CN116033086B - Reversible neural network-based image hiding method - Google Patents
Reversible neural network-based image hiding method Download PDFInfo
- Publication number
- CN116033086B CN116033086B CN202211625266.6A CN202211625266A CN116033086B CN 116033086 B CN116033086 B CN 116033086B CN 202211625266 A CN202211625266 A CN 202211625266A CN 116033086 B CN116033086 B CN 116033086B
- Authority
- CN
- China
- Prior art keywords
- image
- pseudo
- expression
- follows
- input
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000002441 reversible effect Effects 0.000 title claims abstract description 35
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 27
- 238000013507 mapping Methods 0.000 claims abstract description 20
- 238000009792 diffusion process Methods 0.000 claims abstract description 9
- 230000006835 compression Effects 0.000 claims abstract description 8
- 238000007906 compression Methods 0.000 claims abstract description 8
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims description 36
- 238000010168 coupling process Methods 0.000 claims description 35
- 230000008878 coupling Effects 0.000 claims description 31
- 238000005859 coupling reaction Methods 0.000 claims description 31
- 230000014509 gene expression Effects 0.000 claims description 24
- 239000011159 matrix material Substances 0.000 claims description 21
- 238000012545 processing Methods 0.000 claims description 12
- 238000013527 convolutional neural network Methods 0.000 claims description 8
- 230000000739 chaotic effect Effects 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 230000009467 reduction Effects 0.000 claims description 3
- 101100001672 Emericella variicolor andG gene Proteins 0.000 claims 1
- 230000004931 aggregating effect Effects 0.000 claims 1
- 238000003384 imaging method Methods 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 10
- 230000006872 improvement Effects 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000001174 ascending effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000000465 moulding Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Landscapes
- Image Processing (AREA)
- Facsimile Transmission Control (AREA)
Abstract
The invention relates to the field of image encryption, and discloses an image hiding method based on a reversible neural network, which comprises the following steps: and carrying out coding compression on the target image I with the size of M multiplied by N to obtain a compressed image A. And calculating a hash value H of the target image, and generating a pseudo-random sequence of the three-dimensional hyperchaotic mapping based on the hash value H. And scrambling the compressed image A by using the pseudo-random sequence, and performing exclusive-or diffusion on the scrambled image to obtain a ciphertext image V. And embedding the ciphertext image V into the carrier image C by using a reversible neural network to obtain a carrier image Q containing the ciphertext. According to the invention, the image is subjected to double-layer hiding by utilizing the three-dimensional hyperchaotic mapping and the reversible neural network, the hash value corresponding to the target image is calculated, and the hash value is utilized to generate the pseudo-random sequence serving as the three-dimensional hyperchaotic mapping, so that the subsequently generated secret key is associated with the target image, and the safety of the image hiding method is improved.
Description
Technical Field
The invention relates to the field of image encryption, in particular to an image hiding method based on a reversible neural network.
Background
With various advances in network and digital communication technologies, digital images and video based on digital images have become the primary sources of information storage and transmission. The security threat and technical weakness of network communication make it difficult to perform large-scale information transmission in specific fields such as civil use, military use, medical use and the like. Therefore, encrypting and hiding the image, and further protecting important data, has become a hotspot for information security research.
An existing image information hiding method takes secret information as input at an encoder, generates a general secret disturbance and adds the general secret disturbance to different carrier images, and utilizes an attention module to enable the encoder to suppress the secret disturbance which possibly causes higher attention in the channel dimension; the encoder is prompted to learn to generate secret-containing countermeasure disturbance through countermeasure training, so that a secret-containing image is simultaneously used as a countermeasure sample of an attack steganalysis model, and image information hiding is achieved. However, the above method only performs one layer of encryption hiding on the image, and still has the problem of low security.
Disclosure of Invention
The invention provides an image hiding method based on a reversible neural network, which aims to overcome the defect of low safety of the existing image hiding method.
In order to solve the technical problems, the technical scheme of the invention is as follows:
in a first aspect, the present invention proposes an image hiding method, including:
and carrying out coding compression on the target image I with the size of M multiplied by N to obtain a compressed image A.
And calculating a hash value H of the target image, and generating a pseudo-random sequence of the three-dimensional hyperchaotic mapping based on the hash value H.
