CN117611422B - Image steganography method based on Moire pattern generation - Google Patents
Image steganography method based on Moire pattern generation Download PDFInfo
- Publication number
- CN117611422B CN117611422B CN202410089255.3A CN202410089255A CN117611422B CN 117611422 B CN117611422 B CN 117611422B CN 202410089255 A CN202410089255 A CN 202410089255A CN 117611422 B CN117611422 B CN 117611422B
- Authority
- CN
- China
- Prior art keywords
- image
- sub
- mole
- pattern
- secret
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 31
- 230000006870 function Effects 0.000 claims description 16
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims description 2
- 230000015572 biosynthetic process Effects 0.000 abstract description 4
- 238000013461 design Methods 0.000 abstract description 4
- 238000003786 synthesis reaction Methods 0.000 abstract description 4
- 230000004913 activation Effects 0.000 description 6
- 238000012549 training Methods 0.000 description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/0021—Image watermarking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0475—Generative networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/094—Adversarial learning
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Editing Of Facsimile Originals (AREA)
- Image Processing (AREA)
Abstract
The invention discloses an image steganography method based on mole pattern generation, which belongs to the field of image information hiding and comprises the following steps: acquiring a clean image and a noise vector; inputting the clean image and the noise vector into a condition generation countermeasure network to obtain sub-mole pattern with different layers; acquiring a secret message, hiding the secret message in the sub-mole pattern based on an encoder, and obtaining a sub-mole pattern of the hidden message; and generating a secret-containing image based on the sub-mole pattern of the hidden message and the clean image. The invention uses the conditional gan frame to generate the mole pattern, which simplifies the artificial design flow in the traditional synthesis mode; the secret message is hidden in the sub-mole pattern through the encoder, so that the sub-mole pattern of the hidden message is obtained; the embedding strategy of the present invention can improve the non-detectability of conventional steganalysis tools.
Description
Technical Field
The invention belongs to the technical field of image information hiding, and particularly relates to an image steganography method based on Moire pattern generation.
Background
Steganography is a secret communication technique. It uses common media such as text, audio, image or video to hide the secret information so that any third party other than the sender and the receiver cannot find the existence of the secret information. Therefore, it is widely used as a necessary complement to cryptography to ensure privacy of data.
Most image generation-based steganography schemes currently attempt to synthesize realistic images. In practice, depth synthesis can synthesize not only images, but also noise and distortion on natural images. Since these noise and distortions are characteristic of natural images, secret information therein can be hidden without being suspected. For this, we explore the moire pattern on natural images, which is typically caused by shooting screen content using a smartphone. They are very common and come in a variety of shapes, colors and frequency ranges. This is advantageous for their application in hiding secret messages.
Currently existing moire synthesis is accomplished by simulating the screen shot process. It requires a series of manual designs and takes into account a number of influencing factors, the process is complex and difficult to control.
Disclosure of Invention
The invention provides an image steganography method based on mole pattern generation, which utilizes the advantage of deep learning, reduces the artificial design of mole pattern generation, and improves the robustness and safety of the steganography method so as to solve the technical problems in the prior art.
In order to achieve the above object, the present invention provides an image steganography method based on moire generation, including:
acquiring a clean image and a noise vector;
inputting the clean image and the noise vector into a condition generation countermeasure network to obtain sub-mole pattern with different layers;
Acquiring a secret message, hiding the secret message in the sub-mole pattern based on an encoder, and obtaining a sub-mole pattern of the hidden message;
And generating a secret-containing image based on the sub-mole pattern of the hidden message and the clean image.
Preferably, the condition generating countermeasure network includes: a generator, a discriminator and an analyzer, by which the clean image and the noise vector are generated into different levels of sub-moire patterns.
Preferably, the generator is a multi-branch structure, each branch comprising: a downsampled block, a number of convolutions blocks, and an upsampled block.
Preferably, the method further comprises:
adding the sub moire patterns to obtain moire patterns, and generating moire images based on the moire patterns and the clean images;
And identifying the authenticity of the Moire image through the identifier.
Preferably, hiding the secret message in the sub-moire pattern based on an encoder includes:
and acquiring a binary secret message, converting the secret message into a two-dimensional vector with the same size as the sub-mole pattern, and inputting the two-dimensional vector and the sub-mole pattern into an encoder to obtain the sub-mole pattern of the hidden message.
Preferably, the encoder comprises: a number of convolution blocks and a gaussian low pass filter layer, the convolution blocks comprising: the convolutional layers, batchnorm layers, and relu activate functions.
