Controllable image steganography method and system based on steganography diffusion model
By combining three-dimensional memristor chaotic mapping and steganalysis diffusion model, the contradiction between visual quality and security in existing image steganography techniques is resolved, realizing a highly secure and practical image steganography method.
Patent Information
- Application Number
- CN202511739801.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-25
- Publication Date
- 2026-02-17
AI Technical Summary
Existing image steganography techniques struggle to improve security without compromising the visual quality of the carrier image, and are easily detected by steganalysis tools.
A steganalysis diffusion model based on three-dimensional memristor chaotic mapping is adopted. The color plaintext image is scrambled and diffused by generating a highly random sequence, and the ciphertext image is embedded in the carrier image. The steganalysis diffusion model is then used for denoising.
Without affecting the visual quality of the carrier image, it improves the security and practicality of image steganography, reduces the risk of being identified by steganography analysis technology, and achieves stable embedding and accurate extraction of secret information.
Smart Images
Figure CN121547539A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of image information security, in particular to a controllable image steganography method and system based on a steganography diffusion model. BACKGROUND
[0002] With the popularization of computer technology and the Internet, digital images have become one of the core carriers of information transmission and storage. In this process, a large amount of image data containing highly sensitive information, such as medical images, national defense and military secret images, etc. Once such images are illegally stolen or accessed without authorization in the transmission or storage link, it will directly lead to the leakage of users' key privacy, and even may pose a serious threat to national security. Therefore, how to protect the security and concealment of image information, especially the sensitive content contained therein, has become a key technical problem to be solved in the field.
[0003] Image steganography technology, as an important means of image encryption, aims to hide the existence of secret information (such as another image) by embedding it into a public carrier image, thereby avoiding attracting the attention of attackers. However, the current mainstream image steganography technology has the following significant defects: 1) visual distortion, most of these technologies achieve information hiding by directly modifying the pixel content of the carrier image (such as least significant bit replacement). This modification operation is extremely easy to introduce obvious visual artifacts or distortion, which destroys the original visual quality and natural characteristics of the carrier image, and reduces its reliability and usability as a "cover"; 2) statistical feature exposure, more importantly, the modification of the carrier image content inevitably changes its inherent statistical rules, such as the distribution characteristics of pixel values, frequency energy distribution, etc. These disturbed statistical properties become effective detection basis for steganalysis technology (technology aimed at detecting the existence of hidden information). The statistical anomalies generated by existing steganography methods make the existence of secret information easily detected by advanced steganalysis tools, which weakens the security of steganography.
[0004] In summary, the existing image steganography technology, when implementing covert communication, is difficult to balance the visual quality maintenance of the carrier image and the statistical security against steganalysis at the same time, which greatly limits its practical application value in high security requirement scenarios. SUMMARY
[0005] The present application aims to overcome at least one of the defects in the prior art, and proposes a controllable image steganography method and system based on a steganography diffusion model. The present application can controllably embed secret information into the carrier image without destroying the original characteristics of the carrier image, ensuring that the visual quality of the carrier image is not significantly affected, and reducing the risk of being identified by steganalysis technology. At the same time, it realizes the stable embedding and accurate extraction of secret information, and improves the security and practicality of image steganography.
[0006] To achieve the above object, in a first aspect, the present application provides a controllable image steganography method based on steganographic diffusion model, the method comprising the following steps: Obtaining the hash value of the color plaintext image to be encrypted, and generating the parameters and initial value of the three-dimensional memristor chaotic mapping according to the hash value; Driving the three-dimensional memristor chaotic mapping with the parameters and initial value to iteratively generate a plurality of random sequences; Decomposing the color plaintext image to be encrypted into R, G, and B channel matrices, and performing scrambling and diffusion operations on the channel matrices based on the random sequences to generate a color ciphertext image; Selecting a carrier image and adding noise to the carrier image to obtain a preprocessed carrier image; Respectively performing discrete wavelet transform on the R, G, and B channels of the preprocessed carrier image to obtain a plurality of wavelet coefficient matrices corresponding to the channels; Decomposing the pixel values of each channel of the color ciphertext image by digits, and embedding the decomposed digit values into specified decimal places of the wavelet coefficient matrices to generate a ciphertext-containing wavelet coefficient matrix; Performing inverse discrete wavelet transform on the ciphertext-containing wavelet coefficient matrix to obtain a noisy ciphertext-containing carrier image; Using the steganographic diffusion model to denoise the noisy ciphertext-containing carrier image to generate a final ciphertext-containing carrier image.