And scrambling the compressed image A by using the pseudo-random sequence, and performing exclusive-or diffusion on the scrambled image to obtain a ciphertext image V.
And embedding the ciphertext image V into the carrier image C by using a reversible neural network to obtain a carrier image Q containing the ciphertext.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
(1) Firstly, generating a pseudo random sequence through three-dimensional hyper-chaotic mapping to construct a new key model, and encrypting the compressed image to obtain a ciphertext image, thereby realizing first-layer hiding. Then embedding the ciphertext image into the carrier image by using the trained reversible neural network to obtain a carrier image containing the ciphertext, so as to realize second layer hiding. The security of the image hiding method can be further improved by the two layers of encryption hiding.
(2) By calculating the hash value corresponding to the target image and utilizing the hash value to generate a pseudo-random sequence serving as three-dimensional hyperchaotic mapping, the hash values corresponding to different images are different, so that a subsequently generated secret key is associated with the target image, the secret carrier-containing image can resist a selected plaintext attack and a known plaintext attack, the secret keys are different when the secret carrier-containing image is used each time, the secret carrier-containing image can be ensured not to be broken down due to one-time secret key loss, and the safety of an image hiding method is improved.
Drawings
Fig. 1 is a flowchart of an image hiding method based on a reversible neural network according to an embodiment of the present application.
Fig. 2 is a view of a target image selected in an embodiment of the present application.
Fig. 3 is a schematic diagram of a DenceNet network in an embodiment of the present application.
Fig. 4 is a selected carrier image of an embodiment of the present application.
Fig. 5 is a gray value histogram of a selected carrier image according to an embodiment of the present application.
Fig. 6 is a gray value histogram of a dense carrier image of an embodiment of the present application.
Fig. 7 is a schematic diagram of a line forward affine coupling process according to an embodiment of the present application.
Fig. 8 is a schematic diagram of the inverse affine coupling process of the embodiment of the present application.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the present patent;
the technical scheme of the invention is further described below with reference to the accompanying drawings and examples.
Example 1
Referring to fig. 1, the present embodiment provides an image hiding method based on a reversible neural network, including:
s1: and carrying out coding compression on the target image I with the size of M multiplied by N to obtain a compressed image A.
S2: and calculating a hash value H of the target image, and generating a pseudo-random sequence of the three-dimensional hyperchaotic mapping based on the hash value H.
S3: and scrambling the compressed image A by using the pseudo-random sequence, and performing exclusive-or diffusion on the scrambled image to obtain a ciphertext image V.
S4: and embedding the ciphertext image V into the carrier image C by using a reversible neural network to obtain a carrier image Q containing the ciphertext.
According to the image hiding method based on the reversible neural network, firstly, a pseudo-random sequence is generated through three-dimensional hyperchaotic mapping to construct a new key model, a compressed image is encrypted, a ciphertext image is obtained, and first-layer hiding is achieved. Then embedding the ciphertext image into the carrier image by using the trained reversible neural network to obtain a carrier image containing the ciphertext, so as to realize second layer hiding. The security of the image hiding method can be further improved by the two layers of encryption hiding. By calculating the hash value corresponding to the target image and utilizing the hash value to generate a pseudo-random sequence serving as three-dimensional hyperchaotic mapping, the hash values corresponding to different images are different, so that a subsequently generated secret key is associated with the target image, the secret carrier-containing image can resist a selected plaintext attack and a known plaintext attack, the secret keys are different when the secret carrier-containing image is used each time, the secret carrier-containing image can be ensured not to be broken down due to one-time secret key loss, and the safety of an image hiding method is improved.
Example two
The present embodiment is an improvement on the basis of the reversible neural network-based image hiding method proposed in the first embodiment. The method comprises the following steps:
s1: and carrying out coding compression on the target image I with the size of M multiplied by N to obtain a compressed image A. The target image I selected in this embodiment is shown in fig. 2.
In this embodiment, an automatic encoder is used to encode and compress a target image I having a channel number of 1 and a size of mxn; the automatic encoder comprises 4 layers of convolutional neural networks which are connected in sequence; the number of channels is 1, the target image I with the size of MxN is input into the encoder and is sequentially processed by 3 layers of convolutional neural networks to obtain a characteristic diagram with the number of channels being 32 and the size of MxN, the characteristic diagram with the number of channels being 32 and the size of MxN is subjected to dimension reduction by the last layer of convolutional neural network to obtain the characteristic diagram with the number of channels being 1 and the size of MxNIs included in the compressed image a.
