CN109859091B - Image steganography detection method based on Gabor filtering and convolutional neural network - Google Patents
Image steganography detection method based on Gabor filtering and convolutional neural network Download PDFInfo
- Publication number
- CN109859091B CN109859091B CN201811583343.XA CN201811583343A CN109859091B CN 109859091 B CN109859091 B CN 109859091B CN 201811583343 A CN201811583343 A CN 201811583343A CN 109859091 B CN109859091 B CN 109859091B
- Authority
- CN
- China
- Prior art keywords
- image
- steganography
- steganography detection
- gabor
- neural network
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 147
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 47
- 238000001914 filtration Methods 0.000 title claims abstract description 40
- 238000012549 training Methods 0.000 claims abstract description 28
- 238000000605 extraction Methods 0.000 claims abstract description 23
- 238000013139 quantization Methods 0.000 claims description 19
- 238000013528 artificial neural network Methods 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 16
- 238000005070 sampling Methods 0.000 claims description 9
- 238000009825 accumulation Methods 0.000 claims description 4
- 238000000034 method Methods 0.000 abstract description 12
- 238000010276 construction Methods 0.000 abstract description 5
- 238000010586 diagram Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- 238000011176 pooling Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 239000000969 carrier Substances 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
An image steganography detection method based on Gabor filtering and convolutional neural network belongs to the technical field of information hiding, and is characterized in that: selecting a carrier image and a secret image to generate a sample image; extracting a steganography detection feature of a sample image; training the steganography detection characteristics and class marks of the sample image through an integrated classifier to obtain a steganography detector; and extracting the steganography detection characteristics of the image to be detected, and inputting the steganography detection characteristics to the steganography detector for image steganography detection. The method comprises the steps of utilizing a filter to carry out image filtering to construct a plurality of deep convolutional neural networks to carry out steganography detection feature learning, realizing extraction of diversified learning type steganography detection features, simultaneously utilizing a filter coefficient to carry out construction type steganography detection feature extraction, finally combining the learning type steganography detection features and the construction type steganography detection features as steganography detection features and utilizing an integrated classifier to carry out steganography detection, and the steganography detection method obviously reduces the detection error rate of self-adaptive steganography on images.
Description
Technical Field
The invention belongs to the technical field of information hiding, and particularly relates to an image steganography detection method based on Gabor filtering and a convolutional neural network.
Background
Information hiding technology refers to technology that embeds information into a host signal (such as an image, audio, video, or text file) using insensitivity of human senses and redundancy existing in the signal itself, and can detect or extract hidden information as necessary. The information hiding technology mainly comprises digital steganography, digital watermarking technology, visual passwords, protocol steganography and the like. Digital steganography is a technique for concealing the existence of secret information by embedding the secret information in redundant data of digital image, audio, video, etc. Digital steganography analysis is a reverse technology of digital steganography and is mainly used for detecting a steganography carrier, discovering steganography behaviors, extracting steganography messages and the like. In a network environment, the properties of digital images such as universality, availability, sufficient redundant space and the like make the digital images one of the most widely used carrier types in digital steganography, and therefore the digital steganography analysis research is the main object.
Since the JPEG image is one of the image formats with the widest application range, the current mainstream JPEG image self-adaptive steganography algorithm comprises UED, J-UNIWARD and the like, and the corresponding steganography detection method comprises a steganography detection method based on DCTR, PHARM, GFR and other high-dimensional characteristics and an integrated classifier, a steganography detection method based on a deep convolutional neural network and the like. For JPEG image self-adaptive steganography detection, the existing detection method can be divided into two frameworks: one is based on the framework of a 'steganographic detection feature+integrated classifier', and research focuses on steganographic detection feature extraction, wherein GFR features are steganographic detection features of the current mainstream. Another framework is based on "image high-pass filtering+depth convolutional neural network", and the research emphasis is on how to design the depth convolutional neural network according to the requirements of steganography detection, and the current mainstream methods include a depth convolutional neural network steganography detection method based on JPEG phase information and a depth convolutional neural network steganography detection method based on image residual quantization truncation and proposed by Zeng et al.
The GFR feature is one of the current JPEG image self-adaptive steganography detection features with the lowest detection error rate, a multi-scale and multi-directional 2D Gabor filter is utilized to filter the decompressed JPEG image, then each filtered residual image is subjected to histogram feature extraction according to 64 phases in an 8X 8DCT block of the JPEG image, histogram feature accumulation combination is carried out according to the symmetry of projection vectors of different phases, the 2D Gabor filter in the symmetry direction is combined to generate the histogram feature of the image residual, and finally all the extracted histogram features are combined to be used as the final steganography detection feature.
The deep convolution neural network steganography detection method based on JPEG phase information, which is proposed by Fridrich et al, mainly comprises three parts: the first part is an image processing layer, and 4 high-pass filters are utilized to carry out filtering operation on the image to obtain 4 residual images; the second part is a convolutional neural network layer, which totally comprises 5 processing modules and mainly performs operations such as convolution, data standardization, activation, pooling and the like; the third part is a classifier layer comprising a full link layer and a softmax layer, which realizes soft decision on an input sample image. The detection method is mainly characterized in that a filtering residual image is split into 64 sub-images according to JPEG image phase information.
The depth convolution neural network steganography detection method based on image residual quantization truncation, which is proposed by Zeng et al, mainly comprises four parts: the first part is an image processing layer, and 25 high-pass filters are utilized to filter images to obtain 25 residual images; the second part is a quantization and cut-off layer, 3 quantization operations are carried out on 25 residual images, and the cut-off threshold is 4; the third part is a convolutional neural network layer, and 3 quantization operations correspond to 3 convolutional neural networks; the fourth part is the classifier layer, which includes a full link network and softmax layer. The detection method is mainly characterized in that image filtering and residual image quantization and truncation operations are added in a convolutional neural network structure.
However, the existing JPEG image adaptive steganography detection technology has some disadvantages:
1. although GFR is a steganographic detection feature with the lowest current detection error rate, the detection error rate is also higher than the steganographic detection method based on deep learning;
the steganography detection method based on the depth convolution neural network, which is proposed by Fridrich et al and Zeng et al, has the problem that the types of the filters of the image processing layers are not abundant enough;
3. the prior art does not utilize deep learning to realize automatic learning and extraction of steganography detection characteristics;
4. the complementarity of the structured steganographic detection feature and the learning steganographic detection feature is to be studied.
Disclosure of Invention
The invention aims to solve the problems and provides an image steganography detection method based on Gabor filtering and a convolutional neural network.
The invention discloses an image steganography detection method based on Gabor filtering and convolutional neural network, which comprises the steps of selecting a carrier image and a carrier image to generate a sample image;
extracting the steganography detection characteristics of the sample image by using a Gabor filter and a deep convolutional neural network;
training the steganography detection characteristics and class marks of the sample image through an integrated classifier to obtain a steganography detector; and extracting the steganography detection characteristics of the image to be detected, and inputting the steganography detection characteristics to the steganography detector for image steganography detection.
The invention relates to an image steganography detection method based on Gabor filtering and convolutional neural network, which comprises the steps of generating a carrier image and a carrier image sample, extracting learning steganography detection characteristics and structural steganography detection characteristics of a sample image, and combining the learning steganography detection characteristics and the structural steganography detection characteristics to serve as steganography detection characteristics of the sample image;
forming a training sample by using the steganographic detection characteristics and class marks of the sample image;
training the integrated classifier by using a training sample, and taking the trained classifier as a steganography detector;
extracting learning type steganography detection characteristics and structural type steganography detection characteristics of an image to be detected, and combining the learning type steganography detection characteristics and the structural type steganography detection characteristics to be used as steganography detection characteristics of the image to be detected;
inputting the steganography detection characteristics of the image to be detected into a trained steganography detector for judgment;
and judging the image to be detected as a carrier image or a secret image according to the output result of the steganographic detector.
The invention relates to an image steganography detection method based on Gabor filtering and a convolutional neural network, wherein the class mark comprises a carrier image class mark 1 and a carrier image class mark-1.
The invention relates to an image steganography detection method based on Gabor filtering and convolutional neural network, wherein the extraction steps of learning type steganography detection characteristics comprise: firstly, filtering a sample image by utilizing a multi-scale multi-directional 2D Gabor filter, and then training a depth convolution neural network by utilizing filtered images obtained by the 2D Gabor filters with the same scale parameters respectively to obtain a plurality of trained depth convolution neural networks; and when the learning type features are extracted, performing multi-scale multi-directional 2D Gabor filtering on the sample image, inputting different scale filtering images into a corresponding trained deep convolutional neural network to obtain a plurality of corresponding learning type steganography detection features, and finally combining the obtained learning type steganography detection features to obtain a final learning type steganography detection feature.
The invention relates to an image steganography detection method based on Gabor filtering and convolutional neural network, wherein the extraction steps of the structural steganography detection feature comprise: firstly, decompressing an image, and then performing high-pass filtering operation on the decompressed image by utilizing a multi-scale multi-directional 2D Gabor filter to obtain a residual image; secondly, carrying out multiple times of quantization operation on each residual image by utilizing multiple quantization step sizes, and further respectively carrying out histogram feature extraction and accumulation combination on the multiple quantized images to serve as histogram features corresponding to one residual image; and then merging the 2D Gabor filters in the symmetrical direction to generate histogram features of the residual image, and finally combining all the merged histogram features to serve as final structural steganography detection features.
The image steganography detection method based on Gabor filtering and convolutional neural network provided by the invention has the image format of JPEG.
The invention relates to an image steganography detection method based on Gabor filtering and convolutional neural network, wherein the learning type steganography detection feature extraction comprises the following steps:
step1: generating 2D Gabor filters with different parameters; a 2D Gabor filter is generated using a 2D Gabor function shown in the following formula,
where x '=xcos θ+ysin θ, y' = -xsin θ+ycos θ, σ=0.56λ, γ=0.5;
step2: generating a residual image; performing high-pass filtering operation on the input training sample image by using the 2D Gabor filter generated in the previous step to obtain a filtered residual image;
step3: training a deep convolutional neural network; respectively training a deep convolutional neural network by using the filtered residual images;
step4: extracting learning type steganography detection characteristics; and respectively carrying out convolution operation on the input image and the 2D Gabor filter, respectively inputting the obtained residual images into corresponding deep convolution neural networks, taking the characteristics output by each deep convolution neural network, and finally combining the obtained multiple learning type steganography detection characteristics into a final learning type steganography detection characteristic.
The invention relates to an image steganography detection method based on Gabor filtering and convolutional neural network, wherein the generation steps of the 2DGabor filter are as follows: (1) generating sampling points; assuming that the 2D Gabor filter size is M×N, the value range of x is
The value range of y is +.>
Generating sampling points (x, y) in step size 1; (2) determining filter parameters; determination of parameters sigma, theta and +.>
Calculating the parameter lambda value, gamma=0.5;(3) Generating a 2D Gabor filter; according to the 2D Gabor function expression and the parameter value, calculating a function value at a sampling point to obtain a corresponding 2D Gabor filter; (4) normalizing the 2D Gabor filter; subtracting the average value of all its elements from the 2D Gabor filter yields a zero-average high-pass filter.
The invention discloses an image steganography detection method based on Gabor filtering and convolutional neural network, which comprises the steps of carrying out image filtering by utilizing a multi-scale multi-directional 2D Gabor filter, then constructing a plurality of deep convolutional neural networks to carry out steganography detection feature learning, realizing extraction of diversified learning type steganography detection features, simultaneously carrying out construction type steganography detection feature extraction by utilizing a 2D Gabor filter coefficient of a JPEG image, finally combining the learning type steganography detection features and the construction type steganography detection features as steganography detection features and carrying out steganography detection by utilizing an integrated classifier, wherein the steganography detection method obviously reduces the detection error rate of self-adaptive steganography on the JPEG image, reduces the detection error rate of a quality factor of 75 and an embedding rate of 0.3 by 12 percent compared with the detection error rate of GFR and reduces the detection error rate by 6 percent compared with methods such as Jessesia.
Drawings
FIG. 1 is a schematic diagram of a steganography detection feature construction process according to the present invention;
FIG. 2 is a schematic diagram of learning steganography detection feature extraction according to the present invention;
FIG. 3 is a schematic representation of the extraction of structured steganography detection features of the present invention;
FIG. 4 is a schematic diagram of a classifier training process according to the present invention;
FIG. 5 is a schematic diagram of a process flow of an image to be detected according to the present invention;
FIG. 6 is a schematic diagram of a deep convolutional neural network according to the present invention;
FIG. 7 is a schematic diagram showing the combination of 64 (T+1) -dimensional histogram features when the 2D Gabor filter direction parameter θ is 0 or pi/2;
fig. 8 is a schematic diagram showing the combination of 64 (t+1) -dimensional histogram features when the direction parameter θ of the 2D Gabor filter of the present invention takes other values except 0 and pi/2.
Detailed Description
The method for detecting the image steganography based on the filter and the deep convolution neural network is described in detail below with reference to the accompanying drawings and the embodiments.
Example 1
The invention discloses an image steganography detection method based on Gabor filtering and convolutional neural network, which comprises the following steps: generating a carrier image and a carrier secret image sample, wherein the learning type steganography detection feature and the structural type steganography detection feature of the sample image are extracted as shown in fig. 1, and the learning type steganography detection feature and the structural type steganography detection feature are combined to be used as the steganography detection feature of the sample image; as shown in fig. 2, the extraction step of the learning steganography detection feature includes: firstly, filtering a JPEG image by utilizing a multi-scale multi-directional 2D Gabor filter, and then training a depth convolution neural network by utilizing filtered images obtained by the 2D Gabor filters with the same scale parameters respectively to obtain a plurality of trained depth convolution neural networks; and secondly, carrying out multi-scale multi-direction 2D Gabor filtering on the sample image, inputting the sample image into a corresponding trained deep convolutional neural network to obtain a plurality of learning type steganography detection features, and finally combining the obtained plurality of learning type steganography detection features into a final learning type steganography detection feature.
As shown in fig. 3, the extraction step of the structured steganography detection feature includes: firstly, decompressing a JPEG image, and then performing high-pass filtering operation on the decompressed image by utilizing a multi-scale multi-directional 2D Gabor filter to obtain a residual image; and secondly, carrying out multiple quantization operations on each residual image by utilizing multiple quantization step sizes, respectively carrying out histogram feature extraction and accumulation combination on the multiple quantized images to be used as a histogram feature corresponding to one residual image, then combining the 2D Gabor filters in the symmetrical directions to generate the histogram feature of the residual image, and finally combining all the combined histogram features to be used as a final structural steganography detection feature.
As shown in fig. 4, the steganographic detection feature and the class mark of the sample image are combined into a training sample; the carrier image class is marked as 1, and the carrier image class is marked as-1;
training the integrated classifier by using a training sample, and taking the trained classifier as a steganography detector;
as shown in fig. 5, learning-type steganography detection features and structural-type steganography detection features of an image to be detected are extracted, and the learning-type steganography detection features and the structural-type steganography detection features are combined as steganography detection features of the image to be detected;
inputting the steganography detection characteristics of the image to be detected into a trained steganography detector for judgment;
and judging the image to be detected as a carrier image or a secret image according to the output result of the steganographic detector.
Example two
On the basis of the first embodiment, the residual images are generated by using different sets of 2D Gabor filters, and then the residual images are used for training a plurality of depth convolutional neural networks and realizing learning type feature extraction, which comprises the following steps:
step1: generating four sets of 2D Gabor filters of different parameters
A 2D Gabor filter is generated using a 2D Gabor function shown in the following formula,
where x '=xcos θ+ysin θ, y' = -xsin θ+ycos θ, σ=0.56λ, γ=0.5.
The generation steps of the 2D Gabor filter are as follows: (1) generating sampling points. Assuming that the 2D Gabor filter size is M×N, the value range of x is
The value range of y is +.>
Generating sampling points (x, y) in step size 1; (2) determining filter parameters. Determination of the parameters sigma, theta and +.in the 2DGabor function>
Calculating the parameter lambda value, gamma=0.5; (3) generating a 2D Gabor filter. According to the 2D Gabor function expression and the parameter value, calculating a function value at a sampling point to obtain a corresponding 2D Gabor filter; (4) normalizing the 2D Gabor filter; subtracting the average value of all its elements from the 2D Gabor filter yields a zero-average high-pass filter.
The 2D Gabor filter has scale parameters of 8X8, scale parameters sigma are respectively 0.75, 1, 1.25 and 1.5, and corresponding to each scale parameter sigma, direction parameters theta are respectively {0, pi/4, 2pi/4, 3pi/4 }, and phase parameters
Taking 0 and pi/2 respectively, 8 2D Gabor filters with the same scale parameters form a filter group, and four groups of 2D Gabor filters are combined.
Step2: generating four sets of residual images
As shown in fig. 2, the high-pass filtering operation is performed on the input training sample image by using the four sets of 2D Gabor filters generated in the previous step, so as to obtain four sets of filtered residual images.
Step3: training four deep convolutional neural networks
As shown in fig. 2, four depth convolutional neural networks, each having a structure as shown in fig. 6, are trained using four sets of filtered residual images, respectively. The deep convolutional neural network shown in fig. 6 can be structurally divided into 5 groups, each group comprises a plurality of layers, and the convolutional filter in each layer is expressed as follows: the number of filters x (filter height x filter width x number of input feature maps), the feature maps are expressed as: number of feature patterns x (feature pattern height x feature pattern width); ABS represents the absolute value layer, i.e. takes the absolute value of the filtered image; batch Normalization (BN) is a batch normalization layer, namely, batch normalization operation is carried out on the feature map, so that the mean value is 0 and the variance is 1; the TanH and ReLU are activation functions, and the data can be limited in a smaller value range by using the TanH activation functions in Group1 and Group 2; avg Pooling represents an average Pooling layer; phase Split means that a feature image obtained from a JPEG image is sub-sampled in accordance with 64 phases in an 8×8 block, and 64 sub-images can be obtained from one feature image.
When the deep convolutional neural network is trained, a small batch of random gradient descent algorithm is utilized for carrying out network training, and the loss function is a cross entropy function. In the random gradient descent algorithm, the momentum value is set to 0.9, the parameter learning rate is initially set to 0.001, the learning rate is reduced by 25% every 20 generations in the training process, and the total iteration number of the training algorithm is 240 generations; each batch of training samples contains 40 images, wherein 20 carriers and 20 carriers are closely packed; all convolution filters in the convolution neural network are initialized by Gaussian distribution with the mean value of 0 and the standard deviation of 0.01, bias parameters in the convolution layer are fixed to be 0, learning rate of bias parameters in the full-connection layer is twice that of other parameters of the network, other parameters in the full-connection layer are initialized by using an Xavier method, L2 norm weight attenuation coefficient in the full-connection layer is set to be 0.01, and other layers have no weight attenuation.
Step4: extracting learning steganography detection features
After the four deep convolutional neural networks are trained, the networks can be used for extracting the learning type steganography detection characteristics of the images. The method specifically comprises the following steps: and respectively carrying out convolution operation on the input image and four groups of 2D Gabor filters, then respectively inputting the obtained four groups of residual images into four corresponding deep convolution neural networks, taking 512-dimensional features output by Group5 in each deep convolution neural network as learning-type features, and obtaining 2048-dimensional learning-type steganography detection features by the four deep convolution neural networks.
Example III
In the present invention, as shown in fig. 3, different sets of 2D Gabor filters are used to generate filtered residual images, then different step sizes are used to quantize the residual images, and finally histogram feature extraction is performed on the quantized residual images. The method comprises the following specific steps:
step1: decompressing the JPEG image to a space domain without rounding operation;
step2: generating a 2D Gabor filter group, wherein the scale parameter sigma is 0.75, 1, 1.25 and 1.5, the direction parameter sigma corresponding to each scale is {0, pi/16, 2 pi/16, …,15 pi/16 }, and the phase offset parameter is respectively
Respectively taking 0 and pi/2 to obtain four groups of 2D Gabor filters, wherein the number of the filters in each group is 32, and 128 filters are all obtained;
step3: decompressing the JPEG image with each 2D Gabor filter G in the filter bank σ,θ Convoluting to obtain a filtered residual image U σ,θ ；
Step4: for the filtered residual image U σ,θ Quantization of a plurality of different step sizes is carried out, then histogram feature extraction is carried out on the filtered residual image of each quantization step size according to the following formula,
wherein,,
representing the filtered residual image, q representing the quantization step size, T representing the truncation threshold and taken as 4, Q T Representing the quantization function with a quantization center {0,1,2, …, T }, and (a, b) representing the position in the 8 x 8 image block.
For the quantization step q, q' is calculated as follows,
wherein QF represents a JPEG image quality factor, and then the filtered residual image is quantized with a quantization step size of 0.5q ', q ',1.5q ', respectively.
Step5: according to the above histogram feature extraction formula, 64 (t+1) dimensional histogram features can be extracted from each residual image, when the direction parameter θ of the corresponding 2D Gabor filter is 0, pi/2, the histogram features at the same positions with the same reference numbers are combined according to the symmetrical relation shown in fig. 7, and finally a 25× (t+1) dimensional histogram feature is obtained, and when the direction parameter θ takes other values, a 34× (t+1) dimensional histogram feature is obtained by combining according to the symmetrical relation shown in fig. 8.
Step6: combining the histogram features extracted by the 2D Gabor filters with the same scale and phase offset parameters and symmetrical directions;
step7: combining the histogram features extracted from all the filtered images of the same quantization step to obtain a histogram feature of 2× (4× (2×25+ (16/2-1) ×34) × (4+1)) dimension;
step8: finally, the histogram features extracted from the different quantized step filtered residual images are accumulated to form the final steganographic detection feature, with the feature dimension still being 2× (4× (2×25+ (16/2-1) ×34) × (4+1))= 11520 dimensions.
Claims (5)
1. The image steganography detection method based on Gabor filtering and convolutional neural network is characterized in that:
selecting a carrier image and a secret image to generate a carrier image and a secret image sample;
extracting learning type steganography detection features and structural type steganography detection features of a sample image by using a Gabor filter and a deep convolutional neural network, and combining the learning type steganography detection features and the structural type steganography detection features to serve as steganography detection features of the sample image;
forming a training sample by using the steganographic detection characteristics and class marks of the sample image;
training the integrated classifier by using a training sample, and taking the trained classifier as a steganography detector;
extracting learning type steganography detection characteristics and structural type steganography detection characteristics of an image to be detected, and combining the learning type steganography detection characteristics and the structural type steganography detection characteristics to be used as steganography detection characteristics of the image to be detected;
inputting the steganography detection characteristics of the image to be detected into a trained steganography detector for judgment;
judging the image to be detected as a carrier image or a secret image according to the output result of the steganography detector;
the extracting step of the learning type steganography detection feature comprises the following steps: firstly, filtering a sample image by utilizing a multi-scale multi-directional 2DGabor filter, and then training a depth convolution neural network by utilizing filtered images obtained by the 2D Gabor filters with the same scale parameters respectively to obtain a plurality of trained depth convolution neural networks; when learning type features are extracted, carrying out multi-scale multi-directional 2D Gabor filtering on a sample image, then inputting different scale filtering images into a corresponding trained deep convolutional neural network to obtain a plurality of learning type steganography detection features, and finally combining the obtained plurality of learning type steganography detection features into a final learning type steganography detection feature;
the extraction step of the structural steganography detection feature comprises the following steps: firstly, decompressing an image, and then performing high-pass filtering operation on the decompressed image by utilizing a multi-scale multi-directional 2D Gabor filter to obtain a residual image; and then, carrying out multiple quantization operations on each residual image by utilizing multiple quantization step sizes, further respectively carrying out histogram feature extraction and accumulation combination on the multiple quantized images to be used as a histogram feature corresponding to one residual image, then combining the 2D Gabor filters in the symmetrical directions to generate the histogram feature of the residual image, and finally combining all the combined histogram features to be used as a final structural steganography detection feature.
2. The Gabor filtering and convolutional neural network-based image steganography detection method of claim 1, characterized in that: the class labels comprise carrier image class labels of 1 and carrier image class labels of-1.
3. The Gabor filtering and convolutional neural network-based image steganography detection method of claim 2, characterized in that: the image format is JPEG.
4. The Gabor filtering and convolutional neural network-based image steganography detection method of claim 3, wherein the steps of: the learning steganography detection feature extraction includes the steps of:
step1: generating 2D Gabor filters with different parameters; a 2D Gabor filter is generated using a 2D Gabor function shown in the following formula,
where x '=xcos θ+ysin θ, y' = -xsin θ+ycos θ, σ=0.56λ, γ=0.5;
step2: generating a residual image; performing high-pass filtering operation on the input training sample image by using the 2D Gabor filter generated in the previous step to obtain a filtered residual image;
step3: training a deep convolutional neural network; respectively training a deep convolutional neural network by using the filtered residual images;
step4: extracting learning type steganography detection characteristics; and respectively carrying out convolution operation on the input image and the 2D Gabor filter, respectively inputting the obtained residual images into corresponding deep convolution neural networks, taking the characteristics output by each deep convolution neural network, and finally combining the obtained multiple learning type steganography detection characteristics into a final learning type steganography detection characteristic.
5. The Gabor filtering and convolutional neural network-based image steganography detection method of claim 4, wherein the steps of: the generation steps of the 2D Gabor filter are as follows: (1) generating sampling points; assuming that the 2D Gabor filter size is M×N, the value range of x is
The value range of y is +.>
Generating sampling points (x, y) in step size 1; (2) determining filter parameters; determination of the parameters sigma, theta and +.in the 2DGabor function>
Calculating the parameter lambda value, gamma=0.5; (3) generating a 2D Gabor filter; according to the 2D Gabor function expression and the parameter value, calculating the function value at the sampling point to obtain the corresponding 2D Gan abor filter; (4) normalizing the 2D Gabor filter; subtracting the average value of all its elements from the 2D Gabor filter yields a zero-average high-pass filter. />
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811583343.XA CN109859091B (en) | 2018-12-24 | 2018-12-24 | Image steganography detection method based on Gabor filtering and convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811583343.XA CN109859091B (en) | 2018-12-24 | 2018-12-24 | Image steganography detection method based on Gabor filtering and convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109859091A CN109859091A (en) | 2019-06-07 |
| CN109859091B true CN109859091B (en) | 2023-05-16 |
Family
ID=66892030
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811583343.XA Active CN109859091B (en) | 2018-12-24 | 2018-12-24 | Image steganography detection method based on Gabor filtering and convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109859091B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110245621B (en) * | 2019-06-17 | 2023-10-17 | 深圳Tcl新技术有限公司 | Face recognition device, image processing method, feature extraction model, and storage medium |
| CN110490265B (en) * | 2019-08-23 | 2022-04-15 | 安徽大学 | Image steganalysis method based on double-path convolution and feature fusion |
| CN110619594B (en) * | 2019-09-16 | 2023-02-03 | 中山大学 | Halftone image steganalysis method |
| CN111859897A (en) * | 2019-10-16 | 2020-10-30 | 沈阳工业大学 | Text steganalysis method based on dynamic routing capsule network |
| CN111028308B (en) * | 2019-11-19 | 2022-11-04 | 珠海涵辰科技有限公司 | Steganography and reading method for information in image |