[0007] Further, the three-dimensional memristor chaotic mapping is obtained by introducing a memristor into a two-dimensional chaotic mapping, and the three-dimensional memristor chaotic mapping The mathematical model of the chaotic mapping is as follows:
[0008] wherein, , , , , are constants, is the parameter of the three-dimensional memristor chaotic mapping.
[0009] Further, the generation of the parameters and initial value of the three-dimensional memristor chaotic mapping according to the hash value specifically comprises: Using a hash function to calculate the hash value of the color plaintext image to be encrypted, dividing the hash value into a plurality of groups according to a predetermined number of digits, and converting each group of hash values after grouping into a decimal number; Performing XOR calculation on the obtained decimal numbers to obtain the parameters of the three-dimensional memristor chaotic mapping; Based on the parameters of the three-dimensional memristor chaotic mapping, calculating the initial value of the three-dimensional memristor chaotic mapping.
[0010] Further, the three-dimensional memristor chaotic mapping is driven by the parameter and the initial value, and a plurality of random sequences are iteratively generated , including: substituting the initial value into the three-dimensional memristor chaotic mapping, iteratively calculating a new initial value until an iteration threshold is reached, and outputting a plurality of random sequences.
[0011] Further, the color plaintext image to be encrypted is decomposed into R, G, and B channel matrices, and a scrambling and diffusion operation is performed on the channel matrices based on the random sequence to generate a color ciphertext image, specifically including: The color plaintext image to be encrypted is decomposed into R, G, and B channel matrices, and the three channel matrices are spliced, to obtain a spliced matrix; The random sequence is quantized, integerized, and range-constrained to obtain a first sequence for scrambling and a second sequence for XOR operation; The spliced matrix is expanded into a one-dimensional sequence, the first sequence is used to scramble the one-dimensional sequence, and a third sequence is obtained; The elements in the third sequence are XORed with the elements in the second sequence to obtain a fourth sequence, and a restoration operation is performed on the fourth sequence to obtain a restored matrix; The R, G, and B channel matrices are separated from the restored matrix, , and the three channel matrices are combined to obtain a color ciphertext image.
[0012] Further, a carrier image is selected, and noise is added to the carrier image to obtain a preprocessed carrier image, including: adding Gaussian noise to the carrier image to obtain a preprocessed carrier image.
[0013] Further, the pixel values of each channel of the color ciphertext image are decomposed by digits, and the decomposed digit values are embedded into specified decimal places of the wavelet coefficient matrix to generate a secret-containing wavelet coefficient matrix, specifically including: The pixel values of each channel of the color ciphertext image are decomposed by digits, and the decomposed digit values are embedded into specified decimal places of the wavelet coefficient matrix to generate a secret-containing wavelet coefficient matrix, specifically including: The decomposed digit values are embedded into the specified decimal places of the wavelet coefficient matrix to generate a secret-containing wavelet coefficient matrix.
[0014] Further, the noisy secret-containing carrier image is denoised using a steganographic diffusion model to generate a final secret-containing carrier image, including: using a steganographic diffusion model to denoise the noisy secret-containing carrier image to obtain a secret-containing carrier image.
[0015] In a second aspect, the present application also provides a stego diffusion model-based controllable image steganography system, which is based on the method of the first aspect, and comprises the following modules: An obtaining module is configured to obtain a hash value of a color plaintext image to be encrypted, and generate parameters and initial values of a three-dimensional memristor chaotic map according to the hash value; A first generating module is configured to drive the three-dimensional memristor chaotic map with the parameters and initial values, and iteratively generate a plurality of random sequences; A second generating module is configured to decompose the color plaintext image to be encrypted into three channel matrices of R, G and B, and perform scrambling and diffusion operations on the channel matrices based on the random sequences to generate a color ciphertext image; A noise adding module is configured to select a carrier image, and add noise to the carrier image to obtain a preprocessed carrier image; A discrete wavelet transform module is configured to perform discrete wavelet transform on the R, G and B channels of the preprocessed carrier image respectively to obtain a plurality of wavelet coefficient matrices corresponding to the channels; A third generating module is configured to decompose pixel values of each channel of the color ciphertext image by digits, and embed the decomposed digital values into specified digits of the wavelet coefficient matrices to generate a ciphertext-containing wavelet coefficient matrix; An inverse discrete wavelet transform is configured to perform inverse discrete wavelet transform on the ciphertext-containing wavelet coefficient matrix to obtain a noisy ciphertext-containing carrier image; A denoising module is configured to denoise the noisy ciphertext-containing carrier image by using a stego diffusion model to generate a final ciphertext-containing carrier image.
