CN116743934B - An equal-resolution image hiding encryption method based on deep learning and ghost imaging - Google Patents

Abstract

The invention discloses a depth learning and ghost imaging-based multi-resolution image hiding and encrypting method, which comprises the following steps: s1, training a plurality of image samples by using a deep learning method to obtain an equal resolution image steganography model ERIH-Net based on an encoder-decoder structure; s2, taking two test images (a plaintext image and a carrier image) as input ends of a ERIH-Net model Hide-Net image hiding network, and generating a secret image (a primary ciphertext image) containing plaintext image characteristic information; s3, loading a series of Hadamard phase modulation matrix modulation light fields through a Digital Micromirror Device (DMD) to generate a series of illumination speckles; illuminating the dense-containing image by illuminating the speckle, calculating light intensity information in the spatial range of the image by using a barrel detector device BD without spatial resolution, and generating a series of light intensity values (secondary ciphertext); s4, reconstructing a dense image by using the modulated light field information and the acquired barrel detector value through a compressed sensing image reconstruction algorithm; s5, taking the reconstructed dense image as an input end of a ERIH-Net model image extraction network Extract-Net, and successfully extracting plaintext image information. According to the invention, the security and the information hiding capacity of the optical image encryption system can be improved, and the hiding information quantity of unit pixel points can reach 8 bits.

Description

An equal-resolution image hiding encryption method based on deep learning and ghost imaging

Technical Field

The present invention relates to the technical field of optical image encryption, and in particular to an equal-resolution image hiding encryption method based on deep learning and ghost imaging.

Background technique

Ghost imaging plays an important role in optical image encryption because of its advantages such as non-localized imaging, strong anti-interference and high sensitivity. Information hiding technology is also an important research direction in information security and is now widely used in copyright protection fields such as digital watermarking and optical authentication. However, the current mainstream information hiding technology has the disadvantage of small encryption capacity. For example, digital watermarking technology based on discrete wavelet transform (DWT) can usually only hide 1/4 of the size of the carrier image, or even 1/8 or less image or text information.

Summary of the invention

In view of the shortcomings in the prior art, the purpose of the present invention is to provide an equal-resolution image hiding encryption method based on deep learning and ghost imaging. Different from the previous research on applying deep learning to ghost imaging image reconstruction, deep learning is applied to the process of image hiding encryption, which can realize the hiding and extraction of equal-resolution images. After the hidden image is encrypted by computational ghost imaging technology and compressed sensing decryption, the information of the plaintext image can still be extracted from the encrypted image, indicating that the encryption scheme has good robustness, realizes the hiding encryption of equal-resolution images, and greatly improves the information hiding capacity of the hidden encryption system. In order to achieve the above-mentioned purpose and other advantages according to the present invention, a method for hiding and encrypting equal-resolution images based on deep learning and ghost imaging is provided, comprising:

S1. Design and train a deep learning image steganography model ERIH-Net for equal-resolution image hiding, which consists of two sub-networks, namely the image hiding network Hide-Net and the image extraction network Extract-Net;

S2, hide a plaintext image into a non-secret image through the image hiding network Hide-Net in the pre-trained ERIH-Net, and generate a secret image containing the feature information of the plaintext image;

S3, 4096 Hadamard matrices are loaded through the DMD to modulate the light field and generate illumination speckles, and the modulated illumination speckles are used to illuminate the secret image; the total light intensity value of the secret image is collected through the bucket detector to obtain a sequence of 4096 light intensity values, i.e., the ciphertext information;

S4, using a compressed sensing image reconstruction algorithm to reconstruct the secret image information through the ciphertext sequence and the Hadamard modulation mode (key);

S5. Use the secret image reconstructed by compressed sensing as the input of the extraction network Extract-Net to extract the initial secret image information.

