An image encryption and decryption method based on dynamic compressed sensing

Abstract

This invention relates to an image encryption and decryption method based on dynamic compressed sensing, belonging to the field of image information security technology. The invention includes: uniformly dividing an image of a preset size; inputting the image sub-blocks into a compression network; after passing through a two-dimensional Logistic chaotic mapping, performing sliding scrambling encryption and XOR diffusion encryption to obtain the final ciphertext image; after inverse diffusion and inverse scrambling, the ciphertext image is input into a deep reconstruction network to obtain reconstructed sub-block images; and the sub-block images are then stitched together to achieve image restoration. The method of this invention can realize image compression and encryption / decryption. This invention reduces the requirement for image sparsity during compression and encryption, and utilizes the self-learning capability of deep convolutional neural networks to improve the problem of difficulty in achieving high-quality reconstruction with low sampling rates.

Description

Technical Field

[0001] This invention belongs to the field of image information security technology, specifically relating to an image encryption and decryption method based on dynamic compression sensing. Background Technology

[0002] Images, as an important information carrier, are widely used in various fields. As a result, the security issues in the storage and transmission of sensitive images have arisen, attracting widespread attention from researchers. Typically, images have high pixel redundancy in order to represent a wealth of information, which increases the resource overhead of storage and network transmission. Therefore, encrypting and decrypting images after compression has become a crucial technology.

[0003] Currently, image encryption algorithms based on dynamic compressed sensing theory typically utilize chaotic mapping to generate a compression matrix. The sparse image is then compressed and encrypted to obtain the observed signal, which is further encrypted to obtain the final ciphertext image. This method suffers from several drawbacks, including the lack of dynamic adaptability in the constructed random compression matrix, the need for sparse transformation of the image to be processed, and the inability to achieve high-quality reconstruction at low sampling rates.

[0004] In recent years, artificial intelligence technology has developed rapidly, and convolutional neural networks (CNNs) applicable to the image field have attracted widespread attention from researchers. This invention combines block-based compressed sensing and CNNs to construct a dynamic adaptive compression network; it designs a reconstruction network based on CNNs; thereby eliminating the requirements of encryption algorithms on signal sparsity and sampling rate, while enhancing the image reconstruction quality. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] The technical problem to be solved by this invention is how to provide an image encryption and decryption method based on dynamic compressed sensing to solve problems such as the lack of dynamic adaptability of the constructed random compression matrix, the need for sparse transformation of the image to be processed, and the inability to achieve high-quality reconstruction at low sampling rates.

[0007] (II) Technical Solution

[0008] To address the aforementioned technical problems, this invention proposes an image encryption / decryption method based on dynamic compressed sensing, which includes the following steps:

[0009] Step S1: Input the original image and divide it into multiple sub-images, each sub-image being N×N in size;

[0010] Step S2: Use a dynamic compression network to perform convolutional compression and initial reconstruction on the sub-image to obtain a preliminary encrypted image;

[0011] Step S3: Calculate the 64-byte hash value of the sub-image using the SHA256 algorithm. Divide the hash value into two bytes with a step size of 2 bytes. Use the two numbers in the first 4 bytes as a common factor α, and divide the last 60 bytes into β1 and β2. Perform an XOR operation on β1 and β2 with α to obtain two 2-byte numbers. After converting them to decimal, divide them by 255 to obtain two initial keys as the initial values ​​x0 and y0 of the two-dimensional Logistic chaotic system.

[0012] Step S4: The two-dimensional Logistic chaotic system iterates over the initial values ​​x0 and y0 to generate two chaotic sequences x. n and y n From the chaotic sequence x n Extract 32×2 numbers, enlarge and round to the [0,32] interval. The first 32 numbers are used for the initial row sliding scrambling of the encrypted image, and the last 32 numbers are used for the initial column sliding scrambling of the encrypted image.

[0013] Then from the chaotic sequence y n Take 32×32×2 numbers, magnify and round them to the [0, 255] range, and divide them into two parts to serve as the random key for diffusion encryption;

[0014] After scrambling and diffusion encryption, the final encrypted ciphertext image is obtained.

[0015] Step S5: Decrypt the ciphertext image by using two random sequences with inverse XOR diffusion and inverse sliding scrambling to obtain the decrypted image;

[0016] Step S6: The decrypted image is reconstructed using a nonlinear deep reconstruction network to recover the sub-image;

[0017] Step S7: The sub-images are stitched together in the order of the cut pieces to restore the original image.