[0016] In a third aspect, the present application also provides a computer readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method of the first aspect.
[0017] Compared with the prior art, the present application has the following beneficial effects: The present application uses a three-dimensional memristor chaotic map to iteratively generate random sequences, which have higher randomness than two-dimensional memristor chaotic maps. Using the random sequences with higher randomness for color plaintext image encryption can effectively improve the security of image steganography. By controllably embedding a ciphertext image into a noisy carrier image, the visual quality of the carrier image is not significantly affected, and the risk of being identified by steganalysis technology is reduced. At the same time, stable embedding and accurate extraction of secret information are realized, and the security and practicality of image steganography are improved. BRIEF DESCRIPTION OF DRAWINGS
[0018] Figure 1A flow chart of a controllable image steganography method based on a steganographic diffusion model for Embodiment 1 of the present application; Figure 2 A block diagram of a controllable image steganography system based on a steganographic diffusion model for Embodiment 3 of the present application; Figure 3 A color plaintext image to be encrypted for Embodiment 2 of the present application, wherein, Figure 3 a is a color plaintext Peppers image to be encrypted; Figure 3 b is a color plaintext House image to be encrypted; Figure 3 c is a color plaintext Baboon image to be encrypted; Figure 3 d is a color plaintext Xiaoqiao image to be encrypted; Figure 4 A ciphertext image for Embodiment 2 of the present application, wherein, Figure 4 a is a ciphertext Peppers image; Figure 4 b is a ciphertext House image; Figure 4 c is a ciphertext Baboon image; Figure 4 d is a ciphertext Xiaoqiao image; Figure 5 A carrier image for Embodiment 2 of the present application, wherein, Figure 5 a is a carrier image for encrypting a color plaintext Peppers image; Figure 5 b is a carrier image for encrypting a color plaintext House image; Figure 5 c is a carrier image for encrypting a color plaintext Baboon image; Figure 5 d is a carrier image for encrypting a color plaintext Xiaoqiao image; Figure 6 A ciphertext-containing carrier image for Embodiment 2 of the present application, wherein, Figure 6 a is a ciphertext-containing carrier image for encrypting a color plaintext Peppers image; Figure 6 b is a ciphertext-containing carrier image for encrypting a color plaintext House image; Figure 6 c is a ciphertext-containing carrier image for encrypting a color plaintext Baboon image; Figure 6d is the recovered image of the ciphertext Xiaoqiao image; Figure 7 is the recovered image of the ciphertext Xiaoqiao image of the present application embodiment 2, wherein, Figure 7 a is the recovered image of the ciphertext Xiaoqiao image of the present application embodiment 2, wherein, Figure 7 b is the recovered image of the ciphertext Xiaoqiao image of the present application embodiment 2, wherein, Figure 7 c is the recovered image of the ciphertext Xiaoqiao image of the present application embodiment 2, wherein, Figure 7 d is the recovered image of the ciphertext Xiaoqiao image of the present application embodiment 2, wherein, Figure 8 is the process diagram of adding noise to the carrier image of the present application embodiment 2, wherein, Figure 8 a is the process diagram of adding Gaussian noise to the carrier image of the present application embodiment 2, wherein, b is the process diagram of adding Gaussian noise to the carrier image of the present application embodiment 2, wherein, Figure 8 c is the process diagram of adding Gaussian noise to the carrier image of the present application embodiment 2, wherein, Figure 8 d is the process diagram of adding Gaussian noise to the carrier image of the present application embodiment 2, wherein, e is the process diagram of adding Gaussian noise to the carrier image of the present application embodiment 2, wherein, Figure 8 f is the process diagram of adding Gaussian noise to the carrier image of the present application embodiment 2, wherein, Figure 8 j is the process diagram of adding Gaussian noise to the carrier image of the present application embodiment 2, wherein, h is the process diagram of adding Gaussian noise to the carrier image of the present application embodiment 2, wherein, Figure 8 i is the process diagram of adding Gaussian noise to the carrier image of the present application embodiment 2, wherein, Figure 8 is the process diagram of removing noise of the present application embodiment 2, wherein, a is the process diagram of removing Gaussian noise of the present application embodiment 2, wherein, Figure 8 b is the process diagram of removing Gaussian noise of the present application embodiment 2, wherein, Figure 8 c is the process diagram of removing Gaussian noise of the present application embodiment 2, wherein, d is the process diagram of removing Gaussian noise of the present application embodiment 2, wherein, Figure 9 e is the process diagram of removing Gaussian noise of the present application embodiment 2, wherein, Figure 9 f is the process diagram of removing Gaussian noise of the present application embodiment 2, wherein, Figure 9 j is the process diagram of removing Gaussian noise of the present application embodiment 2, wherein, h is the process diagram of removing Gaussian noise of the present application embodiment 2, wherein, Figure 9 i is the process diagram of removing Gaussian noise of the present application embodiment 2, wherein, Figure 9 d is a Gaussian noise map removed at step Figure 9 e is a Gaussian noise map removed at step Figure 9 f is a Gaussian noise map removed at step Figure 9 g is a Gaussian noise map removed at step Figure 9 h is a Gaussian noise map removed at step Figure 9 i is a Gaussian noise map removed at step DETAILED DESCRIPTION
[0019] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and not to limit the present application.
