CN115358910A - Digital watermark attack method and system based on convolutional neural network denoising algorithm - Google Patents
Digital watermark attack method and system based on convolutional neural network denoising algorithm Download PDFInfo
- Publication number
- CN115358910A CN115358910A CN202210995602.XA CN202210995602A CN115358910A CN 115358910 A CN115358910 A CN 115358910A CN 202210995602 A CN202210995602 A CN 202210995602A CN 115358910 A CN115358910 A CN 115358910A
- Authority
- CN
- China
- Prior art keywords
- neural network
- convolutional neural
- digital watermark
- image
- denoising
- 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.)
- Pending
Links
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 72
- 238000000034 method Methods 0.000 title claims abstract description 36
- 230000000694 effects Effects 0.000 claims abstract description 19
- 238000012549 training Methods 0.000 claims description 35
- 238000005070 sampling Methods 0.000 claims description 19
- 230000006870 function Effects 0.000 claims description 12
- 230000004913 activation Effects 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 7
- 230000003044 adaptive effect Effects 0.000 claims description 5
- 238000011156 evaluation Methods 0.000 claims description 5
- 229910052709 silver Inorganic materials 0.000 claims description 5
- 239000004332 silver Substances 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000008030 elimination Effects 0.000 description 2
- 238000003379 elimination reaction Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000014759 maintenance of location Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013475 authorization Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000000593 degrading effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000005242 forging Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Editing Of Facsimile Originals (AREA)
- Image Processing (AREA)
Abstract
The invention relates to the technical field of digital watermark attack, and provides a digital watermark attack method and a digital watermark attack system based on a convolutional neural network denoising algorithm, wherein the method comprises the following steps: acquiring a digital watermark image; carrying out noise adding operation on the digital watermark image; converting the denoised digital watermark image into a plurality of sub-images through a downsampling operation; and connecting the denoised digital watermark image and all the sub-images in parallel, and attacking the watermark information by adopting a denoising convolutional neural network. The method can remove noise and maintain the detailed structure of the image, thereby improving the watermark attack effect of the network.
Description
Technical Field
The invention belongs to the technical field of digital watermark attack, and particularly relates to a digital watermark attack method and system based on a convolutional neural network denoising algorithm.
Background
The statements in this section merely provide background information related to the present disclosure and may not constitute prior art.
With the continuous development of internet technology and the increasing update of multimedia technology, a large number of digital images are widely spread. It is desirable to obtain the required digital image resources more conveniently, which also results in many users illegally copying, forging and disseminating the digital image content without authorization. Therefore, various watermarking methods have been studied for wide applications such as copyright protection and content authentication, and digital image watermarking technology has become an important technology for copyright protection of image resources. In the last few years, various digital image watermarking algorithms have been proposed to continuously improve the robustness and imperceptibility of the watermarked image.
At present, the digital watermarking technology is continuously improved, but the watermarking attack technology is not rapidly developed all the time, even stagnation is generated, and the watermarking attack technology is always stopped in the conventional attack methods, such as Gaussian noise addition, median filtering, mean filtering, lossy compression and the like. At present, most of digital watermarking algorithms have robustness which can effectively resist the conventional watermarking attack method. And the conventional watermark attack method can damage the original content of the digital image to a certain extent and influence the use of the image. Therefore, the robustness monitoring of the digital watermarking algorithm cannot be effectively improved in the research, the development of the digital watermarking algorithm is slowed down, and the robust digital watermarking technology cannot be effectively innovated.
Disclosure of Invention
In order to solve the technical problems existing in the background technology, the invention provides a digital watermark attack method and a digital watermark attack system based on a convolutional neural network denoising algorithm.
In order to achieve the purpose, the invention adopts the following technical scheme:
the first aspect of the present invention provides a digital watermark attack method based on a convolutional neural network denoising algorithm, which includes:
acquiring a digital watermark image;
carrying out noise adding operation on the digital watermark image;
converting the denoised digital watermark image into a plurality of sub-images through a downsampling operation;
and connecting the denoised digital watermark image with all the sub-images in parallel, and attacking the watermark information by adopting a denoising convolutional neural network.