In the implementation process, as the color image is composed of three channels R, G and B, each channel can be independently compressed and hidden like a gray image, and finally the color visual image is synthesized. Therefore, in order to enable better expression, the present embodiment uses a grayscale image for compression concealment. Let the gray scale object image I be as large as the gray scale carrier image C of size m×n, denoted by C.
In addition, the output end of the automatic encoder is also connected with an automatic decoder, so as to form a DencNet network, as shown in FIG. 3, which is a structure diagram of the DencNet network in the embodiment of the application. The automatic encoder comprises 4 layers of convolutional neural networks which are connected in sequence; the number of channels is 1, the target image I with the size of 256 multiplied by 256 is input into the encoder and sequentially processed by 3 layers of convolutional neural networks to obtain a characteristic image with the number of channels being 32 and the size of 256 multiplied by 256, the characteristic image with the number of channels being 32 and the size of 256 multiplied by 256 is subjected to dimension reduction by the last layer of convolutional neural network to obtain a compressed image A with the number of channels being 1 and the size of 128 multiplied by 256, and the compressed image A is compressed to be 1/2 of the original image. The compressed image A is transmitted to an input layer of an automatic decoder for dimension ascending to obtain a characteristic diagram with the channel number of 32 and the size of 256 multiplied by 256, and then the image with the channel number of 1 and the size of 256 multiplied by 256 is restored through 4 layers of convolution nerves.
In contrast to the conventional compressed sensing, the embodiment uses the deep learning automatic encoder to compress the image, uses the features of the deep learning image to compress, and restores the main features of the image with the highest probability, so that the restoring effect is improved, the restoring speed of the image is also greatly improved, and the problems of long sampling time and long reconstruction time of the conventional compressed sensing are solved, thereby accelerating the decryption time.
S2: and calculating a hash value H of the target image, and generating a pseudo-random sequence of the three-dimensional hyperchaotic mapping based on the hash value H. The method comprises the following specific steps of:
s2.1: and calculating the 256-bit hash value H of the target image by using an SHA-256 hash algorithm.
In a specific implementation, the SHA-256 hash algorithm is a one-way anti-collision method that accepts plain text input of length L (integer multiple of 512) and generates a unique hash value or digest of fixed length n=256 bits (32 bytes)The requirement is that. The maximum length of the input message M of SHA-256 is 2 256 Bits. M (actual length l) is typically padded with p bits so that the length l input is a multiple of 512, with the last 64 bits remaining as 2 to the original input M 32 And (5) molding. The fill bit starts with l followed by the desired 0. Thus, the final input of SHA-256 becomes length l=l+p+64, which is divided into n=l/512 equally sized blocks, each block being 512 bits. SHA-256 initialises using 8 32-bit predefined initial vectors H 0 ＝h 0 ,h 1 ,h 2 ,h 3 ,h 4 ,h 5 ,h 6 ,h 7 Updated every 64 rounds. It ultimately generates a hash value H.
S2.2: based on the hash value H, generating an initial value of three-dimensional hyperchaotic mapping:
s2.2.1: dividing the hash value H into 16 sub-hash values H 1 -H 16 ；
S2.2.2: building three sets of keys xx with 8 sub-hash values as a set 0 、yy 0 And zz 0 The expression is as follows:
wherein E, F and G are intermediate variables of 16 scale calculated, E dec 、F dec And G dec Decimal values of E, F and G respectively,representing that bit exclusive OR operation is carried out, alpha, beta and gamma are random numbers, alpha, beta, gamma epsilon {1, -1}, delta, omega and mu are random large integers;
s2.2.3: since the initial value range of the chaotic map is between (-1, 1), the three sets of keys xx are utilized 0 、yy 0 And zz 0 Respectively generating three-dimensional super-mixingInitial value x of chaotic map 0 、y 0 And z 0 The expression is as follows:
initial value x 0 ，y 0 ，z 0 Evenly distributed in the interval (-1, 0) U (0, 1).