[0020] Example 1 Referring to Figure 1 , the embodiment of the present application preferably comprises the following steps: S1: obtaining a hash value of a color plaintext image to be encrypted, and generating parameters and initial values of a three-dimensional memristive chaotic mapping according to the hash value; S2: driving the three-dimensional memristive chaotic mapping using the parameters and initial values to iteratively generate a plurality of random sequences; S3: decomposing the color plaintext image to be encrypted into R, G and B channel matrices, and performing scrambling and diffusion operations on the channel matrices based on the random sequences to generate a color ciphertext image; S4: selecting a carrier image, and adding noise to the carrier image to obtain a preprocessed carrier image; S5: respectively performing discrete wavelet transform on R, G and B channels of the preprocessed carrier image to obtain a plurality of wavelet coefficient matrices corresponding to the channels; S6: decomposing pixel values of each channel of the color ciphertext image by digits, and embedding the decomposed digit values into specified decimal digits of the wavelet coefficient matrices to generate a ciphertext-containing wavelet coefficient matrix; S7: performing inverse discrete wavelet transform on the ciphertext-containing wavelet coefficient matrix to obtain a noisy ciphertext-containing carrier image; S8: using a steganographic diffusion model to denoise the noisy ciphertext-containing carrier image to generate a final ciphertext-containing carrier image.
[0021] The embodiment uses a three-dimensional memristor chaotic mapping to iteratively generate a random sequence, which has higher randomness than a two-dimensional memristor chaotic mapping. The random sequence with higher randomness is used for color plaintext image encryption, which can effectively improve the security of image steganography. The encrypted image is controllably embedded in a noisy carrier image, which ensures that the visual quality of the carrier image is not significantly affected and reduces the risk of being identified by steganalysis technology. At the same time, the secret information is stably embedded and accurately extracted, which improves the security and practicability of image steganography.
[0022] In an optional embodiment, in step S1, the three-dimensional memristor chaotic mapping is obtained by introducing a memristor into a two-dimensional chaotic mapping. It should be noted that the two-dimensional chaotic mapping refers to a discrete-time chaotic system described by two state variables (for example, x and y). Its mathematical form is usually a set of iterative equations, for example:
[0023] Such chaotic mappings are often used to generate pseudo-random sequences as key streams or for shuffling (shuffling pixel positions) and diffusion (changing pixel values). The memristor, which stands for "memory resistor", is the fourth basic circuit element. Its main feature is that its resistance value is not fixed, but is determined by the history of the charge or magnetic flux flowing through it. In the embodiment, the memristor is introduced into the two-dimensional chaotic mapping. A new variable z representing the state of the memristor is added to the mathematical model (iterative equation) of the original two-dimensional chaotic mapping, and the mathematical model of the obtained three-dimensional memristor chaotic mapping is represented as follows:
[0024] wherein a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, p, q, r, s, t, u, v, w, x, y, z are constants, , , , , is a parameter of the three-dimensional memristor chaotic mapping. Optionally, , , , , The embodiment adds a memristor to the traditional two-dimensional chaotic mapping, which is conducive to increasing the complex nonlinearity of the three-dimensional chaotic mapping and laying a foundation for generating a random sequence with higher randomness. After obtaining the three-dimensional chaotic mapping, the hash value of the color plaintext image to be encrypted is obtained, and the parameters and initial values of the three-dimensional memristor chaotic mapping are generated according to the hash value. This process specifically includes: 1) using a hash function to calculate the hash value of the color plaintext image to be encrypted, dividing the hash value into multiple groups by a predetermined number of digits, and converting each group of hash values after grouping into a decimal number; it should be noted that the hash function includes SHA-256, SHA-3, etc., and the present embodiment does not limit which function is used to process the color plaintext image, for example, using the hash function SHA-256 to calculate the hash value of the color plaintext image The hash value is 256 bits, and then the hash value is divided into 8 groups, each group being 32 bits, and the 8 groups of hash values are converted into corresponding decimals, denoted as .