Further, the training method of the denoising convolutional neural network comprises the following steps:
using a plurality of gray level images as carrier images, and respectively embedding tax-silver information images by using different watermark algorithms to obtain digital watermark image training sets containing different watermark algorithms;
and after the noise adding operation and the down-sampling operation are carried out on the manufactured training set, the noise level graph and the down-sampling subgraph are input into the denoising convolutional neural network in parallel, and the denoising convolutional neural network is subjected to adaptive learning training.
Further, the denoising convolutional neural network comprises a plurality of convolutional layers;
convolution operation and an activation function are adopted by a convolution layer of the first layer;
convolution operation, batch normalization and an activation function are adopted for the convolution layer of the middle layer;
the convolution layer of the last layer only adopts convolution operation.
Further, a denoised image is generated by an upsampling operation after the last layer of the convolutional layer.
Further, the attack effect of the denoising convolutional neural network on the watermark is measured by adopting the image peak signal-to-noise ratio and the bit error rate.
The second aspect of the present invention provides a digital watermark attack system based on a convolutional neural network denoising algorithm, which includes:
an image acquisition module configured to: acquiring a digital watermark image;
a noise module configured to: carrying out noise adding operation on the digital watermark image;
a downsampling module configured to: converting the denoised digital watermark image into a plurality of sub-images through a downsampling operation;
an attack module configured to: and connecting the denoised digital watermark image and all the sub-images in parallel, and attacking the watermark information by adopting a denoising convolutional neural network.
Further, a training module is included that is configured to:
using a plurality of gray level images as carrier images, and respectively embedding tax-silver information images by using different watermark algorithms to obtain digital watermark image training sets containing different watermark algorithms;
and after the noise adding operation and the down-sampling operation are carried out on the manufactured training set, the noise level graph and the down-sampling subgraph are input into the denoising convolutional neural network in parallel, and the denoising convolutional neural network is subjected to adaptive learning training.
Further, the attack effect evaluation module is also included and is configured to: and measuring the attack effect of the denoising convolutional neural network on the watermark by adopting the peak signal-to-noise ratio and the bit error rate of the image.
A third aspect of the present invention provides a computer-readable storage medium, on which a computer program is stored, which when executed by a processor, implements the steps in the digital watermark attack method based on the convolutional neural network denoising algorithm as described above.
A fourth aspect of the present invention provides a computer device, which includes a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the digital watermark attack method based on the convolutional neural network denoising algorithm as described above.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides a digital watermark attack method based on a convolutional neural network denoising algorithm, which converts a noise image into a plurality of sub-images by downsampling the noise image, increases the receptive field of the noise image to improve the performance, connects a noise level image as input in parallel with the sub-images obtained after downsampling, extracts the remarkable characteristics of the image, and maintains the detailed structure of the image while removing noise.
The invention provides a digital watermark attack method based on a convolutional neural network denoising algorithm, which adds batch normalization in a network structure, accelerates network convergence, improves the generalization capability of the network, prevents gradient elimination and overfitting, and increases the robustness of a system by restricting a system parameter search space so as to accelerate convergence.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
Fig. 1 is a structural diagram of a denoising convolutional neural network FFDNet according to a first embodiment of the present invention.
Detailed Description
The invention is further described with reference to the following figures and examples.
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
Example one
The embodiment provides a digital watermark attack method based on a convolutional neural network Denoising algorithm, and the adopted Fast and Flexible Denoising convolutional neural network FFDNet (perceived a Fast and Flexible Solution for CNN based Image Denoising) has the characteristics of Fast training speed, flexibility and effectiveness in Image Denoising, and obtains good balance in training speed and Denoising performance. The noise image is down-sampled to convert the image into four sub-images, the receptive field of the image is increased to improve the performance, a noise level graph is used as input to be connected in parallel with the four sub-images obtained after down-sampling, the operation is utilized to extract the significant features of the image, and the detailed structure of the image is maintained while the noise is removed; the method comprises the steps of applying a denoising convolutional neural network FFDNet to watermark attack, inputting a digital watermark image as a noise level image, taking a watermark in the watermark image as noise, and attacking watermark information in the watermark image through the denoising convolutional neural network FFDNet. The method specifically comprises the following steps:
step 1, acquiring a digital watermark image.