S2.3: substituting the initial value into the three-dimensional hyperchaotic mapping for iteration to generate a three-dimensional hyperchaotic mapping with the length ofThree sets of chaotic sequences XX, YY and ZZ.
In this embodiment, the initial value is substituted into the three-dimensional hyperchaotic mapping for iteration, and the expression is as follows:
wherein x is j 、y j And z j Three groups of hyper-chaotic mapping values subjected to j iterations are respectively a, b, c and h which are different control parameters and a, b, c E [0,10 ] 7 ]，h∈[1,2,3,…,10]Epsilon is a number that tends to 0 such that the denominator is non-zero.
S2.4: based on three sets of the chaotic sequences XX, YY and ZZ, a first pseudo-random sequence R and a second pseudo-random sequence S are generated, and the expressions are as follows:
wherein,representing a round-up, mod (·) represents a modulo operation.
S3: scrambling the compressed image A by using the pseudo-random sequence, and performing exclusive-or diffusion on the scrambled image to obtain a ciphertext image V:
s3.1: throwing out the repeated random numbers in the first pseudo-random sequence R;
s3.2: the values of the integer sets {1,2,3, …,512×n } which are not present in the first pseudo-random sequence R are added to the end of R in order from small to large, resulting in an intermediate pseudo-random sequence R 512×n-i+1 ；
S3.3: expanding the image matrix of the compressed image A into a one-dimensional row vector and utilizing an intermediate pseudo-random sequence R 512×n-i+1 Performing position replacement on the one-dimensional row vector to obtain a scrambling sequence O;
s3.4: performing exclusive or diffusion on the elements in the second pseudo-random sequence S and the elements in the scrambling sequence O in a one-to-one correspondence manner to obtain a ciphertext image V; the expression for performing exclusive or diffusion is as follows:
wherein V is i Represents the ith element in the ciphertext image, S i Represents the ith element, O, in the second pseudo-random sequence i Representing the ith element in the scrambling sequence.
S4: embedding the ciphertext image V into the carrier image C by using a reversible neural network to obtain a carrier image Q containing the ciphertext:
s4.1: extracting a bit element in an image matrix of the ciphertext image V, and constructing a matrix K by using the bit element, wherein the expression is as follows:
K＝Vmod(10)；
s4.2: extracting the residual high-order elements in the image matrix of the ciphertext image V, and constructing a matrix L by using the high-order elements, wherein the expression is as follows:
L＝[V/10]；
s4.3: splicing the matrix K and the matrix L to obtain a matrix D;
s4.4: and (3) scrambling the matrix D, and respectively inputting the carrier image C and the matrix D subjected to scrambling into a trained reversible neural network for embedding treatment to obtain a carrier image Q containing the secret.
In this embodiment, the selected carrier image C is shown in fig. 4, the gray value histogram thereof is shown in fig. 5, and the gray value histogram of the dense carrier image Q is shown in fig. 6.
In this embodiment, the reversible neural network includes Z groups of affine coupling blocks sequentially cascaded, where each group of affine coupling blocks includes two input ends, two output ends, 3 basic convolution modules and 3 reversible operation modules; the input of two input ends of the first group of affine coupling blocks is a matrix D and a carrier image C which are subjected to scrambling processing respectively, each group of affine coupling blocks carries out affine coupling processing on the input of the two input ends by utilizing the basic convolution module and the reversible operation module, the output results of the two output ends are transmitted to the next group of affine coupling blocks to serve as input, and the final group of affine coupling blocks outputs a carrier image Q containing a secret.
In this embodiment, z=14, in each group of affine coupling blocks, the basic convolution module and the reversible operation module perform affine coupling processing on the inputs of the two input ends, including forward affine coupling processing and reverse affine coupling processing;
as shown in fig. 7, which is a schematic diagram of forward affine coupling processing whose expression is as follows:
as shown in fig. 8, which is a schematic diagram of the inverse affine coupling processing whose expression is as follows:
wherein y is 1 For the output of the affine coupling block first output, x 1 For an input of the affine coupling block first input,representing the basic convolution module->Performing convolution operation, x 2 For input to the second input of the affine-coupled block, y 2 For the output of the affine coupling block first output, +.>Representing the kronecker product operation, exp (·) represents the exponential operation, ρ (·) represents the convolution operation performed by the base convolution module ρ, η (·) represents the convolution operation performed by the base convolution module η.