[0025] 2) XOR calculation is performed on the obtained decimal numbers to obtain the parameters of the three-dimensional memristor chaotic mapping; for example, the decimal calculated above is grouped and XOR calculated, and a total of 4 parameters are obtained, adjacent parameters h share part of the decimal number , and through XOR calculation, even if the attacker knows part of the decimal number , the complete hash value cannot be inferred, which can provide higher security, and the XOR calculation formula is as follows:
[0026] wherein represents a bitwise XOR operation.
[0027] 3) Based on the parameters of the three-dimensional memristor chaotic mapping, the initial value of the three-dimensional memristor chaotic mapping is calculated. For example, according to the 4 parameters calculated by the XOR calculation above, the initial values of the three-dimensional chaotic mapping are calculated , and , and the calculation formula is as follows:
[0028] wherein represents a modulo operation. In the present embodiment, each initial value contains 3 parameters , which can increase the nonlinearity of the initial value.
[0029] In an optional embodiment, in step S2, the initial value is substituted into the three-dimensional memristor chaotic mapping, a new initial value is iteratively calculated until the iteration threshold is reached, and a plurality of random sequences are output. In the present embodiment, the iteration threshold is set to 1000, and for example, the initial values , and Substituting into the three-dimensional memristor chaotic mapping, the results of the first 1000 iterations are iterated and discarded to eliminate transient effects, generating three sets of lengths. random sequence , and In this embodiment, the initial evolution stage of the chaotic system is a transient process (an unstable chaotic state), and the output sequence is correlated. The first 1000 results are discarded to ensure that the system enters a stable attractor and generates a chaotic sequence with acceptable statistical properties.
[0030] In an optional embodiment, step S3 specifically includes: S3.1: Decompose the color plaintext image to be encrypted into three channel matrices: R, G, and B, and then concatenate the three channel matrices. The matrix is concatenated to form a matrix; for example, the matrix to be encrypted is... The color plaintext image P is decomposed into three channel matrices: R, G, and B. The size of each of the three channel matrices is [missing information]. Then, these three channel matrices are concatenated to obtain the concatenated matrix. splicing matrix Size is This step converts 3D image data into a 2D operation plane, facilitating unified encryption processing.
[0031] S3.2: Quantize, integerize, and range-constrain the random sequence to obtain a first sequence for scrambling and a second sequence for XOR operation; for example, the aforementioned chaotic sequence... The initial first sequence D is obtained by performing the following processing:
[0032] in, To obtain the absolute value, i.e., quantization, chaotic sequences often contain decimals and negative values. Taking the absolute value ensures the integrity of the chaotic values.
[0033] It is a positive number. To amplify and round down, that is, to convert to integers. This represents the modulo operation on integer results, mapped to [0, m × 3n]. Within the range of [1]. Where m×3n corresponds to the total number of pixels in the color image (m×n is the image size, and 3 represents the three RGB channels). Then, the initial first sequence is sorted to obtain the first sequence used for scrambling. Then, the aforementioned chaotic sequence... Quantization, integerization, and range constraint processing are also performed, as follows:
[0034] Where each type of operator is similar to step S3.2, to avoid repetition, not here again. Finally, the second sequence for XOR operation is obtained ; S3.3: The spliced matrix is expanded into a one-dimensional sequence, and the first sequence is used to scramble the one-dimensional sequence to obtain a third sequence; for example, the aforementioned spliced matrix is expanded into a one-dimensional sequence, and the first sequence is used to scramble the one-dimensional sequence to obtain a third sequence , and the specific scrambling operation is as follows:
[0035] S3.4: The elements in the third sequence are XOR calculated with the elements in the second sequence to obtain a fourth sequence, and the fourth sequence is restored to obtain a restored matrix; for example, the elements in the aforementioned third sequence are XOR calculated with the elements in the second sequence to obtain a fourth sequence , and the specific operation is as follows:
[0036] The fourth sequence is restored to obtain a restored matrix of size .
[0037] S3.5: Separate , and three channel matrices from the restored matrix, and combine the three channel matrices to obtain a color ciphertext image. For example, separate , , and three channel matrices from the aforementioned restored matrix , and combine the three channel matrices to obtain a color ciphertext image .