And 2, preprocessing.
When a denoising convolutional neural network FFDNet is used as a watermark attack scheme to attack a digital watermark image, the content quality of an original image is hopefully maintained while watermark information is damaged.
In the embodiment, the downsampling operation and the noise-complementing operation in the denoising convolutional neural network are taken as the preprocessing of the watermark attack scheme. Thus, the pretreatment mainly comprises:
(1) And carrying out noise adding operation on the digital watermark image.
Watermark information is typically embedded in low frequency regions of a digital image, while noise is more present in high frequency regions of the image. If the high-frequency area containing the watermark image has less noise, the denoising network cannot remove the image noise to a greater extent when attacking the watermark image, and cannot effectively damage the watermark information. Therefore, before the digital watermark image is input into the network, a certain degree of noise adding operation is carried out on the watermark image, after repeated experiments are carried out for many times, the noise adding operation with the noise intensity of sigma =15 or sigma =20 is carried out on the digital watermark image, and the noise added digital watermark image is input into the network for training attack, so that the damage effect on watermark information is best and the original content quality of the digital image is not influenced to a certain degree.
(2) The denoised digital watermark image is converted into four sub-images through downsampling operation, and the strategy not only reduces the degree of layers of the network and the number of filters on each layer, but also improves the fitting capability and the receptive field of the network. Setting the down-sampling factor to 2 allows it to increase the training speed without degrading the modeling capability.
And 3, training the denoising convolutional neural network FFDNet by adopting the preprocessed training set.
When watermark attack is carried out on different known digital watermark algorithms, more than 1000 gray images with the size of 256 multiplied by 256 are used as carrier images, watermark information images with the size of 16 multiplied by 16 are respectively embedded by using different watermark algorithms, a plurality of digital watermark image training sets containing different watermark algorithms are manufactured, and network training parameters are set to be 50 rounds of iterative training. The training environment uses a python-based torch framework, and the hardware conditions are trained using a CPU. After the noise adding operation and the down-sampling operation, namely preprocessing, are carried out on the manufactured training set, the noise level graph and the down-sampling subgraph are input into a denoising convolutional neural network FFDNet in parallel, and the denoising convolutional neural network FFDNet is subjected to self-adaptive learning training.
And 4, inputting the preprocessed digital watermark image into a trained denoising convolutional neural network FFDNet for watermark attack, and attacking watermark information in the digital watermark image.
After the watermark-containing images using different watermark algorithms are preprocessed, the preprocessed watermark-containing images are put into a corresponding trained denoising convolutional neural network FFDNet, and then watermark attack operation is carried out.
As an implementation mode, watermark extraction is carried out on an attacked image output by a denoising convolutional neural network FFDNet to detect the watermark attack effect.
The verification of the watermark attack effect mainly comprises two aspects, and in order to measure the image quality and the attack effect of an attacked image, the same index as the watermark algorithm evaluation is adopted in the embodiment, namely, the peak signal-to-noise ratio (PSNR) of the image:
the I and I' are images to be compared, the higher the PSNR value is, the higher the correlation degree of the front and rear images is, the more complete the image detail retention is, and if the PSNR value is lower, the more serious the image damage is, and the worse the detail retention degree is.
In order to accurately measure the watermark removal effect, namely the watermark extraction effect, the bit error rate BER is selected as an evaluation index, and the BER formula is as follows:
where B represents the number of bits of error information in the extracted watermark, and P × Q represents the total number of bits of the original watermark information. The value range of the bit error rate is [0,1], and when the BER value is closer to 0, the more complete the extracted watermark information is, the poorer the removal effect of the watermark attack is.
The network structure of the denoising convolutional neural network FFDNet is shown in FIG. 1:
the input layer connects four sub-images obtained by down-sampling the digital watermark image with the noise-added digital watermark image (noise level image) in parallel, so that the input image with W × H × C size is reconstructed into an input image with W × H × C size
The tensor of (a); w is the number of pixels in the horizontal direction of the image, H is the number of pixels in the vertical direction of the image, and C represents the number of channels of the image, and the scheme is only applied to the gray-scale digital watermark image, so the number of channels C is 1.