| CN111507884A (en) * | 2020-04-19 | 2020-08-07 | 衡阳师范学院 | Self-adaptive image steganalysis method and system based on deep convolutional neural network |
| CN115147501B (en) * | 2022-09-05 | 2022-12-02 | 深圳市明源云科技有限公司 | Picture decompression method and device, terminal device and storage medium |
| CN115731089B (en) * | 2022-12-16 | 2024-02-23 | 中国人民解放军61660部队 | Dual-task image steganography method based on component energy |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101151622A (en) * | 2005-01-26 | 2008-03-26 | 新泽西理工学院 | System and method for steganalysis |
| CN108961137A (en) * | 2018-07-12 | 2018-12-07 | 中山大学 | A kind of image latent writing analysis method and system based on convolutional neural networks |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5206024B2 (en) * | 2008-02-28 | 2013-06-12 | 富士通株式会社 | Image decryption device, image encryption device, image decryption method, and image decryption program |
-
2018
- 2018-12-24 CN CN201811583343.XA patent/CN109859091B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101151622A (en) * | 2005-01-26 | 2008-03-26 | 新泽西理工学院 | System and method for steganalysis |
| CN108961137A (en) * | 2018-07-12 | 2018-12-07 | 中山大学 | A kind of image latent writing analysis method and system based on convolutional neural networks |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109859091A (en) | 2019-06-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109859091B (en) | Image steganography detection method based on Gabor filtering and convolutional neural network | |
| CN111047551B (en) | Remote sensing image change detection method and system based on U-net improved algorithm | |
| CN113222800B (en) | Robust image watermark embedding and extracting method and system based on deep learning | |
| CN112396607B (en) | Deformable convolution fusion enhanced street view image semantic segmentation method | |
| CN108805002B (en) | Monitoring video abnormal event detection method based on deep learning and dynamic clustering | |
| CN112907598B (en) | Method for detecting falsification of document and certificate images based on attention CNN | |
| CN107506765B (en) | License plate inclination correction method based on neural network | |
| CN113538457B (en) | Video semantic segmentation method utilizing multi-frequency dynamic hole convolution | |
| CN110490265A (en) | A kind of image latent writing analysis method based on two-way convolution sum Fusion Features | |
| CN112862655B (en) | JPEG image steganalysis method based on channel space attention mechanism | |
| CN117274059A (en) | Low-resolution image reconstruction method and system based on image coding-decoding | |
| CN113344110B (en) | Fuzzy image classification method based on super-resolution reconstruction | |
| CN106251375A (en) | A kind of degree of depth study stacking-type automatic coding of general steganalysis | |
| Liang et al. | Image resampling detection based on convolutional neural network | |
| Rana et al. | MSRD-CNN: Multi-scale residual deep CNN for general-purpose image manipulation detection | |
| CN116385281A (en) | Remote sensing image denoising method based on real noise model and generated countermeasure network | |
| Yu et al. | A multi-task learning CNN for image steganalysis | |
| CN113392728B (en) | Target detection method based on SSA sharpening attention mechanism | |
| CN117726954B (en) | Sea-land segmentation method and system for remote sensing image | |
| CN112418229A (en) | Unmanned ship marine scene image real-time segmentation method based on deep learning | |
| CN115994849B (en) | Three-dimensional digital watermark embedding and extracting method based on point cloud up-sampling | |
| CN102881008A (en) | Circular loop statistic characteristic-based anti-rotation image Hash method | |
| CN114926885A (en) | Strong generalization depth counterfeit face detection method based on local anomaly | |
| Huang et al. | Detection and classification of various image operations using deep learning technology | |
| Kaushik et al. | Digital Image Forgery Detection using Local Binary Patterns (LBP) and Histogram of Oriented Gradients (HOG) |
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 |