Preferably, the method further comprises:
and adding the sub-mole pattern of the unhidden message to the clean image to obtain a generated mole pattern image.
Preferably, the method further comprises:
and distinguishing the secret image and the moire image respectively based on whether the secret message is contained or not by the analyzer.
Preferably, the method further comprises: and inputting the secret image to a decoder to obtain a secret message.
Preferably, the decoder comprises: a number of convolution blocks, an adaptive averaging pooling layer and a linear layer, the convolution blocks comprising: the convolutional layers, batchnorm layers, and relu activate functions.
Compared with the prior art, the invention has the following advantages and technical effects:
The invention provides an image steganography method based on mole pattern generation, which is characterized in that the clean image and the noise vector are input into a condition generation countermeasure network to obtain sub-mole pattern patterns with different layers; the invention uses the conditional gan frame to generate the mole pattern, which simplifies the artificial design flow in the traditional synthesis mode;
the secret message is hidden in the sub-mole pattern through the encoder, so that the sub-mole pattern of the hidden message is obtained; the method embeds the message into the moire pattern caused by the shooting screen instead of the image content, and the embedding strategy can improve the undetectability of the traditional steganography analysis tool, and meanwhile, the moire pattern caused by the shooting screen does not cause excessive influence on the content of the clean image, so that the method has certain attractive appearance. Each branch is connected with an encoder, so that messages are repeatedly embedded into sub-moire patterns of different layers, the frequency range of the embedding is increased, and the structure of a plurality of encoders enables the invention to be more robust to noise.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings:
FIG. 1 is a flow chart of an image steganography method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a multi-branch generator network according to an embodiment of the present invention;
Fig. 3 is a schematic diagram of an encoder network structure according to an embodiment of the present invention.
Detailed Description
It should be noted that, without conflict, the embodiments of the present application and features of the embodiments may be combined with each other. The application will be described in detail below with reference to the drawings in connection with embodiments.
It should be noted that the steps illustrated in the flowcharts of the figures may be performed in a computer system such as a set of computer executable instructions, and that although a logical order is illustrated in the flowcharts, in some cases the steps illustrated or described may be performed in an order other than that illustrated herein.
Example 1
As shown in fig. 1, in this embodiment, an image steganography method based on moire generation is provided, which specifically includes:
Step 1, reading 50000 the clean image and the corresponding real moire image from the moire data set tip2018, performing central clipping on the image according to the size of 600x600, and uniformly scaling the clipped image to the size of 256x 256.
Step 2, randomly sampling a noise vector of 3x256x256 from normal distribution; the noise vector and the clean image are spliced according to the channel dimension and input to the generator.
As shown in fig. 2, the generator is a multi-branch structure with 6 convolutions blocks and corresponding upsampling blocks for each branch. The first convolution block consists of a layer of convolution layers with the convolution kernel size of 3 and the step length of 1, relu activation functions and a layer of convolution layers with the convolution kernel size of 3 and the step length of 1 in sequence; the remaining convolution blocks consist of a sequence of convolutions of 3 convolutions of step size 1 and relu activation functions. The first convolution blocks of the other branches, except the first branch, all perform a downsampling operation of halving the input features. The output of the first convolution block of each branch is the input of the next branch. The first branch inputs a noise vector and a clean image. The upsampling block consists of a layer of deconvolution layers and relu activation functions. Each branch needs to be enlarged to the same size as the original image and has a channel number of 3. Thus, five different levels of sub-moire patterns can be obtained.
The loss function of the generator is:
where A denotes a discriminator, G denotes a generator, Z is a noise vector, I C is a clean image, T C is a true moire pattern (obtained by subtracting the true moire image from the clean image), L C is an L1 distance, and λ 0 is set to 0.1.
Step 3, adding the five sub-mole pattern patterns with different layers to obtain a generated mole pattern; adding a clean image to the generated moire pattern to obtain a generated moire image; the generated moire image and the real moire image are input into a discriminator; the discriminator judges the authenticity and improves the generation quality of the generator.
The discriminator network architecture is PatchGAN of the discriminator. The discriminator has five convolutions in total. The first four convolution blocks consist of a layer of downsampled convolutions, batchnorm layers and leakrelu activation functions, and the last convolution block consists of a layer of downsampled convolutions and sigmoid functions.
The loss function of the discriminator is:
in this embodiment, in the training process of the generator, the Adam optimizer is used to update the network parameters, and the learning rate is 0.0002. After training, a pre-training model of the generator is obtained.