The next denoised convolutional neural network FFDNet is composed of a series of 3 × 3 convolutional layers, each of which is composed of a convolution operation (Conv), an activation function (ReLU), and Batch normalization (Batch); the denoising convolutional neural network FFDNet comprises 15 convolutional layers, and the number of convolutional filters is set to be 64; the convolution operation and the ReLU activation function are adopted in the first layer of convolution layer, the convolution operation, the Batch normalization and the ReLU activation function are adopted in the convolution layer of the middle layer, and only the convolution operation is adopted in the last layer of convolution layer;
and obtaining four sub-images after denoising through the last convolution layer, and generating the four sub-images after denoising into a denoised image with the size of W multiplied by H multiplied by C by adopting an up-sampling operation after the last convolution layer.
In order to make reasonable interpretation of the noise level map as network input, firstly most model-based image denoising methods aim at handling the flexibility of different noise levels, and the solution targets are as follows:
wherein the content of the first and second substances,
a learning mapping function representing a model, x representing the noise value of the image itself, y representing the noise value of the input,
for data fidelity term with noise level σ, Φ (x) is a regularization term related to the image prior, λ controls the weight between the data fidelity term and the regularization term, and at the same time, controls the balance between the degree of denoising and detail preservation, when it is too small, it leaves much unremoved noise, and when it is too large, it leaves much unremoved noiseMuch of the detail of the image will be too smooth as the noise is removed. To optimize the algorithm, the function is remapped to
Where Θ represents the model parameters of FFDNet.
Therefore, most denoising networks will take the noise image and noise level as input, but since the dimensions of the input y and λ are different, the denoising convolutional neural network FFDNet solves the problem of dimension mismatch by extending the noise level σ into a noise level map M, so the mapping function can be changed to:
the denoising convolutional neural network FFDNet has the characteristics of high training speed, flexibility and effectiveness in image denoising, achieves good balance in the training speed and the denoising performance, converts an image into four sub-images by down-sampling a noise image, increases the receptive field of the image to improve the performance, connects a noise level image serving as input with the four sub-images obtained after the down-sampling in parallel, extracts the significant characteristics of the image by using the operation, and keeps the detailed structure of the image while removing noise. In addition, the denoising convolutional neural network FFDNet improves denoising performance through downsampling operation, and reduces network depth. In addition, batch normalization is added to the network structure, network convergence is accelerated, the generalization capability of the network is improved, gradient elimination and overfitting are prevented, and system robustness is improved by restricting a system parameter search space, so that convergence can be accelerated. Therefore, it does not take much time to train the network using the CPU. The denoising network is applied to watermark attack, watermark information is input as a noise level graph, a watermark in a watermark image is taken as noise, and the watermark information in the watermark image is attacked through the denoising network.
In the digital watermark attack method based on the convolutional neural network denoising algorithm provided by the embodiment, the denoising convolutional neural network FFDNet is applied to digital watermark attack; the denoising convolutional neural network FFDNet has a good effect on the image denoising aspect, and the network depth is reduced by adopting the downsampling operation, so that the training time of the network is greatly shortened, the receptive field of network learning is increased by the downsampling operation, and the influence of the network depth on the denoising effect is balanced; in view of the advantages of the algorithm, the method is applied to a watermark attack method, watermark information is used as a part of noise in an image, and the denoising algorithm is used for destroying the watermark information while removing the image noise, so that the effect of watermark attack is achieved.
Example two
The embodiment provides a digital watermark attack system based on a convolutional neural network denoising algorithm, which specifically comprises the following modules:
an image acquisition module configured to: acquiring a digital watermark image;
a noise module configured to: carrying out noise adding operation on the digital watermark image;
a downsampling module configured to: converting the denoised digital watermark image into a plurality of sub-images through a downsampling operation;
an attack module configured to: and connecting the denoised digital watermark image and all the sub-images in parallel, and attacking the watermark information by adopting a denoising convolutional neural network.