Example III
The present embodiment is an improvement on the basis of the reversible neural network-based image hiding method proposed in the second embodiment.
In this embodiment, after three sets of keys are obtained, the keys are encrypted by using an RSA encryption algorithm, and the specific steps include:
randomly selecting two large prime numbers p and q, multiplying the large prime numbers p and q to obtain an integer n, and calculating an Euler function of the integer nThe expression is as follows:
randomly selecting an integer e as a public key, wherein the integer e is equal toMutually plain->
Using the integer e and Euler functionThe key d is encrypted as follows:
using the public key pair key m= (xx) 0 ,y 0 ,z 0 ) Encryption is performed, and the expression is as follows:
in summary, the image hiding method based on the reversible neural network is provided by the invention. Firstly, compressing a target image by using an automatic encoder, then encrypting the compressed image by using a chaos sequence generated by hyper-chaos mapping, finally embedding a ciphertext image into a carrier image by using a reversible neural network, encrypting an initial value of the hyper-chaos mapping by using a public key RSA, and then transmitting the encrypted initial value to a receiver. The automatic encoder can learn to compress data based on attributes, and the correlation between input feature vectors found from the data in the training process, so that the constructed model can reconstruct data similar to that observed in the training, a better effect can be achieved by using the automatic encoder to compress images, and the automatic encoder does not need to iterate for the traditional compression method using compressed sensing, so that the speed can be higher. In the aspect of ciphertext embedding, the ciphertext is embedded by using the reversible neural network, and the reversible neural network can be used for embedding and extracting the ciphertext in the same model, so that the complexity of the model is reduced, the realization is easier, the embedding amount of the ciphertext can be improved, and the image containing the ciphertext carrier is safer.
The terms describing the positional relationship in the drawings are merely illustrative, and are not to be construed as limiting the present patent;
it is to be understood that the above examples of the present invention are provided by way of illustration only and not by way of limitation of the embodiments of the present invention. Other variations or modifications of the above teachings will be apparent to those of ordinary skill in the art. It is not necessary here nor is it exhaustive of all embodiments. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the invention are desired to be protected by the following claims.
Claims (5)
1. An image hiding method based on a reversible neural network, comprising:
for a size ofTarget image of (a)IPerforming coding compression to obtain a compressed imageA；
Calculating hash value of the target imageHAnd based on the hash valueHGenerating a pseudo-random sequence of a three-dimensional hyper-chaotic map, comprising:
calculating 256-bit hash value of target image by SHA-256 hash algorithmH；
Based on the hash valueHGenerating an initial value of the three-dimensional hyper-chaotic map;
substituting the initial value into the three-dimensional hyperchaotic mapping for iterationAfter the second time, the length is +.>Is of chaotic sequence of (2)XXAndYY；
based on the chaotic sequenceXXAndYY，generating a first pseudo-random sequenceRAnd a second pseudo-random sequenceSThe expression is as follows:
wherein,representing a round up->Representing modulo arithmetic;
said hash value basedHThe specific steps of generating the initial value of the three-dimensional hyper-chaotic map include:
the hash value is processedHEqually divided into 16 sub-hash valuesH 1 -H 16 ；
Three sets of keys are constructed with 8 sub-hash values as a set of each、/>And->The expression is as follows:
，
wherein,E、FandGfor the calculated intermediate variable in 16-ary,E dec 、F dec andG dec respectively isE、FAndGis a decimal value of (c) in the code,representing the execution of a bit exclusive OR operation, ">Is a random number and->，/>、/>And->Is a random large integer;
