comprises:

S321: Use the first row and the first column of the first secret-carrying picture as fixed pixels, and the predicted values of the fixed pixels remain unchanged; other non-fixed pixels are predicted by the median predictor MED to obtain a predicted image;

Preferably, the predicted pixels by MED include:

Wherein, the left-upper adjacent three pixels of p pixel are respectively c, b, and a.

S322: Compare the binary representations of the pixel values of the non-fixed pixels in the first secret-carrying picture and the predicted image from the highest bit to the lowest bit one by one, and use the number of bits of the same pixel value as the corresponding mark position value, and the mark position of the fixed pixel The value is set to -1, and the mark position value of the non-fixed pixel is converted into an eight-bit binary representation:

S323: Count the probability of occurrence of each mark position value, construct Huffman tree data, calculate the Huffman code corresponding to each mark position value according to the rule of left 0 right 1, and record the position map information; wherein, the position map information It consists of additional information and the binary representation of the Huffman code sequence. The additional information includes the binary representation of the length of the Huffman code corresponding to the mark position value, the correspondence between the Huffman code sequence and the mark position value, and the Huffman code sequence the binary representation of the total length;

S324: Use the preset encryption key _e1 to generate a pseudo-random matrix R of the same size as the first secret-carrying picture, multiply the pseudo-random matrix R and the corresponding pixels of the first secret-carrying picture to obtain an encrypted image, and encrypt the pixel value of the image expressed in binary;

Among them, x ^k (i,j) represents the pixel in the first classified image, Represents the pixel in the encrypted image, k represents the binary digits of the pixel value, i and j represent the row and column of the image.

S325: Select the first N rows and the first M columns of pixels in the encrypted image as encrypted fixed pixels, and embed the recorded position map information from left to right and from top to bottom in order to encrypt fixed pixels; *2 Divide into multiple pixel blocks, and embed the Huffman coding sequence corresponding to the marker position value in the four pixels in the pixel block to the pixel block. Binary representation of encrypted fixed pixels from left to right from top to bottom; if the embedding space is insufficient, embed in the next pixel block until the binary representation values of all encrypted fixed pixels are embedded;

S326: Use the preset encryption key e ₂ to encrypt the second secret information. After the encrypted fixed pixels are embedded, the extra embedding space in the pixel block starts to embed the encrypted second secret information. After the embedding is completed, the second secret information is obtained. Upload encrypted pictures.

Preferably, embedding the Huffman coding sequence corresponding to the marker position value in the four non-fixed pixels in the pixel block into the pixel block includes:

Among them, t represents the marked position value of the non-fixed pixel, and b _s represents the Huffman coding sequence corresponding to the marked position value. When the marked position value t of the non-fixed pixel in the pixel block is less than or equal to 7, the embedding of the pixel in the pixel block The space size is t+1 bits;

Please refer to Figure 4, S4: The sender encrypts the hash plaintext digest based on the certificateless public key cryptography system to obtain the hash ciphertext digest and signature information, and encodes the hash ciphertext digest and signature information using Base58 encoding respectively. The string "0" is added between the final hash ciphertext digest and the signature information for splicing to obtain the signature information string δ';

S41: KGC generates a random number λ, uses the random number λ as the system private key s, multiplies the system private key s by the base point G on the elliptic curve, and obtains the system public key P _pub =sG;

S42: The sender's user ID is ID, and KGC generates a random number r _ID to calculate R _ID = r _ID G, where G represents a fixed base point on the elliptic curve, and generates a partial private key D _ID = r _ID + s H (ID, R _ID ), and return the R _ID and part of the private key D _ID to the sender, and H represents the hash function;

S43: The sender verifies whether D _ID ·G=R _ID +P _pub ·H(ID,R _ID ) is established, if established, D _ID is valid, otherwise invalid;

S44: The sender randomly selects the secret number x _ID , calculates X _ID = x _ID G, publishes the sender's public key PK _ID = (R _ID , X _ID ), and calculates the private key as SK _ID = (D _ID , x _ID );

S45: The sender encrypts and transmits the private key SK _ID to the receiver through the negotiated session key, and uses the public key PK _ID = (R _ID , X _ID ) to encrypt the hash plaintext digest m to obtain the hash ciphertext digest m ';

S46: The sender selects a random number t, and calculates respectively T=t·G, v=H(m,T,R _ID ,X _ID ) and u=t+x _ID (H(ID,R _ID )+v)+ D _ID generates signature information δ=(u, v).