[0018] (III) Beneficial Effects

[0019] This invention proposes an image encryption and decryption method based on dynamic compressed sensing. The beneficial effects of this invention are as follows:

[0020] (1) Reduced requirements for image sparsity during compression and encryption: This invention uses convolutional layers with removed activation functions and biases to compress images, thus ignoring whether the image is sparsity;

[0021] (2) By utilizing the self-learning capability of deep convolutional neural networks, the problem of low sampling rate makes it difficult to achieve high-quality reconstruction is improved: This invention obtains an image reconstruction network model that achieves nonlinear operation by training a deep convolutional neural network. This model can obtain more feature information from images with low sampling rate, thereby achieving high-quality image reconstruction. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating an image encryption / decryption method based on dynamic compression sensing according to the present invention.

[0023] Figure 2 This is a block diagram illustrating the principle of an image encryption / decryption method based on dynamic compressed sensing according to the present invention.

[0024] Figure 3 This is a schematic diagram of a reconstructed network model structure according to the present invention. Detailed Implementation

[0025] To make the objectives, contents, and advantages of the present invention clearer, the specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples.

[0026] In view of the above problems, the present invention provides an image encryption and decryption method based on dynamic compressed sensing.

[0027] This invention provides an image encryption / decryption method based on dynamic compressed sensing, comprising:

[0028] Step S1: Input the original image and divide it into multiple sub-images, each sub-image being N×N in size; in one embodiment, N = 32.

[0029] Step S2: Use a dynamic compression network to perform convolutional compression and initial reconstruction on the sub-image to obtain a preliminary encrypted image;

[0030] Step S3: Calculate the 64-byte hash value of the sub-image using the SHA256 algorithm. Divide the hash value into 2-byte increments. The first 4 bytes are used as a common factor α, and the remaining 60 bytes are divided into β1 and β2. XOR α with β1 and β2 to obtain two 2-byte numbers. After converting them to decimal, divide by 255 to obtain two initial keys, which serve as the initial values ​​x0 and y0 of the two-dimensional Logistic chaotic system. Due to the perturbation sensitivity of the SHA256 algorithm, it ensures that... Figure 1 dense;

[0031] Step S4: The two-dimensional Logistic chaotic system iterates over the initial values ​​x0 and y0 to generate two chaotic sequences x. n and y n From the chaotic sequence x n Extract 32 × 2 numbers, round them to the nearest integer in the range [0, 32]. The first 32 numbers are used for the initial row-sliding scrambling of the encrypted image, and the last 32 numbers are used for the initial column-sliding scrambling of the encrypted image. Then, from the chaotic sequence y... nTake 32×32×2 numbers, enlarge and round them to the range [0, 255] and divide them into two parts, which are used as random keys for diffusion encryption; after sliding scrambling and diffusion encryption, the final encrypted ciphertext image is obtained;

[0032] Step S5: Decrypt the ciphertext image by using two random sequences with inverse XOR diffusion and inverse sliding scrambling to obtain the decrypted image;

[0033] Step S6: The decrypted image is reconstructed using a nonlinear deep reconstruction network to recover the sub-image;

[0034] Step S7: The sub-images are stitched together in the order of the cut pieces to restore the original image.

[0035] Furthermore, the process of using a dynamic compression network to perform convolutional compression and initial reconstruction of the sub-image specifically includes:

[0036] Step S21: Use convolution kernel W com ∈R N×N×M (Without setting activation functions and biases for the convolutional layers) for sub-images X∈R N ×N (N=32) Perform non-sliding convolution to achieve linear compression and obtain the observed signal.

[0037] Step S22: Use convolution kernels (Without setting activation functions and biases for the convolutional layers) for the observed signal Linear initialization and reconstruction are achieved by performing convolution operations. Here, M is the number of channels, and B... 2 For quantity.