using the three sets of keys、/>And->Respectively generating initial values of three-dimensional hyper-chaotic map>、/>And->The expression is as follows:
the initial value is substituted into the three-dimensional hyperchaotic mapping for iteration, and the expression is as follows:
wherein,respectively three groups of passesjHyper-chaotic mapping value of multiple iterations, +.>Respectively different control parameters +.>，/>，/>A number approaching 0;
-compressing said image using said pseudo-random sequenceAScrambling and performing exclusive or diffusion on the scrambled image to obtain a ciphertext imageVComprising:
will first pseudo-random sequenceRThrowing out the repeated random numbers;
aggregating integersIs not present in the first pseudo-random sequenceRThe values of (2) are added to +.>At the end of (2) to obtain an intermediate pseudo-random sequenceR 512×k-i+1 ；
Will compress the imageAIs spread into one-dimensional row vectors and uses an intermediate pseudo-random sequenceR 512×k-i+1 Performing position replacement on the one-dimensional row vector to obtain a scrambling sequenceO；
Second pseudo-random sequenceSIn (a) and scrambling sequence(s)OThe elements in the ciphertext image are subjected to exclusive OR diffusion in one-to-one correspondence to obtain the ciphertext imageVThe method comprises the steps of carrying out a first treatment on the surface of the The expression for performing exclusive or diffusion is as follows:
wherein,representing the first of the ciphertext imagesiElement(s)>Representing the first in the second pseudo-random sequenceiElement(s)>Representing the first in the scrambling sequenceiAn element;
using a reversible neural network to image the ciphertextVEmbedded into a carrier imageCIn (3) obtaining an image containing a dense carrierQComprising:
extracting ciphertext imageVBit elements in an image matrix of (a), with which the matrix is constructedK，The expression is as follows:
；
extracting ciphertext imageVRemaining high order bits in the image matrix of (2)Elements, constructing a matrix using the high-order elementsLThe expression is as follows:
；
the matrix is processedKSum matrixLPerforming splicing treatment to obtain a matrixD；
Pair matrixDScrambling and imaging the carrierCAnd a matrix subjected to scramblingDRespectively inputting the obtained images into a trained reversible neural network to perform embedding treatment to obtain an image containing a dense carrierQ。
2. The image hiding method according to claim 1, wherein the reversible neural network includesZThe affine coupling blocks are sequentially cascaded, and each affine coupling block comprises two input ends, two output ends, 3 basic convolution modules and 3 reversible operation modules;
wherein the inputs of the two input ends of the affine coupling block of the first group are respectively matrix after scrambling processingDAnd a carrier imageCEach group of affine coupling blocks performs affine coupling processing on the input of two input ends by utilizing the basic convolution module and the reversible operation module, and transmits the output results of the two output ends to the next group of affine coupling blocks as input, and the final group of affine coupling blocks outputs the dense carrier-containing imageQ。
3. The image hiding method according to claim 2, wherein in each group of affine coupling blocks, the basic convolution module and the reversible operation module perform affine coupling processing on inputs of two input terminals, including forward affine coupling processing and reverse affine coupling processing;
the expression of the forward affine coupling process is as follows:
the expression of the inverse affine coupling process is as follows:
wherein,y 1 for affine coupling the output of the first output of the block,x 1 for an input of the affine coupling block first input,representing the basic convolution module->A convolution operation is performed and,x 2 for the input of the affine coupling block second input,y 2 for the output of the affine coupling block first output, +.>Representing the Cronecker product operation, < ->Representing an exponential operation, ++>Representing the basic convolution module->Performing convolution operation, ++>Representing the basic convolution module->A convolution operation is performed.