A training module configured to:
using a plurality of gray level images as carrier images, and respectively embedding tax-silver information images by using different watermark algorithms to obtain digital watermark image training sets containing different watermark algorithms;
and after the noise adding operation and the down-sampling operation are carried out on the manufactured training set, the noise level graph and the down-sampling subgraph are input into the denoising convolutional neural network in parallel, and the denoising convolutional neural network is subjected to adaptive learning training.
An attack effect evaluation module configured to: and measuring the attack effect of the denoising convolutional neural network on the watermark by adopting the image peak signal-to-noise ratio and the bit error rate.
The denoising convolutional neural network comprises a plurality of convolutional layers;
convolution operation and an activation function are adopted by a convolution layer of the first layer;
convolution operation, batch normalization and activation functions are adopted by the convolution layer of the middle layer;
the convolution layer of the last layer only adopts convolution operation.
And generating a denoised image by an up-sampling operation after the convolution layer of the last layer.
It should be noted that, each module in the present embodiment corresponds to each step in the first embodiment one to one, and the specific implementation process is the same, which is not described herein again.
EXAMPLE III
The present embodiment provides a computer-readable storage medium, on which a computer program is stored, which when executed by a processor implements the steps in the digital watermark attack method based on the convolutional neural network denoising algorithm as described in the first embodiment.
Example four
The embodiment provides a computer device, which includes a memory, a processor and a computer program stored in the memory and executable on the processor, and when the processor executes the program, the processor implements the steps in the digital watermark attack method based on the convolutional neural network denoising algorithm as described in the first embodiment.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. The digital watermark attack method based on the convolutional neural network denoising algorithm is characterized by comprising the following steps:
acquiring a digital watermark image;
carrying out noise adding operation on the digital watermark image;
converting the denoised digital watermark image into a plurality of sub-images through downsampling operation;
and connecting the denoised digital watermark image with all the sub-images in parallel, and attacking the watermark information by adopting a denoising convolutional neural network.
2. The digital watermark attack method based on the convolutional neural network denoising algorithm as claimed in claim 1, wherein the training method of the denoising convolutional neural network is:
using a plurality of gray level images as carrier images, and respectively embedding tax-silver information images by using different watermark algorithms to obtain digital watermark image training sets containing different watermark algorithms;
and after the noise adding operation and the down-sampling operation are carried out on the manufactured training set, the noise level graph and the down-sampling subgraph are input into the denoising convolutional neural network in parallel, and the denoising convolutional neural network is subjected to adaptive learning training.
3. The digital watermark attack method based on the convolutional neural network denoising algorithm as claimed in claim 1, wherein the attack effect of the denoising convolutional neural network on the watermark is measured by using an image peak signal-to-noise ratio and a bit error rate.
4. The method of claim 1, wherein the denoised convolutional neural network comprises a plurality of convolutional layers;
convolution operation and an activation function are adopted by a convolution layer of the first layer;
convolution operation, batch normalization and activation functions are adopted by the convolution layer of the middle layer;
the convolution layer of the last layer only adopts convolution operation.
5. The digital watermark attack method based on the convolutional neural network denoising algorithm of claim 4, wherein a denoised image is generated by an upsampling operation after the convolutional layer of the last layer.
6. The digital watermark attack system based on the convolutional neural network denoising algorithm is characterized by comprising the following steps:
an image acquisition module configured to: acquiring a digital watermark image;
a noise module configured to: carrying out noise adding operation on the digital watermark image;
a downsampling module configured to: converting the denoised digital watermark image into a plurality of sub-images through downsampling operation;
an attack module configured to: and connecting the denoised digital watermark image with all the sub-images in parallel, and attacking the watermark information by adopting a denoising convolutional neural network.
7. The convolutional neural network denoising algorithm-based digital watermark attack system of claim 6, further comprising a training module configured to:
using a plurality of gray level images as carrier images, and respectively embedding tax-silver information images by using different watermark algorithms to obtain digital watermark image training sets containing different watermark algorithms;
and after the noise adding operation and the down-sampling operation are carried out on the manufactured training set, the noise level graph and the down-sampling subgraph are input into the denoising convolutional neural network in parallel, and the denoising convolutional neural network is subjected to adaptive learning training.