[0038] Further, step S3 includes:

[0039] The sub-image is processed by SHA256 to obtain a 64-byte hash value. The hash value is then divided into segments with a step size of 2, resulting in a sequence K consisting of 32 two-byte values. I ={k i |i∈[1,32],k i ∈[0,2 8 ]}.make As a public key factor, the remaining 30 k... i The sequence is divided into two subsequences. Each subsequence is XORed with a public key factor to obtain two 2-byte numbers. These numbers are then normalized by dividing by 255 to obtain x0 and y0, as shown in the following formula:

[0040]

[0041]

[0042] Furthermore, the iterative process of the two-dimensional Logistic chaotic system for initial values ​​x0 and y0 specifically includes:

[0043] Discrete iteration is performed using x0 and y0 as initial values ​​according to the following formula:

[0044] x n+1 =[4μ1x n (1-x n )+μ2x n y n mod 1

[0045] y n+1 =[4μ1y n (1-y n )+μ2x n y n mod 1

[0046] Where, x n ,y n Let μ1 and μ2 be the discrete iterative values ​​of the two-dimensional Logistic chaotic map, and let x be the system parameters of the two-dimensional Logistic chaotic map. n y n This is a secondary coupling term.

[0047] Furthermore, the sliding scrambling and diffusion process specifically includes:

[0048] Step C1: Perform row sliding scrambling and then column sliding scrambling on the initially encrypted image according to the following formula.

[0049] img rs [m i ,:]=roll(img[m i ,:],rs i )

[0050] img cs [:,n j ] = roll(img[:,n j ],cs j )

[0051] Among them, img rs [·]、img cs [·] represents the image after row and column sliding scrambling, respectively; roll(·) represents the sliding scrambling operation; img[·] represents the image before the scrambling operation (size m×n), m i Represents the i-th row of the image, n j rs represents the j-th column of the image. i rs represents the sliding step size for the sliding scramble in the i-th row. i∈[0,n-1], cs j cs represents the sliding step size for sliding scrambling the j-th column. j ∈[0,m-1].

[0052] Step C2: From the chaotic sequence y n Take 32×32×2 numbers from the middle, magnify and round them to the range [0, 255] and divide them into two parts, which are used as random keys K1 and K2 for diffusion encryption;

[0053] Step C3: Segment the image obtained in Step C1 by rows and concatenate them into a one-dimensional vector. First, perform a backward XOR diffusion, that is, the value of the (i+1)th pixel is obtained by XORing the value of the ith pixel with the ith element of the key K1 (obtained in Step S4). Further, perform a forward XOR diffusion on the resulting vector, that is, starting from the last pixel value, the value of the ith pixel is obtained by XORing the (i+1)th pixel with the (i+1)th element of the key K2 (obtained in Step S4).

[0054] Furthermore, the specific reconstruction process of the nonlinear deep reconstruction network includes:

[0055] Step D1: Create a dataset to train the reconstructed network model.

[0056] Step D2: The decrypted image obtained by decrypting the ciphertext image using inverse XOR diffusion and inverse sliding scrambling is used as input to reconstruct the network model (model training has been completed).

[0057] Step D3: The decrypted image undergoes convolutional and nonlinear operations through multiple convolutional layers of the reconstruction network model to obtain the reconstructed sub-image.

[0058] Example 1:

[0059] An image encryption / decryption method based on dynamic compressed sensing includes:

[0060] Step 1: Input the original image and divide it into multiple sub-images, each of which has a size of N×N (N=32);

[0061] Step 2: Use a dynamic compression network to perform convolutional compression and initial reconstruction on the sub-images to obtain a preliminary encrypted image;

[0062] Step 3: Calculate the 64-byte hash value of the sub-image using the SHA256 algorithm. Divide the hash value into 2-byte increments. Use the two numbers in the first 4 bytes as a common factor α, and divide the remaining 60 bytes equally into β1 and β2. XOR α with β1 and β2 to obtain two 2-byte numbers. Convert these numbers to decimal and divide by 255 to obtain two initial keys, which serve as the initial values ​​x0 and y0 of the two-dimensional Logistic chaotic system. Due to the perturbation sensitivity of the SHA256 algorithm, it ensures that... Figure 1 dense;

[0063] Step 4: The two-dimensional Logistic chaotic system iterates over initial values ​​x0 and y0 to generate two chaotic sequences. From the chaotic sequence x... n Extract 32 × 2 numbers, round them to the nearest integer in the range [0, 32]. The first 32 numbers are used for the initial row-sliding scrambling of the encrypted image, and the last 32 numbers are used for the initial column-sliding scrambling of the encrypted image. Then, from the chaotic sequence y... n Take 32×32×2 numbers, enlarge and round them to the range [0, 255] and divide them into two parts, which are used as random keys for diffusion encryption; after sliding scrambling and diffusion encryption, the final encrypted ciphertext image is obtained;

[0064] Step 5: Decrypt the ciphertext image by using two random sequences with inverse XOR diffusion and inverse sliding scrambling to obtain the decrypted image;

[0065] Step 6: The decrypted image is reconstructed using a non-linear deep reconstruction network to recover the sub-image;

[0066] Step 7: Piece the sub-images together in the order of the cut pieces to restore the original image.

[0067] Furthermore, the process of using a dynamic compression network to perform convolutional compression and initial reconstruction of the sub-image specifically includes:

[0068] Step 1: Use convolution kernel W com ∈R N×N×M (Without setting activation functions and biases for the convolutional layers) for sub-images X∈R N×N (N=32) Perform non-sliding convolution to achieve linear compression and obtain the observed signal.

[0069] Step 2: Use convolution kernels (Without setting activation functions and biases for the convolutional layers) for the observed signal Perform convolution operations to achieve linear initialization and reconstruction.

[0070] Furthermore, the iterative process of the two-dimensional Logistic chaotic system for initial values ​​x0 and y0 specifically includes:

[0071] Step 1: After the sub-image is processed by SHA256, a 64-byte hash value is obtained. The hash value is then divided into segments with a step size of 2, resulting in a sequence K consisting of 32 two-byte values. I ={k i |i∈[1,32],k i ∈[0,2 8 ]}.make As a public key factor, the remaining 30 k... i The sequence is divided into two subsequences. Each subsequence is XORed with a public key factor to obtain two 2-byte numbers. These numbers are then normalized by dividing by 255 to obtain x0 and y0, as shown in the following formula:

[0072]

[0073]

[0074] Step 2: Using x0 and y0 as initial values, perform discrete iterations according to the following formula:

[0075] x n+1 =[4μ1x n (1-x n )+μ2x n y n mod 1

[0076] y n+1 =[4μ1y n (1-y n )+μ2x n y n mod 1

[0077] Where, x n ,y n Let μ1 and μ2 be the discrete iterative values ​​of the two-dimensional Logistic chaotic map, and let x be the system parameters of the two-dimensional Logistic chaotic map. n y n This is a secondary coupling term.

[0078] Furthermore, the sliding scrambling and diffusion process specifically includes:

[0079] Step 1: Perform row sliding scrambling and then column sliding scrambling on the initially encrypted image according to the following formula.

[0080] img rs [m i ,:]=roll(img[m i ,:],rs i )

[0081] img cs [:,nj ] = roll(img[:,n j ],cs j )

[0082] Among them, img rs [·]、img cs [·] represents the image after row and column sliding scrambling, respectively; roll(·) represents the sliding scrambling operation; img[·] represents the image before the scrambling operation (size m×n), m i Represents the i-th row of the image, n j rs represents the j-th column of the image. i (rs i ∈[0,n-1]) represents the sliding step size for sliding scrambling in the i-th row, cs j (cs j ∈[0,m-1]) represents the sliding step size for sliding scrambling of the j-th column.

[0083] Step 2: Segment the image obtained in Step 1 by row and concatenate them into a one-dimensional vector. First, perform a backward XOR diffusion, that is, the value of the (i+1)th pixel is obtained by XORing the value of the ith pixel with the ith element of the key K1 (obtained in Step 4). Further, perform a forward XOR diffusion on the resulting vector, that is, starting from the last pixel value, the value of the ith pixel is obtained by XORing the (i+1)th pixel with the (i+1)th element of the key K2 (obtained in Step 4).

[0084] Furthermore, the specific reconstruction process of the nonlinear deep reconstruction network includes:

[0085] Step 1: Create a dataset to train the reconstructed network model.

[0086] Step 2: Decrypt the encrypted image by performing inverse XOR diffusion and inverse sliding scrambling, and use the decrypted image as input to reconstruct the network model (model training has been completed).

[0087] Step 3: The decrypted image undergoes convolutional and nonlinear operations through multiple convolutional layers of the reconstruction network model to obtain the reconstructed sub-image.