Preferably, in step S1, the medium-resolution image steganography model ERIH-Net performs feature extraction and downsampling through multiple convolutional layers to obtain a high-dimensional feature map; based on the high-dimensional feature map, upsampling and feature reconstruction are performed through a deconvolution layer to obtain an imaging result with the same size as the original image.

Preferably, the plaintext image and the non-secret image input in step S2 have the same resolution, and in the experiment, both are 64×64×1 in size.

Preferably, the optical ghost imaging system in step S3 includes a He-Ne laser with a wavelength of 532.8 nm, a beam expander arranged on one side of the laser, a beam collimating lens, a digital micromirror device DMD, a focusing lens arranged on one side of the DMD, a bucket detector BD without spatial resolution capability, and a computer PC.

Preferably, in step S4, the ciphertext is received through a public channel, the key is received through a secure channel, and the compressed sensing image reconstruction algorithm uses an orthogonal matching pursuit (OMP) algorithm.

Preferably, in step S5, the Extract-Net is composed of 6 different convolutional layers, the first 5 convolutional layers are connected to the Relu activation function, and the last convolutional layer is followed by a tanh activation function.

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

(1) The deep learning-based image steganography model is applied to the camouflage encryption process in computational ghost imaging encryption. Compared with other traditional steganography methods, the image hiding process based on deep learning is simple to operate and has strong generalization ability.

(2) The equal-resolution image steganography model can realize hiding and extraction between images of the same resolution size. After the ghost image reconstruction, the extraction network of the steganography model can also extract the information of the plaintext image well, indicating that the model has good robustness.

(3) The combination of encoder-decoder based deep learning image steganography network and computational ghost imaging encryption, as well as the use of compressed sensing algorithm, can improve the information hiding capacity, security, and imaging quality of camouflaged image encryption.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a system flow chart of an equal-resolution image hiding encryption method based on deep learning and ghost imaging according to the present invention;

FIG2 is a structural diagram of the ERIH-Net of the equal-resolution image hiding encryption method based on deep learning and ghost imaging according to the present invention;

FIG3 is a ghost imaging optical path structure diagram of an equal-resolution image hiding encryption method based on deep learning and ghost imaging according to the present invention;

FIG4 is a system encryption structure diagram of an equal-resolution image hiding encryption method based on deep learning and ghost imaging according to the present invention;

FIG5 is a system decryption structure diagram of the equal-resolution image hiding encryption method based on deep learning and ghost imaging according to the present invention;

Detailed ways

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.

1-5, a method for equal-resolution image hiding and encryption based on deep learning and ghost imaging includes the following steps:

S1. Design and train a deep learning image steganography model ERIH-Net for equal-resolution image hiding;

S2, hide a plaintext image into a non-secret image through the image hiding network Hide-Net in the pre-trained ERIH-Net, and generate a secret image containing the feature information of the plaintext image;

S3, 4096 Hadamard matrices are loaded through the DMD to modulate the light field and generate illumination speckles, and the modulated illumination speckles are used to illuminate the secret image; the total light intensity value of the secret image is collected through the bucket detector to obtain a sequence of 4096 light intensity values, i.e., the ciphertext information;

S4, using a compressed sensing image reconstruction algorithm to reconstruct the secret image information through the ciphertext sequence and the Hadamard modulation mode (key);

S5. Use the secret image reconstructed by compressed sensing as the input of the extraction network Extract-Net to extract the initial secret image information.