8. The convolutional neural network denoising algorithm-based digital watermark attack system of claim 6, further comprising an attack effect evaluation module configured to: and measuring the attack effect of the denoising convolutional neural network on the watermark by adopting the peak signal-to-noise ratio and the bit error rate of the image.
9. A computer-readable storage medium, on which a computer program is stored, which program, when being executed by a processor, implements the steps of the method for digital watermark attack based on a convolutional neural network denoising algorithm according to any one of claims 1 to 5.
10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method for digital watermark attack based on the convolutional neural network denoising algorithm according to any one of claims 1-5 when executing the program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210995602.XA CN115358910A (en) | 2022-08-18 | 2022-08-18 | Digital watermark attack method and system based on convolutional neural network denoising algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210995602.XA CN115358910A (en) | 2022-08-18 | 2022-08-18 | Digital watermark attack method and system based on convolutional neural network denoising algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115358910A true CN115358910A (en) | 2022-11-18 |
Family
ID=84001741
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210995602.XA Pending CN115358910A (en) | 2022-08-18 | 2022-08-18 | Digital watermark attack method and system based on convolutional neural network denoising algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115358910A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116308986A (en) * | 2023-05-24 | 2023-06-23 | 齐鲁工业大学(山东省科学院) | Hidden watermark attack algorithm based on wavelet transformation and attention mechanism |
-
2022
- 2022-08-18 CN CN202210995602.XA patent/CN115358910A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116308986A (en) * | 2023-05-24 | 2023-06-23 | 齐鲁工业大学(山东省科学院) | Hidden watermark attack algorithm based on wavelet transformation and attention mechanism |
| CN116308986B (en) * | 2023-05-24 | 2023-08-04 | 齐鲁工业大学(山东省科学院) | Hidden watermark attack algorithm based on wavelet transformation and attention mechanism |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111275637B (en) | Attention model-based non-uniform motion blurred image self-adaptive restoration method | |
| CN110517195B (en) | Unsupervised SAR image denoising method | |
| CN116051408B (en) | Image depth denoising method based on residual error self-coding | |
| CN111696046A (en) | Watermark removing method and device based on generating type countermeasure network | |
| Wang et al. | Mapping based residual convolution neural network for non-embedding and blind image watermarking | |
| Huang et al. | Two-step approach for the restoration of images corrupted by multiplicative noise | |
| CN110111266B (en) | Approximate information transfer algorithm improvement method based on deep learning denoising | |
| CN115908095A (en) | Hierarchical attention feature fusion-based robust image watermarking method and system | |
| CN115809953A (en) | Attention mechanism-based multi-size image robust watermarking method and system | |
| CN115345768A (en) | Robust watermark attack method and system based on neural network | |
| Zhang et al. | Adaptive reversible image watermarking algorithm based on IWT and level set | |
| CN115984979A (en) | Unknown-countermeasure-attack-oriented face counterfeiting identification method and device | |
| CN115358910A (en) | Digital watermark attack method and system based on convolutional neural network denoising algorithm | |
| Lu et al. | Wavelet-based CNN for robust and high-capacity image watermarking | |
| CN113763268B (en) | Blind restoration method and system for face image | |
| CN115358909A (en) | Hidden digital watermark attack method and system based on SAD network | |
| CN111145125A (en) | Image denoising method based on residual learning and convolutional neural network | |
| Zhang et al. | A blind watermarking system based on deep learning model | |
| Yadav et al. | Lagrangian twin support vector regression and genetic algorithm based robust grayscale image watermarking | |
| CN109584139B (en) | Safety embedding method suitable for batch self-adaptive steganography | |
| CN113592697B (en) | Digital watermark attack method and system based on convolutional neural network denoising thought | |
| CN116523794A (en) | Low-light image enhancement method based on convolutional neural network | |
| CN116308986A (en) | Hidden watermark attack algorithm based on wavelet transformation and attention mechanism | |
| CN116205776A (en) | Tamper positioning method based on semi-fragile reversible watermark | |
| Wang et al. | Wavelet-FCWAN: fast and covert watermarking attack network in wavelet domain |
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 |