[0088] Example 2:

[0089] The dataset uses 8-bit grayscale images from the 91-images dataset as the training set for the reconstruction network, and 8-bit grayscale images from the Set14 dataset as the test set. Pixel values ​​in the images are compressed to the [0,1] interval. The nonlinear deep reconstruction network is MSR-Net. The system parameters of the two-dimensional Logistic chaotic system are set to μ1 = 0.88 and μ2 = 0.94. The training loss function for the network model is the squared loss function, the activation function is the ReLU function, the network layer weights are initialized using a Gaussian random distribution, and an adaptive learning rate (adam) is used with an initial value of 0.0001 and 300 iterations. After the dynamic compression network and the reconstruction network are trained, according to... Figure 1 and Figure 2 The forward encryption / decryption process of the network model is as follows:

[0090] Step 1: Take the test set image as the input original image and divide it into multiple sub-images, each sub-image being 32×32 in size;

[0091] Step 2: Use a dynamic compression network to perform convolutional compression and initial reconstruction on the sub-images to obtain a preliminary encrypted image;

[0092] Step 3: Calculate the 64-byte hash value of the sub-image using the SHA256 algorithm. Divide the hash value into two-byte increments. Use the two numbers in the first four bytes as a common factor α, and divide the last 60 bytes into β1 and β2. Perform an XOR operation on β1 and β2 with α to obtain two two-byte numbers. Convert them to decimal and divide by 255 to obtain two initial keys as the initial values ​​x0 and y0 of the two-dimensional Logistic chaotic system.

[0093] Step 4: The two-dimensional Logistic chaotic system iterates over initial values ​​x0 and y0 to generate two chaotic sequences. From the chaotic sequence x... n Extract 32 × 2 numbers, magnify and round them to the [0, 32] interval. Use the first 32 numbers for row sliding scrambling and the last 32 numbers for column sliding scrambling. Then, from the chaotic sequence y... n Take 32×32×2 numbers, enlarge and round them to the range [0, 255] and divide them into two parts, which are used as random keys for diffusion encryption; after sliding scrambling and diffusion encryption, the final encrypted ciphertext image can be output;

[0094] Step 5: Decrypt the ciphertext image by using two random sequences with inverse XOR diffusion and inverse sliding scrambling to obtain the decrypted image;

[0095] Step 6: Decrypt the image using a non-linear deep reconstruction network (e.g., Figure 3 Reconstruct the sub-image as shown;

[0096] Step 7: Piece the sub-images together in the order of the cut pieces to restore the original image.

[0097] The beneficial effects of this invention are:

[0098] (1) Reduced requirements for image sparsity during compression and encryption: This invention uses convolutional layers with removed activation functions and biases to compress images, thus ignoring whether the image is sparsity;

[0099] (2) By utilizing the self-learning capability of deep convolutional neural networks, the problem of low sampling rate makes it difficult to achieve high-quality reconstruction is improved: This invention obtains an image reconstruction network model that achieves nonlinear operation by training a deep convolutional neural network. This model can obtain more feature information from images with low sampling rate, thereby achieving high-quality image reconstruction.

[0100] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An image encryption / decryption method based on dynamic compressed sensing, characterized in that, The method includes the following steps: Step S1: Input the original image and divide it into multiple sub-images, each sub-image being N×N in size; Step S2: Use a dynamic compression network to perform convolutional compression and initial reconstruction on the sub-image to obtain a preliminary encrypted image; Step S3: Calculate the 64-byte hash value of the sub-image using the SHA256 algorithm. Divide the hash value into two bytes with a step size of 2 bytes. Use the two numbers in the first 4 bytes as a common factor α, and divide the last 60 bytes into β1 and β2. Perform an XOR operation on β1 and β2 with α to obtain two 2-byte numbers. After converting them to decimal, divide them by 255 to obtain two initial keys as the initial values ​​x0 and y0 of the two-dimensional Logistic chaotic system. Step S4: The two-dimensional Logistic chaotic system iterates over the initial values ​​x0 and y0 to generate two chaotic sequences x. n and y n From the chaotic sequence x n Extract 32×2 numbers, enlarge and round to the [0,32] interval. The first 32 numbers are used for the initial row sliding scrambling of the encrypted image, and the last 32 numbers are used for the initial column sliding scrambling of the encrypted image. Then from the chaotic sequence y n Take 32×32×2 numbers, magnify and round them to the [0, 255] range, and divide them into two parts to serve as the random key for diffusion encryption; After scrambling and diffusion encryption, the final encrypted ciphertext image is obtained. Step S5: Decrypt the ciphertext image by using two random sequences with inverse XOR diffusion and inverse sliding scrambling to obtain the decrypted image; Step S6: The decrypted image is reconstructed using a nonlinear deep reconstruction network to recover the sub-image; Step S7: The sub-images are stitched together in the order of the cut pieces to restore the original image.