Example 1

Step S1, firstly, an equal-resolution image steganography model ERIH-Net based on the encoder-decoder structure is designed on the pytorch deep learning framework by using the python language. The structure diagram of ERIH-Net is shown in FIG2, which includes two sub-networks, the image hiding network Hide-Net and the image extraction network Extract-Net, wherein Hide-Net includes a preprocessing convolution layer, a downsampling convolution layer, and a transposed convolution upsampling layer; and Extract-Net consists of 6 convolution layers. The preprocessing block can adjust the size of the secret image to the same resolution as the carrier image and extract the high-dimensional features of the original image; the convolution upsampling block and the transposed convolution downsampling block can perform feature fusion on the high-dimensional feature map after the two images are spliced, and finally a secret image containing secret image information is generated through the Tanh activation function. The MSE loss function is set as the loss function of the network, and the sum of the mean square error loss between the carrier image c and the secret image c' and the mean square error loss between the secret image s and the extracted secret image s' is used as the total loss function of the equal-resolution image steganography network for training iteration (β is the weight of the reconstruction error). This network uses the Adam optimizer to accelerate training. The initial learning rate lr is set to 0.001, and the training cycle is 200 rounds. After the 80th and 150th training, the learning rate will be 1/10 of the original. The batch size is set to 8, the image size is 64×64×1, the data set is 7300 Oxford-IIIT_Pets images processed by Matlab2018b software, and the test set is several Set12, Fashion Mnist and binary images, all scaled to 64×64.

Step S2, taking the secret image and carrier image of 64×64 pixels as the input of Hide-Net, calling the pre-trained model parameters to hide the secret image, and generating a carrier image containing the features of the secret image.

The optical path structure diagram of the computational ghost imaging in step S3 is shown in FIG3 , wherein the light beam emitted by the laser is expanded and collimated to illuminate the image of the transmitted object, and the light beam carrying the amplitude information of the object is transmitted through the object and then irradiated onto the DMD (Digital Micromirror Device) loaded with a series of random phase modulation matrices. After the phase information of the light field is modulated by the DMD, the reflected light intensity information is collected by the bucket detector (Bucket Detector), which is recorded as _Di. The DMD loads the random phase modulation matrix N times in sequence, and then obtains N bucket detector values, which the sender transmits to the receiver through the public channel as ciphertext, and flattens the N phase modulation matrices into N one-dimensional vectors as keys and transmits them to the receiver through the secure channel.

Step S4, after the receiver receives the ciphertext and key from the public channel and the secure channel, the corresponding light field distribution information I _i (x, y) can be calculated according to the Fresnel diffraction theorem and the random phase modulation matrix in the key. The receiver performs a second-order correlation operation on the decrypted light field information and the ciphertext information to obtain the decrypted image information. The traditional second-order correlation reconstruction algorithm has the disadvantages of low decryption efficiency and poor imaging quality in the decryption process. In order to improve the imaging quality, this paper adopts a compressed sensing optimization decryption algorithm to achieve high-quality reconstruction of secret information. The compressed sensing algorithm can break through the Nyquist sampling theorem, and the important information of the signal can be retained by compressing and sampling the image signal using the sparsity of natural images. The original signal can be restored from the compressed data by solving a minimum L1 norm optimization problem.

In step S5, the receiver uses the secret image reconstructed by the compressed sensing algorithm as the input of the extraction network to extract the secret image information reconstructed by the associated imaging.

In summary, the current mainstream computational ghost imaging camouflage image encryption research has the disadvantages of small image hiding capacity and complex image hiding process. In view of this phenomenon, the present invention analyzes various image hiding methods, such as digital watermarking, image steganography and other methods, and finds the advantages of hiding and extracting images based on deep learning image steganography models, which have the advantages of high information hiding capacity, strong generalization ability, and simple operation. Then the compressed sensing image reconstruction method is analyzed, and it is found that compared with other image reconstruction algorithms such as second-order correlation calculation, this method has better imaging quality, excellent performance, good robustness, and strong anti-interference ability, which can improve the image reconstruction effect of computational ghost imaging.

The number of devices and processing scales described here are used to simplify the description of the present invention, and the application, modification and variation of the present invention will be obvious to those skilled in the art.

Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and implementation modes. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.