S5: The sender removes the repeated characters in the signature information string δ', re-splicing to obtain an intermediate signature string δ" with different characters, and combines the transaction address of the node in the blockchain network with the intermediate signature string δ" Match, and record the index of the successfully matched character in δ' and the node transaction address, fill in the extraData field of the node transaction in the order of the index; the node transaction will be packaged into a block after being verified, when the block in the After all transactions are verified by other nodes, the block will be transmitted through the blockchain network;

S6: The receiver obtains the index information filled in the extraData field of the node transaction from the block, obtains the hash ciphertext summary and signature information by decoding the index information, verifies the signature information and uses the sender's private key to encrypt the hash summary The text is decrypted into a hash plaintext summary;

Preferably, said decoding the index information to obtain hash ciphertext digest and signature information includes

S61: The receiver checks all transaction information related to the sender from the block, filters out the index information contained in the extraData field of all transaction information; reassembles and restores the signature information string δ according to the index information contained in the extraData field of all transaction information ';

When reassembling, since the index information includes the index of the character in δ' and the node transaction address, the character is restored through the index information contained in the extraData field and the node transaction address, and δ' is restored according to the index of the character in δ'.

S62: The receiver separates the restored signature information string δ' through the string "0" to separate the Base58-encoded signature information and hash ciphertext digest; use Base58 decoding to obtain the sender's signature information and hash ciphertext digest ;

S63: The receiver uses the sender's public key to verify the signature information, and if the verification is passed; uses the sender's private key to decrypt the hashed ciphertext digest to obtain the hashed plaintext digest.

Preferably, the recipient uses the sender's public key to verify the signature information includes:

h=H(ID,R _ID ), T'=u·GX _ID *(h+v)R _id -P _pub *h,v'=H(m,T',R _ID ,X _ID ), if v =v', the verification is passed.

S7: The receiver uses the hash plaintext abstract as the input of the IPFS file system, and obtains the second encrypted image embedded in the sender;

S8: The receiver decrypts the second secret-carrying picture (secret-carrying picture 2) by using the space domain steganography algorithm and the ciphertext domain reversible information hiding algorithm according to the encryption process of the private information by the sender to obtain the secret information.

Preferably, said decrypting the second secret-carrying picture to obtain the secret information includes:

(1) Extract position map information from the first N rows and the first M columns of pixels;

(2) According to the corresponding relationship between the Huffman code sequence and the position mark value, the total length of the Huffman code sequence, obtain the position mark values of all non-fixed pixels;

(3) Extract the pixel values of all fixed pixels; Put the fixed pixels back into the first N rows and the first M columns to obtain the image _I'e ;

(4) After extracting the Huffman coding sequence and the pixel value of the fixed pixel from the non-fixed pixel, the receiver obtains the decryption key e' ₂ corresponding to the encryption key e ₂ , and uses the decryption key e' ₂ to extract a part of the second Secret information, get image I'_e;

(5) The receiver obtains the decryption key e ₁ ' corresponding to the encryption key e ₁ , and uses the decryption key e ₁ ' to generate a pseudo-random matrix R to perform an XOR operation on I' _e to obtain the decrypted I", in the image I The first t bit or t+1 bit of each non-fixed pixel binary in "is different from the I' _e of the non-fixed pixel binary of the image. If t≤7, the t+1 bit of the image I' _e is predicted from the image I" pixel The values are opposite, and each non-fixed pixel is restored sequentially in the order of first row and then column; the restoration process is as follows: According to the restored fixed pixels of the image I", the predicted value of the non-fixed pixel p' _e1 (x, y) is sequentially calculated from row to column If the pixel position mark value t≤7, then the first t bits of the pixel value of the image I' _e and the non-fixed pixel prediction value> Same, the pixel value t+1 of the image I' _e is the same as the non-fixed pixel prediction value /> On the contrary, the pixel value t+2 to the 8th bit of the image I' _e is the same as the non-fixed pixel /> Similarly, after restoring the pixel value of the non-fixed pixel, replace the pixel value of the decrypted image I" with the restored pixel value, and continue to restore each non-fixed pixel from left to right from top to bottom, and finally get the first Confidential pictures, the formula for image recovery is as follows:

in, Represents the value of the high t bits of the predicted pixel, /> Indicates the value of the t+1th bit of the predicted pixel, express yes Execute exclusive or negation operation.

(6) Obtain the remaining first secret information from the first secret-carrying picture by using the decoding method of the spatial domain steganography algorithm, and finally obtain the private information by combining the two parts of the secret information.

In the method proposed by the present invention, steganalysis detection based on deep learning is effectively avoided through confrontational networks and steganalysis, and greater embedding rate and non-destructive It can effectively restore the original image while avoiding probabilistic feature detection. In addition, in the blockchain network, the number of required transactions is reduced by recording the index transmission in the extraData field, and the certificate-free public key key system is used for encryption and signature, which protects the identity privacy of the sender.

It should be noted that those skilled in the art can understand that the implementation of all or part of the processes in the above method embodiments can be completed by instructing related software and hardware through computer programs, and the programs can be stored in computer-readable memory In the medium, when the program is executed, it may include the processes of the above-mentioned method embodiments. Wherein, the storage medium may be a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM) or a random access memory (Random Access Memory, RAM) and the like.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention shall be included in the protection of the present invention. within range.

Finally, it is noted that the above embodiments are only used to illustrate the technical solutions of the present invention without limitation. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be carried out Modifications or equivalent replacements, without departing from the spirit and scope of the technical solution, should be included in the scope of the claims of the present invention.

Claims (8)

1. A block chain covert communication method based on generative steganography network and image double steganography, characterized in that, comprising: S1: The sender uses the generator in the WGAN confrontation network to generate an image sample G(z), and embeds random values into the image sample through the spatial steganography algorithm to obtain the generated image Steg(G(z)) as the input of the discriminator. The image data set trains the WGAN confrontation network; S2: After the WGAN confrontation network training is completed, the sender will generate the picture Steg(G(z)) as the input of the convolutional steganalysis network, and use the real picture data set to train the convolutional steganalysis network. The convolutional steganalysis network screens out the unembedded image from the generated image Steg(G(z)) as the carrier image; S3: The sender divides the secret information into two parts and uses the space domain steganography algorithm and the ciphertext domain reversible information hiding algorithm to sequentially embed the two parts of the secret information into the carrier image to obtain the second secret image; compress and store the second secret image in IPFS file system, return hash plaintext digest, and set access control permissions; S4: The sender encrypts the hash plaintext digest based on the certificate-free public key cryptosystem to obtain the hash ciphertext digest and signature information, and encodes the hash ciphertext digest and signature information using Base58 encoding, and the encoded hash password The string "0" is added between the text abstract and the signature information for splicing to obtain the signature information string δ'; S5: The sender removes the repeated characters in the signature information string δ', re-splicing to obtain an intermediate signature string δ" with different characters, and combines the transaction address of the node in the blockchain network with the intermediate signature string δ" Match, and record the index of the successfully matched character in δ' and the node transaction address, fill in the extraData field of the node transaction in the order of the index; the node transaction will be packaged into a block after being verified, when the block in the After all transactions are verified by other nodes, the block will be transmitted through the blockchain network; S6: The receiver obtains the index information filled in the extraData field of the node transaction from the block, obtains the hash ciphertext summary and signature information by decoding the index information, verifies the signature information and uses the sender's private key to encrypt the hash summary The text is decrypted into a hash plaintext summary; S7: The receiver uses the hash plaintext abstract as the input of the IPFS file system, and obtains the second encrypted image embedded in the sender; S8: According to the encryption process of the sender's private information, the receiver uses the space domain steganography algorithm and the ciphertext domain reversible information hiding algorithm to decrypt the second encrypted picture to obtain the secret information. 2. A kind of block chain covert communication method based on generative steganography network and image double steganography according to claim 1, it is characterized in that, described WGAN confrontation network is trained to comprise: Calculate the distance between the data distribution of Steg(G(z)) and the data distribution of the real picture, so as to distinguish the real picture from the generated picture, and convert the Wasserstein distance into: Limit the variation range of the continuous function f _w , that is, the gradient is w, and the value of w is [-c,-c]. The loss function transformed into the WGAN confrontation network is: in, Indicates the probability that the discriminator distinguishes Steg(G(z)) under the distribution of random noise as z～p _z , Indicates the probability of discriminating the real image under the distribution of the real data set x is x~p _r ; The purpose of the generator is to make the Wasserstein distance smaller, and its loss function is The purpose of the discriminator is to make the distance between the real data distribution and the generated data distribution larger, and its loss function is The generator and the discriminator perform gradient descent to update the parameters and optimize iteratively until the loss function converges. 3. A block chain covert communication method based on generative steganography network and image double steganography according to claim 1, wherein the carrier image comprises: When the SRNet network discriminates the generated Steg(G(z)) as a non-embedded picture, use the Steg(G(z)) picture as the carrier picture; otherwise, continue to train the SRNet network until the Steg(G(z)) ) until it is judged that there is no embedded image. 4. A block chain covert communication method based on generative steganography network and image double steganography according to claim 1, characterized in that, the use of space domain steganography algorithm and ciphertext domain reversible information hiding algorithm in turn Embedding two parts of secret information into the carrier picture to obtain the second secret-carrying picture includes: S31: Using an airspace steganography algorithm to embed the first secret information into the carrier picture to obtain the first secret-carrying picture; S32: Using the ciphertext domain reversible information hiding algorithm to embed the second secret information into the first encrypted image to obtain the second encrypted image; Said embedding the second secret information into the first secret-carrying picture to obtain the second secret-carrying picture comprises: S321: Use the first row and the first column of the first secret-carrying picture as fixed pixels, and the predicted values of the fixed pixels remain unchanged; other non-fixed pixels are predicted by the median predictor MED to obtain a predicted image; S322: Compare the binary representations of the pixel values of the non-fixed pixels in the first secret-carrying picture and the predicted image from the highest bit to the lowest bit one by one, and use the number of bits of the same pixel value as the corresponding mark position value, and the mark position of the fixed pixel The value is set to -1, and the mark position value of the non-fixed pixel is converted into an eight-bit binary representation; S323: Count the probability of occurrence of each mark position value, construct Huffman tree data, calculate the Huffman code corresponding to each mark position value according to the rule of left 0 right 1, and record the position map information; wherein, the position map information It consists of additional information and the binary representation of the Huffman code sequence. The additional information includes the binary representation of the length of the Huffman code corresponding to the mark position value, the correspondence between the Huffman code sequence and the mark position value, and the Huffman code sequence the binary representation of the total length; S324: Use the preset encryption key _e1 to generate a pseudo-random matrix R of the same size as the first secret-carrying picture, multiply the pseudo-random matrix R and the corresponding pixels of the first secret-carrying picture to obtain an encrypted image, and encrypt the pixel value of the image expressed in binary; S325: Select the first N rows and the first M columns of pixels in the encrypted image as encrypted fixed pixels, and embed the recorded position map information from left to right and from top to bottom in order to encrypt fixed pixels; *2 Divide into multiple pixel blocks, and embed the Huffman coding sequence corresponding to the marker position value in the four pixels in the pixel block to the pixel block. Binary representation of encrypted fixed pixels from left to right from top to bottom; if the embedding space is insufficient, embed in the next pixel block until the binary representation values of all encrypted fixed pixels are embedded; S326: Use the preset encryption key e ₂ to encrypt the second secret information. After the encrypted fixed pixels are embedded, the extra embedding space in the pixel block starts to embed the encrypted second secret information. After the embedding is completed, the second secret information is obtained. Upload encrypted pictures. 5. A block chain covert communication method based on generative steganography network and image double steganography according to claim 4, characterized in that the four non-fixed pixels in the pixel block correspond to the tag position values The Huffman coding sequences embedded on the pixel block include: Among them, t represents the marked position value of the non-fixed pixel, and b _s represents the Huffman coding sequence corresponding to the marked position value. When the marked position value t of the non-fixed pixel in the pixel block is less than or equal to 7, the embedding of the pixel in the pixel block The size of the space is t+1 bits. 6. A block chain covert communication method based on generative steganography network and image double steganography according to claim 5, wherein said encryption of hash plaintext digest comprises: S41: KGC generates a random number λ, uses the random number λ as the system private key s, multiplies the system private key s by the base point G on the elliptic curve, and obtains the system public key P _pub =sG; S42: The sender's user ID is ID, and KGC generates a random number r _ID to calculate R _ID = r _ID G, where G represents a fixed base point on the elliptic curve, and generates a partial private key D _ID = r _ID + s H (ID, R _ID ), and return the R _ID and part of the private key D _ID to the sender, and H represents the hash function; S43: The sender verifies whether D _ID ·G=R _ID +P _pub ·H(ID,R _ID ) is established, if established, D _ID is valid, otherwise invalid; S44: The sender randomly selects the secret number x _ID , calculates X _ID = x _ID G, publishes the sender's public key PK _ID = (R _ID , X _ID ), and calculates the private key as SK _ID = (D _ID , x _ID ); S45: The sender encrypts and transmits the private key SK _ID to the receiver through the negotiated session key, and uses the public key PK _ID = (R _ID , X _ID ) to encrypt the hash plaintext digest m to obtain the hash ciphertext digest m '; S46: The sender selects a random number t, and calculates respectively T=t·G, v=H(m,T,R _ID ,X _ID ) and u=t+x _ID (H(ID,R _ID )+v)+ D _ID generates signature information δ=(u, v). 7. A block chain covert communication method based on generative steganography network and image double steganography according to claim 6, characterized in that said decoding the index information to obtain the hash ciphertext summary and signature information include: S61: The receiver checks all transaction information related to the sender from the block, filters out the index information contained in the extraData field of all transaction information; reassembles and restores the signature information string δ according to the index information contained in the extraData field of all transaction information '; S62: The receiver separates the restored signature information string δ' through the string "0" to separate the Base58-encoded signature information and hash ciphertext digest; use Base58 decoding to obtain the sender's signature information and hash ciphertext digest ; S63: The receiver uses the sender's public key to verify the signature information, and if the verification is passed; uses the sender's private key to decrypt the hashed ciphertext digest to obtain the hashed plaintext digest. 8. A block chain covert communication method based on generative steganography network and image double steganography according to claim 7, characterized in that the recipient uses the sender's public key to verify the signature information includes: h=H(ID,R _ID ),T'=u·GX _ID *(h+v)-R _id -P _pub *h,v'=H(m,T',R _ID ,X _ID ), if v=v', then the verification is passed.