CN107563155A - A kind of safe steganography method and device based on generation confrontation network - Google Patents
A kind of safe steganography method and device based on generation confrontation network Download PDFInfo
- Publication number
- CN107563155A CN107563155A CN201710670786.1A CN201710670786A CN107563155A CN 107563155 A CN107563155 A CN 107563155A CN 201710670786 A CN201710670786 A CN 201710670786A CN 107563155 A CN107563155 A CN 107563155A
- Authority
- CN
- China
- Prior art keywords
- network
- generation
- image
- steganography
- carrier image
- 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.)
- Granted
Links
Landscapes
- Editing Of Facsimile Originals (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to a kind of safe steganography method and device based on generation confrontation network.This method includes:In the case where generation resists network frame, the carrier image of embedding information is wanted by generating network generation, by differentiating that network is judged the true and false property of the carrier image of generation;Pass through the dynamic game process for generating network and differentiating network so that the carrier image of generation network generation is close to true picture;The insertion of row information is entered to the carrier image of generation network generation;Then two classification are carried out to the image after the carrier image and steganography of input using steganalysis network, obtains being categorized as the accuracy rate of artwork and steganography figure.The carrier image that the present invention generates is visually closer to true picture, and the speed generated is faster, it is possible to increase the security of steganography.
Description
Technical field
The invention belongs to areas of information technology, are related to information steganography technology, and in particular to one kind is based on generation confrontation network
Safe steganography method and device.
Background technology
Information steganography technology, it is one of Main Branches of Information hiding.Information hiding is exactly to utilize mankind's sense organ pair
The sensation redundancy of data signal, one group of secret information (authorization sequence number, message or copyright information etc.) is hidden into carrier information
In, in the case where not influenceing the sensory effect of host signal and use value so that possible attacker is difficult to therefrom judge
Secret information whether there is, and more be difficult to intercept and capture, so as to ensure the security of information transmission.With the development of science and technology, believe
Cease the concealing technology extensive use full of vitality as new study hotspot, especially digital media technology again so that information
It has been concealed with further development and broader intension.Digital watermarking, digital signature for copyright protection etc. also by
Bring the category of Information hiding into, the present invention is concerned with traditional Information hiding i.e. steganography (Steganography)
Application in Digital Media, the steganography mainly using image as carrier.
At present, when people design steganographic algorithm, generally consider heuristically in terms of steganalysis.It is for example, secret
Close message should be embedded into safer picture noise and texture region.In Denis and Burnaev (ICLR 2016Open
Review, 2016) in work, DCGAN (Deep Convolutional Generative Adversarial are used
Networks, depth convolution generation confrontation network) to generate artwork, and detect steganography using single steganalysis network
Effect, this work is carried out under the framework of confrontation study, utilizes DCGAN's to differentiate network with generating the confrontation of network
Property, the generation of generation network is set to be adapted to the image-carrier for doing steganography, and generated using independent steganalysis network evaluation
Whether image-carrier can do steganography.
The method proposed in Denis and Burnaev (ICLR 2016Open Review, 2016) has some limitations, real
Test and show that its steganography method is not safe enough.Network in this article is applied to embedded identical key, when using random key, net
Network differentiates ineffective.In addition, the DeGrain of its steganalysis network.
The content of the invention
It is an object of the invention to provide a kind of safe steganography method and device based on generation confrontation network, the load of generation
Body image is visually closer to true picture, and the speed generated is faster, it is possible to increase the security of steganography.
The present invention firstly generates carrier image, and secret information is embedded in into carrier image by offline mode, then will
Image after embedding information is input to steganalysis network i.e. GNCNN (Gaussian-Neuron together with the carrier image of generation
Convolutional Neural Network, Gauss-neuron convolutional neural networks) in network, the steganalysis network pair
Original vector image and hidden image carry out discriminant classification.Then, HUGO (Highly Undetectable steGO, height are reused
Spend undetectable steganography figure) algorithm, embed of information into the carrier image of generation, recycle steganalysis network to be sentenced
Not.Not only fixed key is embedded in effectively, also random key is embedded in effective so as to demonstrate the method for the present invention.