2. The image encryption / decryption method based on dynamic compressed sensing as described in claim 1, characterized in that, N＝32。 3. The image encryption / decryption method based on dynamic compressed sensing as described in claim 1 or 2, characterized in that, The method of using a dynamic compression network to perform convolutional compression and initial reconstruction of sub-images includes: Step S21: Use convolution kernel W com ∈R N×N×M Pair of sub-images X∈R N×N The observed signal is obtained by performing non-sliding convolution operations to achieve linear compression. Step S22: Use convolution kernels For the observed signal Perform convolution operations to achieve linear initialization and reconstruction, where M is the number of channels and B... 2 For quantity.

4. The image encryption / decryption method based on dynamic compressed sensing as described in claim 3, characterized in that, Convolutional layers do not have activation functions or biases set.

5. The image encryption / decryption method based on dynamic compressed sensing as described in claim 1, characterized in that, Step S3 includes: The sub-image is processed by SHA256 to obtain a 64-byte hash value. The hash value is then divided into segments with a step size of 2, resulting in a sequence K consisting of 32 two-byte values. I ={k i |i∈[1,32],k i ∈[0,2 8 ]};make As a public key factor, the remaining 30 k i The sequence is divided into two subsequences. Each subsequence is XORed with a public key factor to obtain two 2-byte numbers. These numbers are then normalized by dividing by 255 to obtain x0 and y0, as shown in the following formula:

6. The image encryption / decryption method based on dynamic compressed sensing as described in claim 5, characterized in that, The two-dimensional Logistic chaotic system iterates over initial values ​​x0 and y0 to generate two chaotic sequences x. n and y n include: Discrete iteration is performed using x0 and y0 as initial values ​​according to the following formula: x n+1 =[4μ1x n (1-x n )+μ2x n y n [mod 1 and n+1 =[4μ1y n (1-and n )+μ2x n and n ]mod 1 Where, x n ,y n Let μ1 and μ2 be the discrete iterative values ​​of the two-dimensional Logistic chaotic map, and let x be the system parameters of the two-dimensional Logistic chaotic map. n y n This is a secondary coupling term.

7. The image encryption / decryption method based on dynamic compressed sensing as described in claim 6, characterized in that, The system parameters of the two-dimensional Logistic chaotic system are set to μ1 = 0.88 and μ2 = 0.

94.

8. The image encryption and decryption method based on dynamic compressed sensing as described in claim 6, characterized in that, The sliding scrambling and diffusion process described in S4 specifically includes: Step C1: Perform row-sliding scrambling and then column-sliding scrambling on the initially encrypted image according to the following formula; img rs [m i ,:]=roll(img[m i ,:],rs i ) img cs [:,n j ]=roll(img[:,n j ],cs j ) Among them, img rs [·]、img cs [·] represents the row and column images after sliding scrambling, respectively; roll(·) represents the sliding scrambling operation; img[·] represents the image before the scrambling operation; m i Represents the i-th row of the image, n j rs represents the j-th column of the image. i rs represents the sliding step size for the sliding scramble in the i-th row. i ∈[0,n-1], cs j cs represents the sliding step size for sliding scrambling the j-th column. j ∈[0,m-1]; Step C2: From the chaotic sequence y n Take 32×32×2 numbers from the middle, magnify and round them to the range [0, 255] and divide them into two parts, which are used as random keys K1 and K2 for diffusion encryption; Step C3: Divide the image obtained in step C1 into rows and concatenate them into a one-dimensional vector; first, perform backward XOR diffusion on it, that is, the value of the (i+1)th pixel is obtained by XORing the value of the ith pixel with the ith element of key K1; further, perform forward XOR diffusion on the result vector of backward XOR diffusion, that is, starting from the last pixel value, the value of the ith pixel is obtained by XORing the (i+1)th pixel with the (i+1)th element of key K2.

9. The image encryption / decryption method based on dynamic compressed sensing as described in claim 8, characterized in that, The reconstruction via a nonlinear deep reconstruction network includes: Step D1: Create a dataset to train the reconstructed network model; Step D2: The decrypted image obtained by decrypting the ciphertext image using inverse XOR diffusion and inverse sliding scrambling is used as the input to the reconstructed network model; Step D3: The decrypted image undergoes convolutional and nonlinear operations through multiple convolutional layers of the reconstruction network model to obtain the reconstructed sub-image.

10. The image encryption and decryption method based on dynamic compressed sensing as described in claim 9, characterized in that, MSR-Net was selected as the nonlinear deep reconstruction network.

Images