Step 4, randomly sampling 0/1 binary message. Each bit of message is copied into a two-dimensional vector consistent with the image size, and the copied message vector is input to an encoder together with the sub-mole pattern.
Each sub-moire pattern corresponds to one encoder. Multiple encoders are advantageous to improve the robustness of the method.
As shown in fig. 3, the encoder consists of 6 convolution blocks and a gaussian low pass filter layer. Each convolution block consists of a layer of convolutions, batchnorm layers and relu activation function sequences. The first convolution block inputs a sub-moire pattern. The fifth convolution block inputs the concatenation of the intermediate feature and the message. The last output is the sub-moire pattern of the hidden message.
The sub-mole pattern of the hidden message is multiplied by a weight coefficient to adjust the robustness and security of the steganography.
The encoder loss function is:
Where S is the analyzer, E i is the i-th encoder, T i is the i-th generated sub-moire pattern, m represents a secret message, λ 1 is 1, and λ 2 is 0.001.
And 5, adding the sub-mole pattern of the hidden message and the sub-mole pattern of the non-hidden message to the clean image respectively to obtain a secret image and a generated mole pattern image.
And 6, inputting the secret image and the generated moire image into an analyzer, and distinguishing by the analyzer to improve the safety of the method.
The analyzer loss function is:
And 7, inputting the secret image into a decoder through Gaussian noise and jpeg compressed disturbance.
The decoder extracts the secret message from the secret-containing image. The decoder consists of seven convolution blocks, an adaptive average pooling layer and a linear layer. Each convolution block consists of a layer of convolutions, batchnorm layers and relu activation function sequences.
The decoder loss function is:
Wherein the method comprises the steps of Is a secret image and D is a decoder.
In this embodiment, in the decoder training process, an Adam optimizer is used to update network parameters, and the learning rate is 0.001.
The present application is not limited to the above-mentioned embodiments, and any changes or substitutions that can be easily understood by those skilled in the art within the technical scope of the present application are intended to be included in the scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.
Claims (7)
1. An image steganography method based on moire generation, comprising:
acquiring a clean image and a noise vector;
inputting the clean image and the noise vector into a condition generation countermeasure network to obtain sub-mole pattern with different layers;
the condition generating countermeasure network includes: a generator, a discriminator, and an analyzer, through which the clean image and the noise vector are generated into different levels of sub-moire patterns;
The generator is a multi-branch structure, each branch comprising: a downsampling block, a plurality of convolution blocks and an upsampling block; the output of each branch in the multi-branch structure is a sub-mole pattern with different layers;
Acquiring a secret message, hiding the secret message in the sub-mole pattern based on an encoder, and obtaining a sub-mole pattern of the hidden message;
And generating a secret-containing image based on the sub-mole pattern of the hidden message and the clean image.
2. The mole-based image steganography method of claim 1, further comprising:
adding the sub moire patterns to obtain moire patterns, and generating moire images based on the moire patterns and the clean images;
And identifying the authenticity of the Moire image through the identifier.
3. The mole-based generated image steganography method of claim 1, wherein hiding the secret message in the sub-mole pattern based on an encoder comprises:
and acquiring a binary secret message, converting the secret message into a two-dimensional vector with the same size as the sub-mole pattern, and inputting the two-dimensional vector and the sub-mole pattern into an encoder to obtain the sub-mole pattern of the hidden message.
4. The mole-based image steganography method of claim 1, wherein the encoder comprises: a number of convolution blocks and a gaussian low pass filter layer, the convolution blocks comprising: the convolutional layers, batchnorm layers, and relu activate functions.
5. The mole-based image steganography method of claim 2, further comprising:
and distinguishing the secret image and the moire image respectively based on whether the secret message is contained or not by the analyzer.
6. The mole-based image steganography method of claim 1, further comprising: and inputting the secret image to a decoder to obtain a secret message.