[0038] The embodiment can destroy the spatial correlation of the color plaintext image through the scrambling operation, which is beneficial to resist cropping / compression attacks; at the same time, through the diffusion operation, the statistical characteristics are destroyed, which is beneficial to resist frequency analysis, thereby improving the security and reliability of image steganography.
[0039] In an optional embodiment, in step S4, specifically comprising: adding Gaussian noise of step to the carrier image to obtain a preprocessed carrier image. For example, a Gaussian noise of size carrier image For carrier images Add to The preprocessed carrier image is obtained by stepping through Gaussian noise. ,as follows:
[0040] in, , , , , , , ~N(0,1) is noise that follows a Gaussian distribution. By adding Gaussian noise to the carrier image, the preprocessed carrier image is made to have a visually imperceptible but statistically detectable noise floor, while providing an ideal masking environment for subsequent ciphertext embedding in the wavelet domain.
[0041] In an optional embodiment, in step S5, discrete wavelet transforms are performed on the R, G, and B channels of the preprocessed carrier image to obtain multiple wavelet coefficient matrices for the corresponding channels; for example, the aforementioned preprocessed carrier image... Discrete wavelet transforms were performed on the R, G, and B channels respectively, yielding results of magnitudes of... wavelet coefficient matrix , , , , , , , , , , , .
[0042] In an optional embodiment, step S6 specifically includes: S6.1: Perform digit decomposition on the pixel values of each channel of the color encrypted image according to the units, tens, and hundreds digits; for example, the aforementioned obtained color encrypted image... Three channels red , green , blue Extract the units, tens, and hundreds digits separately to obtain a matrix. , , , , , , , , .
[0043] S6.2: Embed the decomposed digit values into a specified decimal place of the wavelet coefficient matrix to generate a dense wavelet coefficient matrix. In this embodiment, the specified decimal place is the second decimal place. For example, a color ciphertext image... The units, tens, and hundreds digits of the corresponding channel are embedded in the image respectively. The dense wavelet coefficient matrix is obtained by taking the second decimal place of the coefficients. , , , , , , , , ,as follows:
[0044]
[0045]
[0046]
[0047]
[0048]
[0049]
[0050]
[0051] .
[0052] In this embodiment, the decomposed digit values are embedded into the second decimal place of the wavelet coefficient matrix. The second decimal place is the least significant decimal place. After modification, it will not have a significant impact on the image visual quality, which is beneficial to improving the image visual quality.
[0053] In an optional embodiment, in step S7, an inverse discrete wavelet transform is performed on the dense wavelet coefficient matrix to obtain a noisy and dense carrier image; for example, the dense wavelet coefficients are... , , , , , , , , , , , Inverse discrete wavelet transform is performed to obtain noisy and dense images. .
[0054] In an optional embodiment, in step S8, the noisy and dense carrier image is processed using a stegoDiffusion model. After denoising, a dense carrier image is obtained.
[0055] In this embodiment, a random sequence is generated iteratively using a three-dimensional memristor chaotic map, which has higher randomness than a two-dimensional memristor chaotic map. Using this random sequence with higher randomness for color plaintext image encryption can effectively improve the security of image steganography. Furthermore, by controllably embedding the ciphertext image into a noisy carrier image, the visual quality of the carrier image is not significantly affected, and the risk of being identified by steganography analysis technology is reduced. At the same time, stable embedding and accurate extraction of secret information are achieved, improving the security and practicality of image steganography.
[0056] Example 2 Please see Figures 3-9 Based on the method proposed in Example 1, a simulation experiment was conducted in this embodiment of the invention. Figure 3 As shown, from left to right, the color plaintext images to be encrypted are (a) Peppers (64×64×3), (b) House (64×64×3), (c) Baboon (64×64×3), and (d) Xiaoqiao (64×64×3). Using the method proposed in Example 1, the plaintext images are encrypted... Figure 3 The color plaintext images to be encrypted (a) Peppers, (b) House, (c) Baboon and (d) Xiaoqiao are encrypted to obtain Figure 4 The encrypted image shown; Figure 5 The carrier image is randomly selected from the database, with a size of 128×128×3; Figure 6 The image is a dense carrier image generated using the method proposed in Example 1, with a size of 128×128×3. Figure 7 To recover the images of (a) Peppers, (b) House, (c) Baboon and (d) Xiaoqiao.
[0057] Furthermore, this embodiment of the invention utilizes the reversibility of the StegoDiffusion model to achieve image steganography and enables a certain degree of control over the content of the dense carrier image. When StegoDiffusion adds noise to a real image, it tends towards Gaussian noise, ultimately becoming completely Gaussian noise. Therefore, the sampled image has randomness. To gain some control over the content of the sampled image, this embodiment adds a certain degree of noise to the real image, so that the final noisy image still retains some information from the real image. Therefore, when generating the sampled image, this information from the real image guides the model to generate an image with a certain similarity to the real image, thereby achieving a certain degree of control over the generated content.
[0058] Assume the total number of iterations from the original data to pure Gaussian noise is... The noise-adding process only adds Gaussian noise occurs when an image contains both noise and some information from the real image, such as... Figure 8 As shown in (i). At this point, the secret information is embedded into the image, and then the sampling and image generation step is performed to finally generate a secret-carrier image, as shown. Figure 9 As shown in (i), the final generated image has a certain similarity to the original real image. This method allows for some control over the content of the generated image. During decryption, the StegoDiffusion method is used to obtain the image before sampling, extract the secret information, and decrypt it to obtain the plaintext image.
[0059] Example 3 Please see Figure 2 An embodiment of the present invention provides a controllable image steganography system based on a steganalysis diffusion model, the system comprising the following modules: Acquisition module: used to acquire the hash value of the color plaintext image to be encrypted, and generate the parameters and initial values of the three-dimensional memristor chaotic map based on the hash value; First generation module: used to drive the three-dimensional memristor chaotic mapping using the parameters and initial values, and iteratively generate multiple random sequences; The second generation module is used to decompose the color plaintext image to be encrypted into three channel matrices R, G, and B, and perform scrambling and diffusion operations on the channel matrices based on the random sequence to generate a color ciphertext image. Noise addition module: used to select a carrier image and add noise to the carrier image to obtain a preprocessed carrier image; Discrete wavelet transform module: used to perform discrete wavelet transform on the R, G, and B channels of the preprocessed carrier image respectively to obtain multiple wavelet coefficient matrices for the corresponding channels; The third generation module is used to decompose the pixel values of each channel of the color encrypted image into digital bits, and embed the decomposed digital values into the specified decimal places of the wavelet coefficient matrix to generate a encrypted wavelet coefficient matrix. Inverse Discrete Wavelet Transform: Used to perform inverse discrete wavelet transform on the dense wavelet coefficient matrix to obtain a noisy and dense carrier image; Denoising module: Denoises the noisy and dense carrier image using a steganalysis diffusion model to generate the final dense carrier image.
[0060] The system proposed in this embodiment is based on the method proposed in Embodiment 1. Therefore, the options proposed in Embodiment 1 are also applicable to this embodiment. To avoid repetition, they will not be described again here.
[0061] Furthermore, this embodiment also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method described in Embodiment 1.
[0062] The readable storage medium can be any available medium or data storage device that the processor can access, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), etc.), optical storage (such as compact disks (CD), digital video discs (DVD), Blu-ray discs (BD), high-definition universal discs (HVD), etc.), and semiconductor storage (such as read-only memory (ROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), non-volatile memory (NAND FLASH), solid-state disks (SSD), etc.).
[0063] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.
Claims
1. A steganographic method of controllable image based on steganographic diffusion model, characterized in that, The method comprises the following steps: Obtaining the hash value of the color plaintext image to be encrypted, and generating the parameters and initial value of the three-dimensional memristor chaotic mapping according to the hash value; Driving the three-dimensional memristor chaotic mapping with the parameters and initial value to iteratively generate a plurality of random sequences; Decomposing the color plaintext image to be encrypted into R, G, and B channel matrices, and performing scrambling and diffusion operations on the channel matrices based on the random sequences to generate a color ciphertext image; Selecting a carrier image and adding noise to the carrier image to obtain a preprocessed carrier image; Performing discrete wavelet transform on the R, G, and B channels of the preprocessed carrier image respectively to obtain a plurality of wavelet coefficient matrices corresponding to the channels; Decomposing the pixel values of each channel of the color ciphertext image by digits, and embedding the decomposed digit values into specified decimal places of the wavelet coefficient matrices to generate a ciphertext-containing wavelet coefficient matrix; Performing inverse discrete wavelet transform on the ciphertext-containing wavelet coefficient matrix to obtain a noisy ciphertext-containing carrier image; Using a steganographic diffusion model to denoise the noisy ciphertext-containing carrier image to generate a final ciphertext-containing carrier image.
2. The method of claim 1, wherein, The three-dimensional memristor chaotic mapping is obtained by introducing a memristor into a two-dimensional chaotic mapping, and the mathematical model of the three-dimensional memristor chaotic mapping is represented as follows: The parameters and initial value of the three-dimensional memristor chaotic mapping are generated according to the hash value, specifically including: wherein, , , , , are constants, is a parameter of the three-dimensional memristive chaotic map.
3. The method of claim 1, wherein, Using a hash function to calculate the hash value of the color plaintext image to be encrypted, dividing the hash value into a plurality of groups by a predetermined number of digits, and converting each group of hash values after grouping into a decimal number; Performing XOR calculation on the obtained decimal numbers to obtain the parameters of the three-dimensional memristor chaotic mapping; Based on the parameters of the three-dimensional memristor chaotic mapping, the initial value of the three-dimensional memristor chaotic mapping is calculated. The three-dimensional memristor chaotic mapping is driven by the parameters and initial value to iteratively generate a plurality of random sequences, including substituting the initial value into the three-dimensional memristor chaotic mapping, iteratively calculating a new initial value, stopping until the iteration threshold is reached, and outputting a plurality of random sequences.
4. The method of claim 2, wherein, The color plaintext image to be encrypted is decomposed into R, G, and B channel matrices, and scrambling and diffusion operations are performed on the channel matrices based on the random sequences to generate a color ciphertext image, specifically including: The color plaintext image to be encrypted is decomposed into R, G, and B channel matrices, and the three channel matrices are spliced to obtain a spliced matrix; 5. The method of claim 1, wherein, Quantizing, integerizing, and range-constraining the random sequences to obtain a first sequence for scrambling and a second sequence for XOR operation; The spliced matrix is expanded into a one-dimensional sequence, the first sequence is used to scramble the one-dimensional sequence to obtain a third sequence; The elements in the third sequence are XOR calculated with the elements in the second sequence to obtain a fourth sequence, and the fourth sequence is subjected to a restoration operation to obtain a restored matrix; The pixel values of each channel of the color ciphertext image are decomposed by digits, and the decomposed digit values are embedded into specified decimal places of the wavelet coefficient matrices to generate a ciphertext-containing wavelet coefficient matrix. separating three channel matrices from the reduced matrix , and combining the three channel matrices to obtain a color ciphertext image.
6. The method of claim 1, wherein, The carrier image is selected, and noise is added to the carrier image to obtain a pretreated carrier image, including: adding Gaussian noise to the carrier image to obtain the pretreated carrier image. The carrier image is selected, and noise is added to the carrier image to obtain a pretreated carrier image, including: adding Gaussian noise to the carrier image to obtain the pretreated carrier image.
7. The method of claim 6, wherein, The pixel values of each channel of the color ciphertext image are digitized by ones, tens and hundreds; The digitized values are embedded into the specified decimal digits of the wavelet coefficient matrix to generate a wavelet coefficient matrix containing the ciphertext.
8. The method of claim 6, wherein, The steganographic diffusion model is used to denoise the noisy stego carrier image to generate a final stego carrier image, including using a steganographic diffusion model to denoise the noisy stego carrier image Step denoising to obtain a stego carrier image.
9. A steganographic system for controllable image steganography based on a diffusion model, the system being based on the method according to any one of claims 1 to 8, characterized in that, The system comprises the following modules: An obtaining module is configured to obtain a hash value of a color plaintext image to be encrypted, and generate parameters and initial values of a three-dimensional memristor chaotic map according to the hash value; A first generating module is configured to drive the three-dimensional memristor chaotic map by using the parameters and initial values, and iteratively generate a plurality of random sequences; A second generating module is configured to decompose the color plaintext image to be encrypted into R, G and B channel matrices, and perform scrambling and diffusion operations on the channel matrices based on the random sequences to generate a color ciphertext image; A noise adding module is configured to select a carrier image, and add noise to the carrier image to obtain a preprocessed carrier image; A discrete wavelet transform module is configured to perform discrete wavelet transform on the R, G and B channels of the preprocessed carrier image respectively to obtain a plurality of wavelet coefficient matrices corresponding to the channels; A third generating module is configured to digitize the pixel values of each channel of the color ciphertext image by digits, and embed the digitized values into the specified decimal digits of the wavelet coefficient matrix to generate a wavelet coefficient matrix containing the ciphertext; An inverse discrete wavelet transform module is configured to perform inverse discrete wavelet transform on the wavelet coefficient matrix containing the ciphertext to obtain a noisy carrier image containing the ciphertext; A denoising module is configured to denoise the noisy carrier image containing the ciphertext by using a steganographic diffusion model to generate a final carrier image containing the ciphertext.
10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by a processor, implements the steps of the method of any one of claims 1 to 8. The computer program, when executed by a processor, implements the steps of the method of any one of claims 1 to 8.