4. The image hiding method according to claim 2, wherein after obtaining three sets of keys, the method further comprises: encrypting the key by using an RSA encryption algorithm, wherein the specific steps comprise:
randomly selecting two big prime numberspAndqand for large prime numberspAndqmultiplying to obtain integernThen calculate the integernEuler function of (2)The expression is as follows:
randomly selecting an integereIs a public key, the integereAnd (3) withMutually plain->；
By using the integereEuler functionPair keydEncryption is performed, and the expression is as follows:
，/>；
key pair using the public keym=Encryption, the expression of which is as follows:
,
,
。
5. the image hiding method according to claim 2, wherein the number of channels is 1 and the size is 1 using an automatic encoderTarget image of (a)IPerforming coding compression;
the automatic encoder comprises an input layer, a hidden layer and an output layer which are sequentially connected; the hidden layer comprises 4 layers of convolutional neural networks which are connected in sequence;
the number of the channels is 1, and the size is 1Target image of (a)IAfter being transmitted to the input layer, the input layer is sequentially processed by 4 layers of convolutional neural networks to obtain the channel number of 32 with the size of +.>The number of the channels is 32, and the size isAfter the feature map of (1) is subjected to dimension reduction by the output layer, the channel number is 1 and is equal to +.>Compressed image of (a)A。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211625266.6A CN116033086B (en) | 2022-12-16 | 2022-12-16 | Reversible neural network-based image hiding method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211625266.6A CN116033086B (en) | 2022-12-16 | 2022-12-16 | Reversible neural network-based image hiding method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116033086A CN116033086A (en) | 2023-04-28 |
| CN116033086B true CN116033086B (en) | 2024-04-09 |
Family
ID=86076906
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211625266.6A Active CN116033086B (en) | 2022-12-16 | 2022-12-16 | Reversible neural network-based image hiding method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116033086B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112417467A (en) * | 2020-10-26 | 2021-02-26 | 南昌大学 | Image encryption method based on anti-neurocryptography and SHA control chaos |
| CN112906043A (en) * | 2021-04-07 | 2021-06-04 | 华侨大学 | Image encryption method based on chaotic mapping and chaotic S-box substitution |
| CN113538203A (en) * | 2021-09-01 | 2021-10-22 | 华侨大学 | Image encryption method and device based on novel two-dimensional composite chaotic mapping and SHA-256 |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070050614A1 (en) * | 2005-08-24 | 2007-03-01 | Wen-Wei Lin | Robust hyper-chaotic encryption-decryption system and method for digital secure-communication |
-
2022
- 2022-12-16 CN CN202211625266.6A patent/CN116033086B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112417467A (en) * | 2020-10-26 | 2021-02-26 | 南昌大学 | Image encryption method based on anti-neurocryptography and SHA control chaos |
| CN112906043A (en) * | 2021-04-07 | 2021-06-04 | 华侨大学 | Image encryption method based on chaotic mapping and chaotic S-box substitution |
| CN113538203A (en) * | 2021-09-01 | 2021-10-22 | 华侨大学 | Image encryption method and device based on novel two-dimensional composite chaotic mapping and SHA-256 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116033086A (en) | 2023-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ahmad et al. | A secure image encryption scheme based on chaotic maps and affine transformation | |
| CN109903212B (en) | Image encryption method based on H geometric fractal and Hilbert curve | |
| Soleymani et al. | A chaotic cryptosystem for images based on Henon and Arnold cat map | |
| CN109660696B (en) | New image encryption method | |
| CN107659751B (en) | Multi-image encryption method based on DNA coding and chaos | |
| CN107292184B (en) | Image encryption method, device and key stream generating method and key stream generator | |
| CN110535624B (en) | Medical image privacy protection method applied to DICOM format | |
| CN107392970B (en) | Digital image encryption method based on bit plane and high-dimensional chaotic system | |
| CN114549266B (en) | Image encryption method based on DNA replacement rule and chaotic system | |
| Kumar et al. | Image cryptography with matrix array symmetric key using chaos based approach | |
| CN113691362A (en) | Bit plane image compression encryption algorithm based on hyperchaotic system and DNA coding | |
| CN113612596A (en) | Image self-decryption method based on hyperchaotic system | |
| CN113099069B (en) | Multi-image encryption method based on three-dimensional spiral model | |
| CN113114870A (en) | Multi-image encryption method based on three-dimensional Zigzag model | |
| Shakir et al. | A new four-dimensional hyper-chaotic system for image encryption | |
| CN116033086B (en) | Reversible neural network-based image hiding method | |
| CN116827509A (en) | Image encryption method based on five-dimensional conserved hyperchaotic system and bit plane segmentation diffusion | |
| CN115103080B (en) | Image encryption method and system based on DNA triploid variation | |
| Prasad et al. | A combined encryption compression scheme using chaotic maps | |
| CN109558701B (en) | Medical CT image secret sharing method | |
| CN117057804B (en) | Financial transaction data secure storage method and system based on hash sequence | |
| Yuanshi et al. | An algorithm for JPEG compressing with chaotic encrypting | |
| CN116318615B (en) | Image encryption method and decryption method based on combination of hyperchaotic system and DNA (deoxyribonucleic acid) coding | |
| CN117114959B (en) | Image encryption method based on key feedback mechanism of multi-parameter one-dimensional chaotic system | |
| CN114362918A (en) | Image encryption method based on central rule and three-dimensional bit plane |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |