Encryption method, device and system based on neural network and correlated imaging

Abstract

The invention discloses an encryption method, device and system based on a neural network and correlated imaging, and the method comprises the steps: inputting a secret image and a carrier image into a hidden network, hiding the secret image in the carrier image through the hidden network, and obtaining a secret-carrying image; placing the secret-carrying image in a correlated imaging system, and performing modulation coding through a modulation matrix to obtain a ciphertext image; performing second-order correlation operation on the received ciphertext image and a modulation matrix of a correlated imaging system to obtain a decrypted secret-carrying image; and inputting the decrypted secret-carrying image into an extraction network for secret image extraction to obtain an original secret image. According to the method, in the encryption process, the existence of the secret image cannot be seen from the secret-carrying image at all, the hiding performance is excellent, and the security and accuracy of transmission information can be ensured. Compared with other encryption schemes, the decrypted reconstructed image is closer to the original secret image, that is, the dual encryption system still has good robustness in a noisy environment.

Description

Technical Field

[0001] The present invention belongs to the field of quantum information encryption, and in particular relates to an encryption method, device and system based on neural network and correlation imaging. Background Art

[0002] Quantum information security technology has attracted widespread attention due to its high speed, parallelism, and multi-dimensionality. The optical encryption technology derived from it is even more of a hot topic. In the field of optics, correlation imaging technology can generate images of objects on an optical path that does not contain the object, which can solve problems that are difficult to solve, such as the failure of conventional imaging. In addition, correlation imaging is easy to combine with other imaging systems, giving it a good application scenario and playing a unique role in encryption fields such as medicine, military, remote sensing, and radar. With the in-depth study of correlation imaging, correlation imaging technology has gradually matured and has been gradually applied to the field of image information encryption.

[0003] Research on optical information security technology is extremely rich and has great research significance and practical value. Traditional optical information encryption schemes have problems such as large amount of ciphertext data and high requirements for instrument precision. At the same time, there is less research, which limits the application scenarios. The correlation imaging encryption scheme solves the above problems to a certain extent. Its ciphertext is an intensity sequence, which is more conducive to popularization and application. However, as a symmetric encryption system, the existing correlation imaging encryption scheme threatens the security of information once the key is stolen. At the same time, the number of keys is huge, which is not conducive to transmission and storage. While correlation imaging is developing rapidly, artificial intelligence has developed rapidly. Since deep learning can better reflect the essential characteristics of data and has made a series of breakthroughs in image processing, natural language processing and other fields, applying it to the field of information encryption has extremely important research significance and application value. Summary of the invention

[0004] In view of the above problems, the present invention proposes an encryption method, device and system based on neural network and associative imaging, which can ensure the security and accuracy of transmitted information.

[0005] In order to achieve the above technical objectives and the above technical effects, the present invention is implemented through the following technical solutions:

[0006] In a first aspect, the present invention provides an encryption method based on a neural network and associated imaging, comprising:

[0007] The secret image and the carrier image are input into the hidden network, and the hidden network hides the secret image into the carrier image to obtain the secret image.

[0008] Placing the encrypted image in a correlation imaging system and performing modulation encoding through a modulation matrix to obtain a ciphertext image;

[0009] The received ciphertext image is subjected to a second-order correlation operation with the modulation matrix of the correlation imaging system to obtain the decrypted ciphertext image.

[0010] The decrypted secret image is input into the extraction network to extract the secret image, thereby obtaining the original secret image.

[0011] Optionally, the method for obtaining the confidential image includes:

[0012] The secret image and the carrier image are subjected to cascade convolution to retain important local information, and the secret image is obtained through encoding.

[0013] Optionally, the modulation matrix in the associated imaging system is a preset modulation key.

[0014] Optionally, the expression of the decrypted secret image is:

[0015]

[0016] D i =∫A(x,y)O(x,y)dxdy

[0017] Where N represents the number of sampling times, A(x,y) represents the preset modulation key, and <·> represents the averaging operation. <d>Represents the average value of the detector detection value, G(x,y) represents the decrypted encrypted image, D i represents the total detection value after modulation and encryption recorded by the detector for the i-th time in the correlation imaging system, as the ciphertext image; O(x,y) represents the ciphertext image, and x and y represent the horizontal and vertical coordinates of the image respectively.

[0018] Optionally, the hidden network is a SegNet deep neural network.

[0019] Optionally, the extraction network is a CNN network.

[0020] Optionally, both the hidden network and the extraction network are pre-trained network models. During the training process, the hyper parameters of the two are adjusted to minimize the difference between the carrier image and the secret image, and between the secret image and the extraction image.

[0021] Optionally, the loss function expression during network model training is:

[0022] L(H,C,S,R)=||HC||+β||SR||

[0023] Where L represents the loss function, H, C, S, and R represent the carrier image, the encrypted image, the secret image, and the extracted image, respectively, and β represents the coefficient of the equalized reconstruction error.

[0024] In a second aspect, the present invention provides an encryption device based on a neural network and associated imaging, comprising:

[0025] An image hiding unit, used for inputting a secret image and a carrier image into a hidden network, and hiding the secret image into the carrier image by the hidden network to obtain a secret image;

[0026] A modulation module, used for placing the encrypted image in a correlation imaging system, performing modulation encoding through a modulation matrix, and obtaining a ciphertext image;

[0027] The demodulation module is used to perform a second-order correlation operation on the received ciphertext image and the modulation matrix of the correlation imaging system to obtain a decrypted ciphertext image;

[0028] The image extraction module is used to input the decrypted secret image into the extraction network to extract the secret image and obtain the original secret image.

[0029] In a third aspect, the present invention provides an encryption system based on neural network and associated imaging, including a transmitting end and a receiving end;

[0030] The transmitting end is configured to perform the following steps:

[0031] The secret image and the carrier image are input into a hidden network, and the hidden network hides the secret image into the carrier image to obtain a secret image;

[0032] Placing the encrypted image in a correlation imaging system and performing modulation encoding through a modulation matrix to obtain a ciphertext image;

[0033] The receiving end is configured to perform the following steps:

[0034] The received ciphertext image is decrypted by performing a second-order correlation operation with the modulation matrix of the correlation imaging system to obtain the ciphertext image.

[0035] The secret image is input into the extraction network to extract the secret image and obtain the original secret image.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] In the encryption process of the present invention, intuitively, the existence of the secret image cannot be seen from the encrypted image at all, the concealment is extremely good, and the security and accuracy of the transmitted information can be guaranteed.

[0038] Even under noise interference, the decrypted reconstructed image in the present invention is closer to the original secret image than other encryption schemes, that is, the double encryption system can still have good robustness in a noisy environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative labor, among which:

[0040] Figure 1 A flowchart of an encryption method based on a neural network and associated imaging in an embodiment of the present invention;

[0041] Figure 2 for Figure 1 Hidden network structure diagram used in the encryption process;

[0042] Figure 3 for Figure 1 The extracted network structure diagram used in the decryption process;

[0043] Figure 4 This is a diagram showing the model training process of the SegNet deep neural network;

[0044] Figure 5 This is a diagram showing the stabilization effect of model training for the SegNet deep neural network. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present invention, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0047] Example 1

[0048] An embodiment of the present invention provides an encryption method based on a neural network and associated imaging, comprising the following steps:

[0049] (1) The secret image and the carrier image are input into the hidden network, and the hidden network hides the secret image into the carrier image to obtain the secret image;

[0050] (2) placing the encrypted image in a correlation imaging system and performing modulation encoding through a modulation matrix to obtain a ciphertext image;

[0051] (3) Performing a second-order correlation operation on the received ciphertext image and the modulation matrix of the correlation imaging system to obtain the decrypted ciphertext image;

[0052] (4) The decrypted secret image is input into an extraction network to extract the secret image, thereby obtaining the original secret image.

[0053] In a specific implementation of the embodiment of the present invention, the method for obtaining the confidential image includes:

[0054] The secret image and the carrier image are subjected to cascade convolution to retain important local information, and the secret image is obtained through encoding.

[0055] In a specific implementation of the embodiment of the present invention, the modulation matrix in the correlation imaging system is a preset modulation key.

[0056] In a specific implementation of the embodiment of the present invention, the expression of the decrypted secret image is:

[0057]

[0058] D i =∫A(x,y)O(x,y)dxdy

[0059] Where N represents the number of sampling times, A(x,y) represents the preset modulation key, and <·> represents the averaging operation. <d>Represents the average value of the detector detection value, G(x,y) represents the decrypted encrypted image, D i represents the total detection value after modulation and encryption recorded by the detector for the i-th time in the correlation imaging system, as the ciphertext image; O(x,y) represents the ciphertext image, and x and y represent the horizontal and vertical coordinates of the image respectively.

[0060] In a specific implementation of the embodiment of the present invention, the hidden network is a SegNet deep neural network.

[0061] In a specific implementation of the embodiment of the present invention, the extraction network is a CNN network.

[0062] In a specific implementation of the embodiment of the present invention, the hidden network and the extraction network are both pre-trained network models. During the training process, the hyperparameters of the two are adjusted to minimize the difference between the carrier image and the secret image, and between the secret image and the extraction image. The loss function expression during the network model training process is:

[0063] L(H,C,S,R)=||HC||+β||SR||

[0064] Where L represents the loss function, H, C, S, and R represent the carrier image, the encrypted image, the secret image, and the extracted image, respectively, and β represents the coefficient of the equalized reconstruction error.

[0065] The method in the embodiment of the present invention is described in detail below in conjunction with a specific implementation manner.

[0066] like Figure 1 As shown, an encryption method based on neural network and associated imaging in an embodiment of the present invention includes steps S1-S4. Specifically, as follows:

[0067] S1: The sender uses a hidden network structure to steganographically steganographically the secret image and the carrier image, thereby obtaining the secret image, which is first-level encryption. In the specific implementation process, the hidden network can select a SegNet deep neural network, which concatenates the secret image and the carrier image through cascade convolution, retains important local information, and obtains the secret image through encoding.

[0068] Specifically: the SegNet deep neural network is converted from the image segmentation processing field to the image encryption processing field. Figure 2 As shown, firstly, two images of size M×N (M=N) (one of which is a secret image and the other is a carrier image) are inputted, and a 2-channel feature tensor (grayscale image) is obtained through a cascaded convolution operation layer; the feature tensor is mapped through an encoder, and each encoder generates a series of feature maps through a convolution operation, followed by batch normalization, ReLU activation function and maximum pooling layer (in the processing process of the entire network, the size of the pooling window is 2×2, and the step size is set to 2), and a secret image of the same size as the original image is obtained after batch operation.

[0069] Since the hidden network and the extraction network in the present invention are designed based on the encoder-decoder structure, it is required that the carrier image of the intermediate representation is as similar as possible to the hidden image, which can be expressed by formula (1).

[0070] L(H,C,S,R)=||HC||+β||SR|| (1)

[0071] In the above formula, H, C, S, and R represent the carrier image, the secret image, the secret image, and the extracted image, respectively. β is a system used to equalize the reconstruction error. The error loss of the carrier image and the secret image is continuously adjusted through training to ensure that the secret image can be completely encoded into the carrier image.

[0072] S2: Place the encrypted image in the associated imaging system, modulate and encode it through the modulation matrix, and finally obtain the encrypted image (detector receives information) after modulation by the associated imaging system. This is secondary encryption. The modulation matrix in the imaging system is the preset modulation key. At the receiving end of the associated imaging encryption system, the detector records the total light intensity value D after modulation and encryption. i , as the ciphertext image, the calculation formula is as follows, A(x,y) represents the secondary modulation key, and O(x,y) represents the encrypted image information.

[0073] D i =∫A(x,y)O(x,y)dxdy (3)

[0074] S3: The receiving end performs a second-order correlation operation on the received ciphertext image and the preset modulation key to obtain the decrypted ciphertext image. This is the first-level decryption. The preset modulation key is the modulation matrix in the correlation imaging system, and the ciphertext image is the ciphertext image after the second-level encryption. The calculation formula is as follows:

[0075]

[0076] Where N represents the number of sampling times, A(x,y) represents the preset modulation key, and <·> represents the averaging operation. <d>Represents the average value of the detector detection value, G(x,y) represents the decrypted encrypted image, D i represents the total detection value after modulation and encryption recorded by the detector for the i-th time in the correlation imaging system, as the ciphertext image; O(x,y) represents the ciphertext image, and x and y represent the horizontal and vertical coordinates of the image respectively.

[0077] S4: The secret image is extracted through the extraction network, and the original secret image is decrypted. This is the second-level decryption. The extraction network is a CNN network, through which the secret image can be extracted.

[0078] Specifically: The network model referenced by the secret image extraction process, such as Figure 3 As shown in the figure. In the CNN network structure, the secret image is extracted through 6 convolutional layers. On the CNN network, the Dropout operation, activation function and pooling layer are used to enhance the nonlinear learning ability of the network. The fitting parameters are learned by using nonlinear features to understand the weighted parameters of each layer of the network to adapt to the mapping between input and output.

[0079] In the extraction network, the filter size of each convolution layer is designed to be 3×3, and each convolution layer is followed by a ReLU activation function and batch normalization operation. On the left and right sides of the network, 3×3 convolutions are used to map the feature vectors of 64 components to the required number of categories, and the Sigmoid activation function is used to calculate the secret image and the carrier image. The stride size in the network is set to 1 and the padding is set to 1 to ensure that the size of the image remains unchanged during the convolution process.

[0080] Research and investigation show that the image data for testing and training is from the ImageNet database. 55,000 images were extracted from it (50,000 for training and 5,000 for testing). The initial learning rate of the entire network was set to 0.001, and the hyperparameter β was set to 0.75. The Adam optimization method was used to automatically adjust the learning rate so that the network parameters can be learned smoothly. The number of training images per batch was set to 8, and the network was trained for 300 iterations. Figure 4 The results of image hiding for three iterations of Epoch0, Epoch50 and Epoch100 are shown. From this figure, we can see that the initial training effect starts with noise, and the training effect will get better and better as the number of training iterations increases. When Epoch is 50, we can see that the noise is greatly reduced. When iterating to Epoch100, the human eye cannot see the noise, and the effect of HVS is difficult to distinguish compared with the original image. After 100 data iterations, the network model at this time tends to be stable. Figure 5 This shows the training effect after the model becomes robust. Intuitively, the secret image is very similar to the carrier image, and the existence of the secret image cannot be seen at all. From this figure, we can see that the encryption and extraction effects of the image are very good.

[0081] In summary, in the dual encryption method of the base neural network and associated imaging provided in the embodiment of the present invention, based on the hidden network, two images are cascaded through convolution to retain important local information and are encoded to achieve primary encryption of the image; the modulation encryption method of the present invention belongs to intensity modulation, and does not modulate the phase, frequency, polarization, amplitude, etc. of the carrier light source, and secondary encryption is achieved through a preset modulation key; the extraction of the secret image is achieved based on the extraction network. Associated imaging does not require accurate sampling of each information at the receiving end, and only needs to receive the energy value of the transmitted information to reconstruct the transmitted information, which has good anti-interference performance.

[0082] Example 2

[0083] An embodiment of the present invention provides an encryption device based on a neural network and associated imaging, comprising:

[0084] An image hiding unit, used for inputting a secret image and a carrier image into a hidden network, and hiding the secret image into the carrier image by the hidden network to obtain a secret image;

[0085] A modulation module, used for placing the encrypted image in a correlation imaging system, performing modulation encoding through a modulation matrix, and obtaining a ciphertext image;

[0086] The demodulation module is used to perform a second-order correlation operation on the received ciphertext image and the modulation matrix of the correlation imaging system to obtain a decrypted ciphertext image;

[0087] The image extraction module is used to input the decrypted secret image into the extraction network to extract the secret image and obtain the original secret image.

[0088] The rest are the same as in Example 1.

[0089] Example 3

[0090] An embodiment of the present invention provides an encryption system based on a neural network and associated imaging, including a transmitting end and a receiving end;

[0091] The transmitting end is configured to perform the following steps:

[0092] The secret image and the carrier image are input into a hidden network, and the hidden network hides the secret image into the carrier image to obtain a secret image;

[0093] Placing the encrypted image in a correlation imaging system and performing modulation encoding through a modulation matrix to obtain a ciphertext image;

[0094] The receiving end is configured to perform the following steps:

[0095] The received ciphertext image is decrypted by performing a second-order correlation operation with the modulation matrix of the correlation imaging system to obtain the ciphertext image.