Claims (1)

1. An equal resolution image encryption method based on deep learning and computational correlation imaging is characterized by comprising the following steps:

S1, designing a deep learning image steganography model ERIH-Net for hiding an equal resolution image, and carrying out feature extraction and downsampling on the equal resolution image steganography model ERIH-Net through a plurality of convolution layers to obtain a high-dimensional feature map; up-sampling and feature reconstruction are carried out through a deconvolution layer on the basis of the high-dimensional feature map, and an imaging result with the same size as the original image is obtained; specifically, an equal resolution image steganography model ERIH-Net based on an encoder and decoder structure is designed on a pytorch deep learning framework through python language, wherein the ERIH-Net structure comprises an image hiding network Hide-Net and an image extraction network Extract-Net, and the Hide-Net comprises a preprocessing convolution layer, a downsampling convolution layer and a transposed convolution upsampling layer; the Extract-Net consists of 6 convolution layers; the size of the secret image can be adjusted to be the same resolution as the carrier image through the preprocessing block, and the high-dimensional characteristics of the original image are extracted; the convolution up-sampling block and the transposition convolution down-sampling block can perform feature fusion on the high-dimensional feature images formed by splicing the two images, and finally generate a secret image containing secret image information through a Tanh activation function; setting MSE loss function as network loss function, training and iterating by using sum of beta times of mean square error loss between carrier image c and carrier image c 'and mean square error loss between secret image s and extracted secret image s' as total loss function of the equal resolution image steganographic network,

S2, training a model, namely hiding a plaintext image into a non-secret image through an image hiding network Hide-Net in a pre-training ERIH-Net to generate a secret image containing plaintext image characteristic information;

S3, loading 4096 Hadamard matrixes through the DMD for modulating a light field and generating illumination speckles, and utilizing the modulated illumination speckles to illuminate a dense image; acquiring total light intensity values of a dense image through a barrel detector to obtain 4096 light intensity value sequences, namely ciphertext information, wherein the optical ghost imaging system comprises a He-Ne laser with 532.8nm wavelength, a beam Expander arranged at one side of the laser, a beam collimating lens L, a digital micromirror device DMD, a focusing lens L arranged at one side of the DMD, a barrel detector BD without space resolution capability and a computer PC; specifically, a light beam emitted by a laser emits a transmission object image after being expanded and collimated, the light beam carrying the object amplitude information is transmitted through the object and irradiates the DMD carrying a series of random phase modulation matrixes, the phase information of a light field is modulated by the DMD, the reflected light intensity information is collected by a barrel detector and recorded as Di, the DMD sequentially loads N random phase modulation matrixes, N barrel detector values are obtained, the N barrel detector values are transmitted to a receiver through a public channel by a sender as ciphertext, and the N phase modulation matrixes are flattened into N one-dimensional vectors to be transmitted to the receiver through a safety channel as keys;

S4, reconstructing the image information containing the secret by using a compressed sensing image reconstruction algorithm through a ciphertext sequence and a Hadamard modulation mode (key), wherein the ciphertext is received through a public channel, the key is received through a secure channel, the compressed sensing image reconstruction algorithm adopts an Orthogonal Matching Pursuit (OMP) algorithm, wherein an image sparse coefficient is set to be 1.5, and if a set value is greater than 1.5, the quality of an image reconstruction result is reduced, and the image extraction network Extract can not completely recover the plaintext image information; if the setting value is smaller than 1.5, the calculated amount is greatly increased, and the imaging time is slow; after receiving ciphertext and a secret key from a public channel and a secure channel, a specific receiver can calculate corresponding light field distribution information according to the Fresnel diffraction theorem and a random phase modulation matrix in the secret key The receiver carries out second-order correlation operation on the obtained light field information and ciphertext information, and decrypted image information can be obtained;

S5, taking the compressed sensing reconstructed dense-containing image as an input end of an extraction network Extract-Net, extracting initial secret image information, wherein the Extract-Net consists of 6 different convolution layers, the first 5 convolution layers are connected with Relu activation functions, and the last convolution layer is added with a tanh activation function.