The technical solution adopted by the present invention is as follows:
A kind of safe steganography method based on generation confrontation network, comprises the following steps:
1) in the case where generation resists network frame, the carrier image of embedding information is wanted by generating network generation, passes through differentiation
Network is judged the true and false property of the carrier image of generation;
2) by generating network and differentiating the dynamic game process of network so that the carrier image of generation network generation approaches
True picture;
3) insertion of row information is entered to the carrier image of generation network generation.
Further, the generation confrontation network is WGAN.
Further, the dynamic game process of the generation network and differentiation network includes:
A) for generating network, make its caused data as consistent with True Data as possible, embedding information is wanted so as to generate
Carrier image;
B) differentiate network area it is mitogenetic into carrier image and true picture, for differentiate network prediction result, to gradient
The direction of change is changed;
C) generation network obtains differentiating gradient and the undated parameter that network is passed back, generates new carrier image.
Further, the telescopiny of information including the use of fixed key telescopiny and uses random key in step 3)
Telescopiny.
Further, it is described to be embedded in using fixed key telescopiny using LSB methods;It is described to use random key
Telescopiny be embedded in using HUGO methods.
Further, two classification are carried out to the image after the carrier image and steganography of input using steganalysis network, obtained
To the accuracy rate for being categorized as artwork and steganography figure.
Further, the steganalysis network is GNCNN.
Further, compared by setting different seeds under different parameters, the carrier image of generation is by letter
To the duplicity of steganalysis network after breath is embedded, so as to improve the security of information insertion.
A kind of safe steganography device based on generation confrontation network, including:
NE is generated, under resisting network frame in generation, the carrier image of embedding information is wanted in generation;
Differentiating NE, the true and false property for the carrier image to generation judges, and by with generating network list
The dynamic game process of member so that the carrier image of generation NE generation is close to true picture;
Steganography unit, for entering the insertion of row information to the carrier image of generation NE generation.
Further, in addition to steganalysis NE, enter for the image after the carrier image to input and steganography
Row two is classified, and obtains being categorized as the accuracy rate of artwork and steganography figure.
Using the present invention method the image being embedded in through random key can be worked, compared with prior art have with
Lower advantage:
1st, the present invention is changed to Denis and Burnaev (ICLR 2016Open Review, 2016) method proposed
Enter, solve the problems, such as the image progress steganalysis DeGrain to being embedded in using random key in its work;
2nd, the present invention using WGAN (Wasserstein Generative Adversarial Networks,
Wasserstein generation confrontation network) to replace DCGAN, and image is generated using WGAN, the image of generation is in visual aspects
Effect is more preferable, closer true picture, and the training time is reduced, and training speed is faster;
3rd, differentiate that network by being confronted with each other with generation network, makes the image of generation be more beneficial for being embedded in, improves steganography
The security of image insertion;Differentiate that network is similar to steganalysis network structure, but function is different, and the network is the same as generation network pair
It is anti-so that the image of generation is truer, is more suitable for doing insertion;
4th, it is hidden if appropriate for doing using a more complicated steganalysis network G NCNN to assess the image of generation by the present invention
Write, the steganalysis network of use is it was proved that more preferable to the steganalysis effect of hidden image.
Brief description of the drawings
Fig. 1 is to carry out image latent writing and the flow chart differentiated using the inventive method.Wherein " data prediction " is
Refer to and unified trimming operation (for example carrying out center cropped operations, be uniformly cut into 64 × 64 size) is carried out to picture.
Fig. 2 is the image generated in the inventive method by generation network.
Fig. 3 is using the flow chart of embedding grammar during the inventive method progress image latent writing, wherein (a) figure is close using fixation
Key embedding grammar, (b) figure use random key embedding grammar.
Embodiment
Below by specific embodiments and the drawings, the present invention is described in further details.
Steganography method provided by the invention based on generation confrontation network, is suitable for use with random key and identical key
It is embedded in, the flow of this method is as shown in figure 1, its key step includes:Embedding information is wanted by the generation of generation network first
Carrier image, the true and false property of the carrier image of generation is judged by WGAN differentiation network, and by generating network and sentencing
The dynamic game process of other network, the carrier image for allowing the target of generation network to be to try to generate closer to true picture go to cheat
Differentiate network, differentiate that the target of network is to try to the carrier image and true picture of the generation of generation network to distinguish.Then it is right
The carrier image of generation network generation enters the insertion (i.e. steganography) of row information, the image (bag using steganalysis network to input
Containing the image after carrier image and steganography) two classification are carried out, obtain being categorized as artwork (i.e. carrier image) and steganography figure (i.e. steganography
Image afterwards) accuracy rate.
Wherein, generating the dynamic game process of network and differentiation network is:
1st, for generating network, to allow its caused data as consistent with True Data as possible, be exactly same data point
Cloth, generate the carrier image of steganography;
2nd, differentiate that e-learning distinguishes the carrier image and true picture of generation, the prediction result for differentiating network is right
The direction of graded is changed, when the output that differentiation network is thought to generate network is True Data and thinks that its output is to make an uproar
When sound data, differentiating the gradient updating direction of network will be changed;The gradient updating direction refers to object function
First derivative negative direction;
3rd, generation network obtains differentiating the gradient that network is passed back, undated parameter, generates new carrier image.Due to using
WGAN, be not in GAN gradient disappearance problem, network can be with continuous updating.Fig. 2 is the example images for generating network generation
Figure.
The process of the inventive method embedding information is carried out offline, is separated with the training process of network.LSB is used first
(Least Significant Bit, least significant bit) method is embedded in, and steganalysis knot is obtained by steganalysis network
After fruit, reuse HUGO (Highly Undetectable steGO, highly undetectable steganography figure) method and be embedded in,
And steganalysis is carried out by steganalysis network.During image is generated, by setting different experiment parameters, compare
Degree of the image generated under different parameters to steganalysis network cheating.
During above-mentioned embedding information, it is carried out separately using LSB and the HUGO process being embedded in twice, LSB is to adopt
Fixed key embedded mode, and HUGO is embedded in using random key, the method for thus demonstrating the present invention is not only right
Fixed key insertion is effective, and also random key is embedded in effectively, more demonstrates effectiveness of the invention.Fig. 3 illustrates to use
The overall procedure of two kinds of embedding grammars of LSB and HUGO, wherein (a) figure uses fixed key embedding grammar, the use of (b) figure is with secret
Key embedding grammar.
Steganalysis network in this method, employ GNCNN (Gaussian-Neuron Convolutional
Neural Network, Gauss-neuron convolutional neural networks), referring to " In IS&T/SPIE Electronic Imaging,
pp.94090J–94090J.”.GNCNN proposes one kind and divided by deep learning model come automatic learning characteristic to carry out steganography
The new method of analysis, the model proposed can capture the complicated dependence to steganalysis useful feature.With it is existing
Steganalysis mechanism is compared, and the model can use the automatic learning characteristic of several convolutional layers to represent.Feature extraction and the step of classification
Suddenly unitized under same architecture, it means that during feature extraction, can be instructed using classification.
Safe steganography method of the example 1 based on generation confrontation network
By taking CelebA human face data collection as an example:
1) in WGAN (Wasserstein Generative Adversarial Networks, Wasserstein generations
Resist network) under framework, noise signal z is input to generation network G, carrier image I is generated by generation network G;
2) network D and the generation mutual game of network G are differentiated so that the carrier image I of G generations is closer to true picture;
3) differentiate that the carrier image I that network D generates to G carries out discriminant analysis, obtain being judged as Zhen Tu and false figure probability;
4) by the offline embedding informations of carrier image I of generation, it is embedded in first by LSB fixed keys, after obtaining steganography
Image I ', by I and I ' while steganalysis network S is input to, judgement is the probability of artwork or steganography figure;
5) and then using HUGO it is randomly-embedded, obtains the image I " after steganography, by I and I " while is input to steganalysis net
Network S, judgement are the probability of artwork or steganography figure;
6) in DCGAN (Deep Convolutional Generative Adversarial Networks, depth convolution
Generation confrontation network) under framework, repeat 1)~5) the step of, artwork and steganography figure are divided by steganalysis network
Class, obtain a probability;
7) probability under two methods is compared, obtained classification accuracy is as shown in table 1:
Table 1. trains the classification accuracy that steganalysis network obtains on true picture
Image type | The inventive method | SGANs |
Artwork | 0.87 | 0.92 |
The image of generation | 0.72 | 0.90 |
In table 1, SGANs is the method for contrast, and the method is realized using DCGAN;" image of generation " refers to generate net
The carrier image of network generation.The experiment is to do steganography respectively in artwork and the carrier image of generation, then carries out steganalysis,
Obtain classification accuracy.After repeatedly training test, method proposed by the present invention is in the classification accuracy to generating image
It is lower, that is, it is not easy the picture of generation and true picture to distinguish, so as to which security is higher compared with the method for contrast.
Safe steganography method of the example 2 based on generation confrontation network
By taking CelebA human face data collection as an example, in this experiment, by setting different seeds, to compare in difference
Under parameter, the image of generation is after the insertion of information to the duplicity of steganalysis network.The seed refers to random
In the case that number is certain, the repeatability of experiment is controlled, is presented as the randomness of the image of control generation in the present invention.
1) under WGAN frameworks, noise signal z is input to generation network G, carrier image I is generated by generation network G;
2) differentiate network D and the generation mutual game of network G, generate the carrier image I closer to true picture;
3) differentiate that network D carries out discriminant analysis to the carrier image I of generation, obtain being judged as Zhen Tu and false figure probability;
4) identical seed is used, to the carrier image I of generation, first by LSB method embedding informations, obtains steganography
Image I ' afterwards, by I and I ' while steganalysis network S is input to, judgement is the probability of artwork or steganography figure;
5) and then using HUGO it is randomly-embedded, obtains the image I " after steganography, by I and I " while is input to steganalysis net
Network S, judgement are the probability of artwork or steganography figure;
6) using the seed generated at random, after the carrier image I embedding informations of generation, the image T after steganography is obtained,
I and T is input to steganalysis network S simultaneously, judgement is the probability of artwork or steganography figure;
7) using the seed generated at random, and WGAN is finely adjusted during training, adjusts its learning rate
And Momentum coefficients etc., after generation image embedding information, the image M after steganography is obtained, I and M is input to simultaneously hidden
Analysis network S is write, judgement is the probability of artwork or steganography figure;
8) classification accuracy under three kinds of Setup Experiments is compared, as shown in table 2:
For table 2. under different Setup Experiments, steganalysis network trains obtained classification accuracy on generation image
Experiment condition | Classification accuracy |
4) | 0.87 |
6) | 0.72 |
7) | 0.71 |
In table 2,4), 6), 7) represent respectively step 4), 6), 7) in Setup Experiments.Test result indicates that scheme in generation
As during, by setting different seeds, lower classification accuracy can be obtained, you can deception steganalysis net
Network, prove that the image of the inventive method generation is more suitable for doing steganography with this, security is higher.
Safe steganography device of the example 3 based on generation confrontation network
The safe steganography device based on generation confrontation network includes:NE is generated, for resisting network in generation
Under framework, the carrier image of embedding information is wanted in generation;Differentiate NE, the true and false property for the carrier image to generation is carried out
Judge, and pass through the dynamic game process with generating NE so that the carrier image of generation NE generation is close true
Real image;Steganography unit, for entering the insertion of row information to the carrier image of generation NE generation.Further, the dress
Putting also includes steganalysis NE, carries out two classification for the image after the carrier image to input and steganography, is divided
Class is artwork and the accuracy rate of steganography figure.
The present invention is carried out under WGAN (Wasserstein generation confrontation network) framework.For regarding for generation image
Feel quality, generation network and differentiation network can be replaced with other generation confrontation networks that generation picture quality is high, generating rate is fast
Generation;For steganography method, the present invention has carried out the experiment of fixed key insertion and random key insertion, passes through the reality of two methods
Result is tested, embodies the validity of steganography method proposed by the invention, while can also enter using other adaptive Steganographies
Row experiment;For steganalysis method, present invention preferably uses GNCNN networks, there is higher steganalysis verification and measurement ratio.
The above embodiments are merely illustrative of the technical solutions of the present invention rather than is limited, the ordinary skill of this area
Technical scheme can be modified by personnel or equivalent substitution, without departing from the spirit and scope of the present invention, this
The protection domain of invention should be to be defined described in claims.
Claims (10)
1. a kind of safe steganography method based on generation confrontation network, it is characterised in that comprise the following steps:
1) in the case where generation resists network frame, the carrier image of embedding information is wanted by generating network generation, by differentiating network
The true and false property of the carrier image of generation is judged;
2) by generating network and differentiating the dynamic game process of network so that the carrier image of generation network generation is close true
Image;
3) insertion of row information is entered to the carrier image of generation network generation.
2. the method as described in claim 1, it is characterised in that the generation confrontation network is WGAN.
3. the method as described in claim 1, it is characterised in that the generation network and the dynamic game process bag for differentiating network
Include:
A) for generating network, make its caused data as consistent with True Data as possible, so as to generate the load for wanting embedding information
Body image;
B) differentiate network area it is mitogenetic into carrier image and true picture, for differentiate network prediction result, to graded
Direction be changed;
C) generation network obtains differentiating gradient and the undated parameter that network is passed back, generates new carrier image.
4. the method as described in claim 1, it is characterised in that the telescopiny of information is including the use of fixed key in step 3)
Telescopiny and the telescopiny using random key.
5. method as claimed in claim 4, it is characterised in that described to be entered using fixed key telescopiny using LSB methods
Row insertion;The telescopiny using random key is embedded in using HUGO methods.
6. the method as described in any claim in claim 1 to 5, it is characterised in that using steganalysis network to defeated
Image after the carrier image and steganography that enter carries out two classification, obtains being categorized as the accuracy rate of artwork and steganography figure.
7. method as claimed in claim 6, it is characterised in that the steganalysis network is GNCNN.
8. method as claimed in claim 6, it is characterised in that compared by setting different seeds in different parameters
Under, the carrier image of generation after information is embedded in the duplicity of steganalysis network, so as to improve the safety of information insertion
Property.
A kind of 9. safe steganography device based on generation confrontation network, it is characterised in that including:
NE is generated, under resisting network frame in generation, the carrier image of embedding information is wanted in generation;
Differentiating NE, the true and false property for the carrier image to generation judges, and by with generating NE
Dynamic game process so that the carrier image of generation NE generation is close to true picture;
Steganography unit, for entering the insertion of row information to the carrier image of generation NE generation.
10. device as claimed in claim 9, it is characterised in that also including steganalysis NE, for the load to input
Image after body image and steganography carries out two classification, obtains being categorized as the accuracy rate of artwork and steganography figure.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710670786.1A CN107563155B (en) | 2017-08-08 | 2017-08-08 | Security steganography method and device based on generation of countermeasure network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710670786.1A CN107563155B (en) | 2017-08-08 | 2017-08-08 | Security steganography method and device based on generation of countermeasure network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107563155A true CN107563155A (en) | 2018-01-09 |
CN107563155B CN107563155B (en) | 2023-02-28 |
Family
ID=60973962
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710670786.1A Active CN107563155B (en) | 2017-08-08 | 2017-08-08 | Security steganography method and device based on generation of countermeasure network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107563155B (en) |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108154547A (en) * | 2018-01-17 | 2018-06-12 | 百度在线网络技术(北京)有限公司 | Image generating method and device |
CN108346125A (en) * | 2018-03-15 | 2018-07-31 | 中山大学 | A kind of spatial domain picture steganography method and system based on generation confrontation network |
CN108596141A (en) * | 2018-05-08 | 2018-09-28 | 深圳大学 | A kind of depth network generates the detection method and system of facial image |
CN108648135A (en) * | 2018-06-01 | 2018-10-12 | 深圳大学 | Hide model training and application method, device and computer readable storage medium |
CN108765246A (en) * | 2018-04-03 | 2018-11-06 | 宁波大学 | A kind of selection method of steganographic system carrier image |
CN109002686A (en) * | 2018-04-26 | 2018-12-14 | 浙江工业大学 | A kind of more trade mark chemical process soft-measuring modeling methods automatically generating sample |
CN109214973A (en) * | 2018-08-24 | 2019-01-15 | 中国科学技术大学 | For the confrontation safety barrier generation method of steganalysis neural network |
CN109348211A (en) * | 2018-08-06 | 2019-02-15 | 中国科学院声学研究所 | The general information of interframe encode hides detection method in a kind of video frame |
CN109359667A (en) * | 2018-09-07 | 2019-02-19 | 华南理工大学 | A kind of feature recalibration convolution method based on WGAN model |
CN109993678A (en) * | 2019-03-26 | 2019-07-09 | 南京联创北斗技术应用研究院有限公司 | It is a kind of to fight the robust steganography method for generating network based on depth |
CN110084734A (en) * | 2019-04-25 | 2019-08-02 | 南京信息工程大学 | A kind of big data ownership guard method being locally generated confrontation network based on object |
CN110334805A (en) * | 2019-05-05 | 2019-10-15 | 中山大学 | A kind of JPEG domain image latent writing method and system based on generation confrontation network |
CN110457910A (en) * | 2018-05-07 | 2019-11-15 | 中国人民武装警察部队工程大学 | A kind of production information concealing method based on image synthesis |
CN111768325A (en) * | 2020-04-03 | 2020-10-13 | 南京信息工程大学 | Security improvement method based on generation of countermeasure sample in big data privacy protection |
CN111768326A (en) * | 2020-04-03 | 2020-10-13 | 南京信息工程大学 | High-capacity data protection method based on GAN amplification image foreground object |
CN112115490A (en) * | 2020-08-14 | 2020-12-22 | 宁波大学 | Carrier image synthesis steganography method based on GAN |
CN112395635A (en) * | 2021-01-18 | 2021-02-23 | 北京灵汐科技有限公司 | Image processing method, device, secret key generating method, device, training method and device, and computer readable medium |
CN114339258A (en) * | 2021-12-28 | 2022-04-12 | 中国人民武装警察部队工程大学 | Information steganography method and device based on video carrier |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105979268A (en) * | 2016-05-05 | 2016-09-28 | 北京智捷伟讯科技有限公司 | Safe information transmission method based on lossless watermark embedding and safe video hiding |
CN106920206A (en) * | 2017-03-16 | 2017-07-04 | 广州大学 | A kind of steganalysis method based on confrontation neutral net |
-
2017
- 2017-08-08 CN CN201710670786.1A patent/CN107563155B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105979268A (en) * | 2016-05-05 | 2016-09-28 | 北京智捷伟讯科技有限公司 | Safe information transmission method based on lossless watermark embedding and safe video hiding |
CN106920206A (en) * | 2017-03-16 | 2017-07-04 | 广州大学 | A kind of steganalysis method based on confrontation neutral net |
Cited By (29)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108154547A (en) * | 2018-01-17 | 2018-06-12 | 百度在线网络技术(北京)有限公司 | Image generating method and device |
CN108346125A (en) * | 2018-03-15 | 2018-07-31 | 中山大学 | A kind of spatial domain picture steganography method and system based on generation confrontation network |
CN108346125B (en) * | 2018-03-15 | 2021-10-08 | 中山大学 | Airspace image steganography method and system based on generation countermeasure network |
CN108765246A (en) * | 2018-04-03 | 2018-11-06 | 宁波大学 | A kind of selection method of steganographic system carrier image |
CN108765246B (en) * | 2018-04-03 | 2019-07-16 | 宁波大学 | A kind of selection method of steganographic system carrier image |
CN109002686A (en) * | 2018-04-26 | 2018-12-14 | 浙江工业大学 | A kind of more trade mark chemical process soft-measuring modeling methods automatically generating sample |
CN109002686B (en) * | 2018-04-26 | 2022-04-08 | 浙江工业大学 | Multi-grade chemical process soft measurement modeling method capable of automatically generating samples |
CN110457910A (en) * | 2018-05-07 | 2019-11-15 | 中国人民武装警察部队工程大学 | A kind of production information concealing method based on image synthesis |
CN108596141A (en) * | 2018-05-08 | 2018-09-28 | 深圳大学 | A kind of depth network generates the detection method and system of facial image |
CN108648135A (en) * | 2018-06-01 | 2018-10-12 | 深圳大学 | Hide model training and application method, device and computer readable storage medium |
CN108648135B (en) * | 2018-06-01 | 2022-05-27 | 深圳大学 | Hidden model training and using method, device and computer readable storage medium |
CN109348211A (en) * | 2018-08-06 | 2019-02-15 | 中国科学院声学研究所 | The general information of interframe encode hides detection method in a kind of video frame |
CN109214973A (en) * | 2018-08-24 | 2019-01-15 | 中国科学技术大学 | For the confrontation safety barrier generation method of steganalysis neural network |
CN109214973B (en) * | 2018-08-24 | 2020-10-27 | 中国科学技术大学 | Method for generating countermeasure security carrier aiming at steganalysis neural network |
CN109359667A (en) * | 2018-09-07 | 2019-02-19 | 华南理工大学 | A kind of feature recalibration convolution method based on WGAN model |
CN109993678A (en) * | 2019-03-26 | 2019-07-09 | 南京联创北斗技术应用研究院有限公司 | It is a kind of to fight the robust steganography method for generating network based on depth |
CN109993678B (en) * | 2019-03-26 | 2020-04-07 | 南京联创北斗技术应用研究院有限公司 | Robust information hiding method based on deep confrontation generation network |
CN110084734A (en) * | 2019-04-25 | 2019-08-02 | 南京信息工程大学 | A kind of big data ownership guard method being locally generated confrontation network based on object |
CN110084734B (en) * | 2019-04-25 | 2023-02-14 | 南京信息工程大学 | Big data ownership protection method based on object local generation countermeasure network |
CN110334805B (en) * | 2019-05-05 | 2022-10-25 | 中山大学 | JPEG domain image steganography method and system based on generation countermeasure network |
CN110334805A (en) * | 2019-05-05 | 2019-10-15 | 中山大学 | A kind of JPEG domain image latent writing method and system based on generation confrontation network |
CN111768326A (en) * | 2020-04-03 | 2020-10-13 | 南京信息工程大学 | High-capacity data protection method based on GAN amplification image foreground object |
CN111768325A (en) * | 2020-04-03 | 2020-10-13 | 南京信息工程大学 | Security improvement method based on generation of countermeasure sample in big data privacy protection |
CN111768326B (en) * | 2020-04-03 | 2023-08-25 | 南京信息工程大学 | High-capacity data protection method based on GAN (gas-insulated gate bipolar transistor) amplified image foreground object |
CN112115490A (en) * | 2020-08-14 | 2020-12-22 | 宁波大学 | Carrier image synthesis steganography method based on GAN |
CN112115490B (en) * | 2020-08-14 | 2023-09-26 | 石坚 | GAN-based carrier image synthesis steganography method |
CN112395635A (en) * | 2021-01-18 | 2021-02-23 | 北京灵汐科技有限公司 | Image processing method, device, secret key generating method, device, training method and device, and computer readable medium |
CN114339258A (en) * | 2021-12-28 | 2022-04-12 | 中国人民武装警察部队工程大学 | Information steganography method and device based on video carrier |
CN114339258B (en) * | 2021-12-28 | 2024-05-10 | 中国人民武装警察部队工程大学 | Information steganography method and device based on video carrier |
Also Published As
Publication number | Publication date |
---|---|
CN107563155B (en) | 2023-02-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107563155A (en) | A kind of safe steganography method and device based on generation confrontation network | |
CN105518708B (en) | For verifying the method for living body faces, equipment and computer program product | |
CN108596141A (en) | A kind of depth network generates the detection method and system of facial image | |
CN105426827B (en) | Living body verification method, device and system | |
CN109977841A (en) | A kind of face identification method based on confrontation deep learning network | |
CN107330444A (en) | A kind of image autotext mask method based on generation confrontation network | |
CN109308450A (en) | A kind of face's variation prediction method based on generation confrontation network | |
CN107368752A (en) | A kind of depth difference method for secret protection based on production confrontation network | |
CN107886064A (en) | A kind of method that recognition of face scene based on convolutional neural networks adapts to | |
CN109492416A (en) | A kind of guard method of big data image and system based on safety zone | |
CN106778506A (en) | A kind of expression recognition method for merging depth image and multi-channel feature | |
CN108596026A (en) | Across the visual angle Gait Recognition device and training method of confrontation network are generated based on double fluid | |
CN108182409A (en) | Biopsy method, device, equipment and storage medium | |
CN106384093A (en) | Human action recognition method based on noise reduction automatic encoder and particle filter | |
CN104361548A (en) | BP neural network digital image compression based image watermark embedding and extracting method | |
CN109584162A (en) | A method of based on the image super-resolution reconstruct for generating network | |
CN106997452A (en) | Live body verification method and device | |
CN106821333A (en) | A kind of cognition dysfunction rehabilitation detection means based on virtual scene, method and therapeutic equipment | |
Zhou et al. | On security enhancement of steganography via generative adversarial image | |
CN107273794A (en) | Live body discrimination method and device in a kind of face recognition process | |
CN107092883A (en) | Object identification method for tracing | |
CN113642621A (en) | Zero sample image classification method based on generation countermeasure network | |
CN109508740A (en) | Object hardness identification method based on Gaussian mixed noise production confrontation network | |
CN107657634A (en) | Shale digital cores three-dimensional reconstruction method based on deep learning and SVMs | |
CN112949469A (en) | Image recognition method, system and equipment for face tampered image characteristic distribution |
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 |