[0096] The secret image is input into the extraction network to extract the secret image and obtain the original secret image.

[0097] The rest are the same as in Example 1.

[0098] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0099] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0100] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0101] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0102] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

[0103] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.< / d> < / d> < / d>

Claims

1. An encryption method based on neural network and associative imaging, characterized in that: include: The secret image and the carrier image are input into the hidden network, and the hidden network hides the secret image into the carrier image to obtain the secret image. The secret image is placed in a correlation imaging system, and modulation coding is performed through a modulation matrix to obtain a ciphertext image; the received ciphertext image and the modulation matrix of the correlation imaging system are subjected to a second-order correlation operation to obtain a decrypted secret image; The decrypted secret image is input into the extraction network to extract the secret image, thereby obtaining the original secret image.

2. The encryption method based on neural network and associative imaging according to claim 1, characterized in that: The method for obtaining the confidential image comprises: The secret image and the carrier image are subjected to cascade convolution to retain important local information, and the secret image is obtained through encoding.

3. The encryption method based on neural network and associative imaging according to claim 1, characterized in that: The modulation matrix in the associated imaging system is a preset modulation key.

4. The encryption method based on neural network and associative imaging according to claim 3, characterized in that: The expression of the decrypted secret image is: D i =∫A(x,y)O(x,y)dxdy Where N represents the number of sampling times, A(x,y) represents the preset modulation key, and <·> represents the averaging operation. <d>Represents the average value of the detector detection value, G(x,y) represents the decrypted encrypted image, D i represents the total detection value after modulation and encryption recorded by the detector for the i-th time in the correlation imaging system, as the ciphertext image; O(x,y) represents the ciphertext image, and x and y represent the horizontal and vertical coordinates of the image respectively.< / d> 5. The encryption method based on neural network and associative imaging according to claim 1, characterized in that: The hidden network is a SegNet deep neural network.

6. The encryption method based on neural network and associative imaging according to claim 1, characterized in that: The extraction network is a CNN network.

7. The encryption method based on neural network and associative imaging according to claim 1, characterized in that: The hidden network and the extraction network are both pre-trained network models. During the training process, the hyper parameters of the two are adjusted to minimize the difference between the carrier image and the secret image, and between the secret image and the extraction image.

8. The encryption method based on neural network and associative imaging according to claim 7, characterized in that: The loss function expression during network model training is: L(H,C,S,R)=||HC||+β||SR|| Where L represents the loss function, H, C, S, and R represent the carrier image, the encrypted image, the secret image, and the extracted image, respectively, and β represents the coefficient of the equalized reconstruction error.

9. An encryption device based on neural network and associative imaging, characterized in that: include: An image hiding unit, used for inputting a secret image and a carrier image into a hidden network, and hiding the secret image into the carrier image by the hidden network to obtain a secret image; A modulation module, used for placing the encrypted image in a correlation imaging system, performing modulation encoding through a modulation matrix, and obtaining a ciphertext image; The demodulation module is used to perform a second-order correlation operation on the received ciphertext image and the modulation matrix of the correlation imaging system to obtain a decrypted ciphertext image; The image extraction module is used to input the decrypted secret image into the extraction network to extract the secret image and obtain the original secret image.

10. An encryption system based on neural network and associative imaging, characterized in that: Includes the sending end and the receiving end; The transmitting end is configured to perform the following steps: The secret image and the carrier image are input into a hidden network, and the hidden network hides the secret image into the carrier image to obtain a secret image; Placing the encrypted image in a correlation imaging system and performing modulation encoding through a modulation matrix to obtain a ciphertext image; The receiving end is configured to perform the following steps: The received ciphertext image is decrypted by performing a second-order correlation operation with the modulation matrix of the correlation imaging system to obtain the ciphertext image. The secret image is input into the extraction network to extract the secret image and obtain the original secret image.

Images