7. The mole-based image steganography method of claim 6, wherein the decoder comprises: a number of convolution blocks, an adaptive averaging pooling layer and a linear layer, the convolution blocks comprising: the convolutional layers, batchnorm layers, and relu activate functions.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410089255.3A CN117611422B (en) | 2024-01-23 | 2024-01-23 | Image steganography method based on Moire pattern generation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410089255.3A CN117611422B (en) | 2024-01-23 | 2024-01-23 | Image steganography method based on Moire pattern generation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN117611422A CN117611422A (en) | 2024-02-27 |
CN117611422B true CN117611422B (en) | 2024-05-07 |
Family
ID=89958242
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202410089255.3A Active CN117611422B (en) | 2024-01-23 | 2024-01-23 | Image steganography method based on Moire pattern generation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117611422B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2002279480A (en) * | 2001-03-19 | 2002-09-27 | Dainippon Printing Co Ltd | Image forming body and device for determining authenticity thereof |
CN113486377A (en) * | 2021-07-22 | 2021-10-08 | 维沃移动通信(杭州)有限公司 | Image encryption method and device, electronic equipment and readable storage medium |
CN114827381A (en) * | 2022-06-30 | 2022-07-29 | 北京大学深圳研究生院 | Strong robustness image steganography method and system based on condition standardization flow model |
WO2023151511A1 (en) * | 2022-02-08 | 2023-08-17 | 维沃移动通信有限公司 | Model training method and apparatus, image moire removal method and apparatus, and electronic device |
-
2024
- 2024-01-23 CN CN202410089255.3A patent/CN117611422B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2002279480A (en) * | 2001-03-19 | 2002-09-27 | Dainippon Printing Co Ltd | Image forming body and device for determining authenticity thereof |
CN113486377A (en) * | 2021-07-22 | 2021-10-08 | 维沃移动通信(杭州)有限公司 | Image encryption method and device, electronic equipment and readable storage medium |
WO2023151511A1 (en) * | 2022-02-08 | 2023-08-17 | 维沃移动通信有限公司 | Model training method and apparatus, image moire removal method and apparatus, and electronic device |
CN114827381A (en) * | 2022-06-30 | 2022-07-29 | 北京大学深圳研究生院 | Strong robustness image steganography method and system based on condition standardization flow model |
Non-Patent Citations (4)
Title |
---|
Image hiding based on time-averaging moiré;Minvydas Ragulskis , Algiment Aleksa;Optics Communications;20091231;第2752-2759页 * |
Minvydas Ragulskis, Algiment Aleksa, Jurate Ragulskiene.Image Hiding Based on Circular Moiré Fringes.WSEAS TRANSACTIONS on MATHEMATICS.第90-99页. * |
基于微结构网点的信息隐藏防伪技术研究;郭凌华 等;西安理工大学学报;20161230(第04期);第416-421页 * |
基于随机相位矩阵的版权信息隐藏方法;曹超;陈汝钧;;计算机应用;20081215(第S2期);第257-259页 * |
Also Published As
Publication number | Publication date |
---|---|
CN117611422A (en) | 2024-02-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108346125B (en) | Airspace image steganography method and system based on generation countermeasure network | |
CN109993678B (en) | Robust information hiding method based on deep confrontation generation network | |
Miller et al. | Applying informed coding and embedding to design a robust high-capacity watermark | |
Wang et al. | Image hiding by optimal LSB substitution and genetic algorithm | |
CN111640444B (en) | CNN-based adaptive audio steganography method and secret information extraction method | |
Wei et al. | Generative steganography network | |
CN112634117B (en) | End-to-end JPEG domain image steganography method based on generation of countermeasure network | |
CN102254295B (en) | Color halftoning image watermarking algorithm based on support vector machine | |
CN110232650B (en) | Color image watermark embedding method, detection method and system | |
CN115131188A (en) | Robust image watermarking method based on generation countermeasure network | |
CN112132737B (en) | Image robust steganography method without reference generation | |
CN114157773B (en) | Image steganography method based on convolutional neural network and frequency domain attention | |
CN111681155A (en) | GIF dynamic image watermarking method based on deep learning | |
CN116091288A (en) | Diffusion model-based image steganography method | |
CN111327785B (en) | Information steganography communication method based on automatic image construction of countermeasure generation network | |
CN107578365B (en) | Wavelet digital watermark embedding and extracting method based on quantum weed optimizing mechanism | |
CN116456037A (en) | Diffusion model-based generated image steganography method | |
Chanchal et al. | A comprehensive survey on neural network based image data hiding scheme | |
CN117611422B (en) | Image steganography method based on Moire pattern generation | |
CN115880125B (en) | Soft fusion robust image watermarking method based on Transformer | |
CN114630130B (en) | Face-changing video tracing method and system based on deep learning | |
Wu et al. | Modified multiway pixel-value differencing methods based on general quantization ranges for image steganography | |
CN116112685A (en) | Image steganography method based on diffusion probability model | |
DING et al. | High quality data hiding in halftone image based on block conjugate | |
CN108416726B (en) | A kind of digital picture steganography method keeping pixel frequency balance |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |