CN113989097A - Information steganography model training method, information steganography device and storage medium - Google Patents

Information steganography model training method, information steganography device and storage medium Download PDF

Info

Publication number
CN113989097A
CN113989097A CN202111637371.7A CN202111637371A CN113989097A CN 113989097 A CN113989097 A CN 113989097A CN 202111637371 A CN202111637371 A CN 202111637371A CN 113989097 A CN113989097 A CN 113989097A
Authority
CN
China
Prior art keywords
steganography
information
loss function
image
function value
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
Application number
CN202111637371.7A
Other languages
Chinese (zh)
Other versions
CN113989097B (en
Inventor
崔新安
孙强
苗功勋
曲志峰
解荣昊
高伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing Zhongfu Information Technology Co Ltd
Original Assignee
Nanjing Zhongfu Information Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing Zhongfu Information Technology Co Ltd filed Critical Nanjing Zhongfu Information Technology Co Ltd
Priority to CN202111637371.7A priority Critical patent/CN113989097B/en
Publication of CN113989097A publication Critical patent/CN113989097A/en
Application granted granted Critical
Publication of CN113989097B publication Critical patent/CN113989097B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Biophysics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Image Processing (AREA)

Abstract

The application provides an information steganography model training method, an information steganography method, equipment and a storage medium, wherein the information steganography model training method comprises the following steps: the method comprises the steps of adopting a steganography network to embed preset steganography information of a plurality of image samples into the plurality of image samples respectively to obtain steganography image samples corresponding to the plurality of image samples, adopting an extraction network to process the steganography image samples to obtain recovery steganography information corresponding to the steganography image samples, adopting a decision network to process the plurality of image samples and the corresponding steganography image samples to obtain steganography evaluation values, calculating target loss function values according to the preset steganography information, the recovery steganography information and the steganography evaluation values, and updating parameters of an information steganography model according to the target loss function values to obtain a target information steganography model. Therefore, the target information steganography model can be used for steganography of information of various files, and the target information steganography model is high in compatibility and good in steganography effect.

Description

Information steganography model training method, information steganography device and storage medium
Technical Field
The present application relates to the field of computer technologies, and in particular, to an information steganography model training method, an information steganography device, and a storage medium.
Background
With the enhancement of data security awareness of people, more and more data security protection means exist at present to ensure the security of data files in the transmission process.
At present, hidden information embedding is carried out according to the format of a format file, and information hiding is carried out according to a mode of replacing words and phrases for a print output file or a word stock. However, the information steganography method is limited in the types of files, insufficient in compatibility, and poor in steganography effect.
Disclosure of Invention
An object of the present application is to provide an information steganography model training method, an information steganography device and a storage medium, aiming at overcoming the shortcomings in the prior art, so as to solve the problems of insufficient steganography compatibility and poor steganography effect in the prior art.
In order to achieve the above purpose, the technical solutions adopted in the embodiments of the present application are as follows:
in a first aspect, an embodiment of the present application provides an information steganography model training method, where the information steganography model includes: a steganographic network, an extraction network, and a decision network, the method comprising:
embedding preset steganography information of a plurality of image samples into the plurality of image samples respectively by adopting the steganography network to obtain steganography image samples corresponding to the plurality of image samples;
processing the steganographic image sample by adopting the extraction network to obtain recovery steganographic information corresponding to the steganographic image sample;
processing the plurality of image samples and the corresponding stego-images samples by adopting the judging network to obtain stego-evaluation values, wherein the stego-evaluation values are used for representing the quality of the stego-images samples, the judging network is used for extracting the image characteristics of the plurality of image samples and the image characteristics of the corresponding stego-images samples, and the stego-evaluation values are the characteristic differences of the image characteristics of the plurality of image samples and the image characteristics of the corresponding stego-images samples;
calculating a target loss function value according to the preset steganography information, the recovery steganography information and the steganography evaluation value;
and updating parameters of the information steganography model according to the target loss function value to obtain a target information steganography model.
Optionally, the calculating a target loss function value according to the preset steganographic information, the recovery steganographic information, and the steganographic evaluation value includes:
calculating a first loss function value according to the preset steganography information and the recovery steganography information;
calculating a second loss function value according to the first loss function value and the steganography evaluation value; the target loss function value comprises: the second loss function value;
the updating the parameters of the information steganography model according to the target loss function value to obtain a target information steganography model comprises the following steps:
and updating the parameters of the steganographic network and the extraction network according to the second loss function value.
Optionally, the calculating a second loss function value according to the first loss function value and the steganography evaluation value includes:
calculating the mean square error of the pixel values of the plurality of image samples and the corresponding stego-images samples in a plurality of color channels;
and calculating the second loss function value according to the mean square error, the first loss function value and the steganography evaluation value.
Optionally, after updating the parameters of the steganographic network and the extraction network according to the second loss function value, the method further includes:
and training the information steganography model after the parameters are updated according to the plurality of image samples until the second loss function value reaches a first loss condition.
Optionally, before the updating the parameters of the information steganography model according to the target loss function value to obtain the target information steganography model, the method further includes:
calculating a third loss function value of the decision network according to the plurality of image samples and the corresponding stego image samples; the target loss function value further comprises: the third loss function value;
the updating the parameters of the information steganography model according to the target loss function value to obtain a target information steganography model comprises the following steps:
and updating the parameters of the decision network according to the third loss function value.
Optionally, after updating the parameter of the decision network according to the third loss function value, the method further includes:
and training the information steganography model after the parameters are updated according to the plurality of image samples until the third loss function value reaches a second loss condition.
In a second aspect, another embodiment of the present application provides an information steganography method, including:
processing the file to be processed to obtain an image to be processed;
training by adopting any one of the training methods of the first aspect to obtain a steganographic network in a target information steganographic model, and embedding preset steganographic information of the image to be processed into the image to be processed to obtain a steganographic image of the image to be processed;
and generating a target file according to the steganographic image.
In a third aspect, another embodiment of the present application provides an information steganography model training apparatus, including:
the processing module is used for respectively embedding preset steganographic information of a plurality of image samples into the plurality of image samples by adopting the steganographic network to obtain steganographic image samples corresponding to the plurality of image samples;
processing the steganographic image sample by adopting the extraction network to obtain recovery steganographic information corresponding to the steganographic image sample;
processing the plurality of image samples and the corresponding stego-images samples by adopting the judging network to obtain stego-evaluation values, wherein the stego-evaluation values are used for representing the quality of the stego-images samples, the judging network is used for extracting the image characteristics of the plurality of image samples and the image characteristics of the corresponding stego-images samples, and the stego-evaluation values are the characteristic differences of the image characteristics of the plurality of image samples and the image characteristics of the corresponding stego-images samples;
the calculation module is used for calculating a target loss function value according to the preset steganography information, the recovery steganography information and the steganography evaluation value;
and the updating module is used for updating parameters of the information steganography model according to the target loss function value to obtain a target information steganography model.
Optionally, the calculation module is specifically configured to:
calculating a first loss function value according to the preset steganography information and the recovery steganography information;
calculating a second loss function value according to the first loss function value and the steganography evaluation value; the target loss function value comprises: the second loss function value;
the update module is specifically configured to:
and updating the parameters of the steganographic network and the extraction network according to the second loss function value.
Optionally, the calculation module is specifically configured to:
calculating the mean square error of the pixel values of the plurality of image samples and the corresponding stego-images samples in a plurality of color channels;
and calculating the second loss function value according to the mean square error, the first loss function value and the steganography evaluation value.
Optionally, the method further comprises:
and the training module is used for training the information steganography model after the parameters are updated according to the plurality of image samples until the second loss function value reaches a first loss condition.
Optionally, the computing module is further configured to:
calculating a third loss function value of the decision network according to the plurality of image samples and the corresponding stego image samples; the target loss function value further comprises: the third loss function value;
the update module is specifically configured to:
and updating the parameters of the decision network according to the third loss function value.
Optionally, the training module is further configured to:
and training the information steganography model after the parameters are updated according to the plurality of image samples until the third loss function value reaches a second loss condition.
In a fourth aspect, another embodiment of the present application provides an information steganography apparatus, including:
the processing module is used for processing the file to be processed to obtain an image to be processed;
training by adopting the training method of any one of the first aspect to obtain a steganographic network in a target information steganographic model, and embedding preset steganographic information of the image to be processed into the image to be processed to obtain a steganographic image of the image to be processed;
and the generating module is used for generating a target file according to the steganographic image.
In a fifth aspect, another embodiment of the present application provides an information steganography model training apparatus, including: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, the processor and the memory communicate via the bus when the information steganography model training apparatus is operated, and the processor executes the machine-readable instructions to perform the method of any one of the first aspect.
In a sixth aspect, another embodiment of the present application provides an information steganography device, including: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, the processor and the memory communicate via the bus when the information steganography device is operated, and the processor executes the machine-readable instructions to perform the method of the second aspect.
In a seventh aspect, another embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored, and the computer program is executed by a processor to perform the method of any one of the first and second aspects.
The beneficial effect of this application is:
the application provides an information steganography model training method, an information steganography method, equipment and a storage medium, wherein the information steganography model training method comprises the following steps: the method comprises the steps of adopting a steganography network to embed preset steganography information of a plurality of image samples into the plurality of image samples respectively to obtain steganography image samples corresponding to the plurality of image samples, adopting an extraction network to process the steganography image samples to obtain recovery steganography information corresponding to the steganography image samples, adopting a decision network to process the plurality of image samples and the corresponding steganography image samples to obtain steganography evaluation values, calculating target loss function values according to the preset steganography information, the recovery steganography information and the steganography evaluation values, and updating parameters of an information steganography model according to the target loss function values to obtain a target information steganography model. Therefore, the target information steganography model can be used for steganography of information of various files, and the target information steganography model is high in compatibility and good in steganography effect.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained from the drawings without inventive effort.
Fig. 1 is a first schematic flow chart of a method for training an information steganography model according to an embodiment of the present application;
fig. 2 is a schematic flow chart diagram of a second method for training an information steganography model according to an embodiment of the present application;
fig. 3 is a schematic flow chart diagram of a third method for training an information steganography model according to an embodiment of the present application;
fig. 4 is a fourth schematic flowchart of a method for training an information steganography model according to an embodiment of the present application;
fig. 5 is a schematic flowchart of an information steganography method according to an embodiment of the present application;
fig. 6 is a schematic structural diagram of an information steganography model training apparatus according to an embodiment of the present application;
fig. 7 is a schematic structural diagram of an information steganography apparatus according to an embodiment of the present application;
fig. 8 is a schematic structural diagram of an information steganography model training device provided in an embodiment of the present application;
fig. 9 is a schematic structural diagram of an information steganography device according to an embodiment of the present application.
Detailed Description
In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it should be understood that the drawings in the present application are for illustrative and descriptive purposes only and are not used to limit the scope of protection of the present application. Additionally, it should be understood that the schematic drawings are not necessarily drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of the present application. It should be understood that the operations of the flow diagrams may be performed out of order, and steps without logical context may be performed in reverse order or simultaneously. One skilled in the art, under the guidance of this application, may add one or more other operations to, or remove one or more operations from, the flowchart.
In addition, the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present application without making any creative effort, shall fall within the protection scope of the present application.
It should be noted that in the embodiments of the present application, the term "comprising" is used to indicate the presence of the features stated hereinafter, but does not exclude the addition of further features.
At present, hidden information embedding is carried out aiming at the format of a format file, information hiding is carried out aiming at printout or a word stock by adopting a mode of replacing words, but the information hiding protection of the format file such as embedding of a large segment of image is deficient; the compatibility of files of OFD and other domestic formats is not enough; the steganography effect on format files which are not printed and output in the traditional way or large-segment pictures is poor, or the information hiding effect cannot be embodied; the coverage of the whole flow of the file circulation is not enough.
Based on the information steganography model training method and the information steganography method, the format file is processed by adopting an antagonistic neural network, and invisible information of human eyes is embedded, so that the problems that the format file which is not printed and output in a traditional mode or has large-section pictures is poor in effect after steganography, is easy to interfere, covers the whole flow of file circulation, does not support a domestic format and the like are solved.
The following describes a message of the information steganography model training method provided by the present application with reference to several specific embodiments.
Fig. 1 is a schematic flow chart of a method for training an information steganography model according to an embodiment of the present disclosure, where an execution subject of the embodiment may be an information steganography model training device, such as an electronic device with data processing capability, e.g., a tablet computer, a notebook computer, and the like.
S101, embedding preset steganographic information of the image samples into the image samples respectively by adopting a steganographic network to obtain steganographic image samples corresponding to the image samples.
The information steganography model comprises: the image processing device comprises an steganographic network (an encoder), an extraction network (a decoder) and a determination network, wherein the steganographic network is used for embedding steganographic information into an image, the extraction network is used for extracting the steganographic information from the image embedded with the steganographic information, and the determination network is used for extracting image characteristics. The information steganography model can be an antagonistic neural network model.
Each image sample corresponds to preset steganographic information, which may be binary information, such as binary information obtained by binary encoding a user identity identifier.
Inputting a plurality of image samples into a steganography network, and respectively embedding preset steganography information of the image samples into the image samples by adopting the steganography network to obtain steganography image samples corresponding to the image samples, wherein the image samples and the steganography image patterns are in one-to-one correspondence.
It should be noted that the image sample can be obtained as follows:
step 1: setting a file operation interception hook in a path of file processing and circulation (such as double-click opening, printing, network transmission and the like) so as to intercept the format file in the path of file processing and circulation.
Step 2: extracting file resources of the layout file, including: extracting new image resources for embedding steganography from the format file, taking the image resources as image samples, or directly performing bitmap conversion on contents in the format file according to pages, taking the images after the bitmap conversion as the image samples, or blocking and bitmapping the pages of the format file, and taking the blocked images as the image samples, wherein the embedding rate can be improved when the steganography information is embedded according to subsequent blocks.
And S102, processing the steganographic image sample by adopting an extraction network to obtain recovery steganographic information corresponding to the steganographic image sample.
And inputting the steganographic image sample into an extraction network, and processing the steganographic image sample by adopting the extraction network to obtain recovery steganographic information corresponding to the steganographic image sample, wherein the recovery steganographic information is the steganographic information recovered from the steganographic image sample by the extraction network. The recovery steganographic information and steganographic image samples are in a one-to-one correspondence.
S103, processing the plurality of image samples and the corresponding steganographic image samples by adopting a decision network to obtain a steganographic evaluation value.
Respectively inputting a plurality of image samples and corresponding stego-images samples into a decision network, processing the plurality of image samples and the corresponding stego-images samples by adopting the decision network, extracting the image characteristics of the plurality of image samples and the image characteristics of the corresponding stego-images samples, wherein the image characteristics are represented in the form of characteristic values, extracting a characteristic matrix of the image samples by the decision network, performing pooling operation on the characteristic matrix of the image samples to obtain the characteristic values of the image samples and extracting the characteristic matrix of the stego-images samples, performing pooling operation on the characteristic matrix of the stego-images samples to obtain the characteristic values of the stego-images samples, then calculating the characteristic difference between the image characteristics of the image samples and the image characteristics of the corresponding stego-images samples according to the characteristic values of the image samples and the characteristic values of the corresponding stego-images samples, and taking the characteristic difference as the stego-evaluation value of the image samples and the corresponding stego-images samples, namely, the steganography evaluation value is the characteristic difference between the image characteristics of a plurality of image samples and the image characteristics of the corresponding steganography image samples, and the steganography evaluation value is used for representing the quality of the steganography image samples, namely, the steganography effect relative to the image samples, wherein the steganography evaluation value can be a value from 0 to 1.
And S104, calculating a target loss function value according to the preset steganography information, the recovered steganography information and the steganography evaluation value.
Each image sample corresponds to a preset steganography information, each steganography image sample corresponds to a recovery steganography information, and each image sample and the corresponding steganography image sample correspond to a steganography evaluation value.
In some embodiments, for each image sample and each steganographic image sample, an steganographic information loss value may be calculated according to preset steganographic information and recovery steganographic information, and then a target loss function value may be calculated according to all steganographic information loss values and all steganographic evaluation values, where the target loss function value may be a sum of all steganographic information loss values and all steganographic evaluation values, or may be a sum of a mean of all steganographic information loss values and a mean of all steganographic evaluation values, which is not limited by this embodiment.
And S105, updating parameters of the information steganography model according to the target loss function value to obtain the target information steganography model.
After the target loss function value is calculated, a parameter updating gradient corresponding to the target loss function can be calculated through a back propagation algorithm, the information steganography model is subjected to parameter updating according to the parameter updating gradient, then the steps S102-S104 are executed in a circulating mode until the target loss function meets certain conditions, and then the information steganography model corresponding to the target loss function meeting the conditions is determined to be the target information steganography model.
Therefore, feedback can be provided for the steganographic network through the steganographic network, the extraction network and the judgment network, and the purpose of counterstudy is achieved.
In the information steganography model training method of this embodiment, a steganography network is used to embed preset steganography information of a plurality of image samples into the plurality of image samples respectively to obtain steganography image samples corresponding to the plurality of image samples, an extraction network is used to process the steganography image samples to obtain recovered steganography information corresponding to the steganography image samples, a decision network is used to process the plurality of image samples and the corresponding steganography image samples to obtain steganography evaluation values, a target loss function value is calculated according to the preset steganography information, the recovered steganography information and the steganography evaluation values, and a parameter of an information steganography model is updated according to the target loss function value to obtain a target information steganography model. Therefore, the target information steganography model can be used for steganography of information of various files, and the target information steganography model is high in compatibility and good in steganography effect.
As an example, the steganographic network consists of 4 convolutional layers, the extraction network consists of 4 convolutional layers, and the decision network consists of 3 output channels
Figure M_211229101853422_422263001
And a convolutional layer of output channel 1, wherein,
Figure M_211229101853484_484827002
indicating the number of hidden channels in the network.
The steganographic image implementation process is as follows:
step 1: original image
Figure M_211229101853516_516049001
Has a width of
Figure M_211229101853562_562900002
Height of
Figure M_211229101853613_613226003
The original image is processed
Figure M_211229101853675_675498004
R, G, B to respectively expand to
Figure M_211229101853706_706960005
On the matrix of (A) to obtain a structure of
Figure M_211229101853738_738236006
Is shown, where 3 represents the three color channels of RGB (i.e., three channels in the depth direction).
Step 2: on the first layer of convolution layer, to tensor
Figure M_211229101853802_802139001
Convolution processing is carried out to obtain a structure of
Figure M_211229101853818_818266002
Tensor of
Figure M_211229101853865_865154003
I.e. by
Figure M_211229101853896_896416004
Expressing that the tensor Dsource in the depth direction is convoluted to generate the number of channels in the depth direction as
Figure M_211229101853927_927667005
Tensor of
Figure M_211229101853958_958913006
。
Wherein information is steganographically
Figure M_211229101853990_990166001
May be binary information that may be in binary form converting an utf-8 encoded string to a corresponding ASCII code, or compressed into binary information using a compression algorithm. Wherein the binary information is encoded as an array comprising 0 or 1,
Figure M_211229101854008_008189002
the number of bits of information, i.e. the length of the array, is hidden for the final binary.
And step 3: adding 4 bytes (32 0) as a separation at the tail of the binary information array, and circularly filling
Figure M_211229101854039_039974001
In an array of lengths, the excess data will be truncated. Will array according to
Figure M_211229101854071_071217002
Is divided into
Figure M_211229101854102_102471003
Group, generating the structure of
Figure M_211229101854118_118074004
Tensor of
Figure M_211229101854149_149340005
。
And 4, step 4: on the second layer of the convolution layer, will
Figure M_211229101854180_180585001
And
Figure M_211229101854217_217219002
performing connection and convolution processing to obtain a structure of
Figure M_211229101854232_232843003
Tensor of
Figure M_211229101854264_264097004
Wherein, in the step (A),
Figure M_211229101854295_295345005
to represent
Figure M_211229101854326_326644006
And
Figure M_211229101854357_357822007
the connection tensor in the depth direction, R, is a tensor having:
Figure M_211229101854401_401731001
hidden_size×W×H
Figure M_211229101854433_433586002
D×W×H
Figure M_211229101854496_496015003
(hideen_size+D)×W×H
then:
Figure M_211229101854527_527250001
representing the number of channels in the depth direction as
Figure M_211229101854558_558490001
Is/are as follows
Figure M_211229101854607_607290002
The tensor is convoluted to generate channels with the number in the depth direction
Figure M_211229101854639_639101003
Tensor of
Figure M_211229101854670_670405004
。
And 5: using dense connections to increase the embedding rate, on the third layer of the convolutional layer, pair
Figure M_211229101854701_701141001
Is convolved to obtain
Figure M_211229101854748_748450002
，
Figure M_211229101854764_764861003
To represent
Figure M_211229101854796_796278004
The connection tensor in the depth direction includes:
Figure M_211229101854828_828111001
hidden_size×W×H
Figure M_211229101854859_859304002
D×W×H
Figure M_211229101854906_906246003
hidden_size×W×H
Figure M_211229101854936_936996004
(2 *hideen_size+D)×W×H
then:
Figure M_211229101854968_968724001
representing the number of channels in the depth direction as
Figure M_211229101855016_016568001
Is/are as follows
Figure M_211229101855064_064184002
Performing convolution processing to generate channels with the number of channels in the depth direction
Figure M_211229101855079_079048003
Tensor of
Figure M_211229101855110_110319004
。
On the fourth layer of the convolution layer, pair
Figure M_211229101855141_141552001
Is convolved to obtain
Figure M_211229101855172_172802002
Figure M_211229101855205_205469001
Representing the number of channels in the depth direction as
Figure M_211229101855253_253040001
Is/are as follows
Figure M_211229101855284_284115002
Performing convolution processing to generate channels with the number of channels in the depth direction
Figure M_211229101855331_331267003
Tensor of
Figure M_211229101855362_362260004
。
Then according to
Figure M_211229101855394_394927001
According to
Figure M_211229101855411_411038002
Channel output tensor, output and original image
Figure M_211229101855457_457933003
Have the same
Figure M_211229101855489_489173004
Steganographic image of
Figure M_211229101855520_520415005
。
The implementation process of recovering the steganographic information is as follows:
the extraction network extracts the recovery steganographic information in the steganographic image, which is the reverse of the process from step 1 to step 5, wherein the recovery steganographic information
Figure M_211229101855551_551696001
Comprises the following steps:
on the first layer of the convolution layer
Figure M_211229101855582_582904001
Represents the number of channels in the depth direction
Figure M_211229101855631_631776002
Steganographic image of
Figure M_211229101855663_663763003
Performing convolution processing to generate channels with the number of channels in the depth direction
Figure M_211229101855694_694249004
Tensor of
Figure M_211229101855741_741128005
。
In the second layer of the convolution layer,
Figure M_211229101855756_756798001
represents the number of channels in the depth direction
Figure M_211229101855805_805105002
Tensor of
Figure M_211229101855821_821420003
Performing convolution processing to generate channels with the number of channels in the depth direction
Figure M_211229101855852_852457004
Tensor of
Figure M_211229101855883_883721005
。
In the third layer of the convolution layer,
Figure M_211229101855915_915265001
represents the number of channels in the depth direction
Figure M_211229101855961_961815002
Is/are as follows
Figure M_211229101855994_994548003
Performing convolution processing to generate channels with the number of channels in the depth direction
Figure M_211229101856026_026308004
Is/are as follows
Figure M_211229101856057_057544005
。
In the fourth layer of the convolution layer,
Figure M_211229101856073_073169001
represents the number of channels in the depth direction
Figure M_211229101856120_120082002
Is/are as follows
Figure M_211229101856151_151272003
Performing convolution processing to generate channels with the number of channels in the depth direction
Figure M_211229101856182_182536004
Is/are as follows
Figure M_211229101856199_199588005
。
With respect to the step S104, fig. 2 provides a possible implementation manner, and fig. 2 is a schematic flow chart of a method for training an information steganography model provided in an embodiment of the present application, as shown in fig. 2, including:
s201, calculating a first loss function value according to preset steganographic information and recovered steganographic information.
And calculating a first loss function value according to the preset steganography information and the recovery steganography information, wherein the first loss function value is a loss value of the steganography information and is used for measuring the difference between the preset steganography information and the recovery steganography information.
In some embodiments, binary cross entropy processing of the preset steganographic information and the recovery steganographic information may be calculated, where both the preset steganographic information and the recovery steganographic information may be binary information, the binary cross entropy is a binary information cross entropy for calculating the preset steganographic information and the recovery steganographic information, and then the calculated binary cross entropy is used as a first loss function value.
S202, calculating a second loss function value according to the first loss function value and the steganography evaluation value.
Wherein the target loss function value comprises: and a second loss function value, the steganographic evaluation value being used to characterize the quality of the steganographic image sample.
Each image sample and the corresponding steganographic image sample correspond to a first loss function value, and each image sample and the corresponding steganographic image sample correspond to a steganographic evaluation value.
In some embodiments, all the first loss function values and all the steganographic information evaluation values may be used to calculate a second loss function value, where the second loss function value may be a sum of all the first loss function values and all the steganographic evaluation values, or may be a sum of a mean value of all the first loss function values and a mean value of all the steganographic evaluation values, which is not limited in this embodiment.
Correspondingly, step S105, updating parameters of the information steganography model according to the target loss function value to obtain a target information steganography model, including:
and S203, updating the parameters of the steganographic network and the extraction network according to the second loss function value.
After the second loss function value is calculated, a parameter updating gradient corresponding to the second loss function can be calculated through a back propagation algorithm, and parameters of the steganographic network and the extraction network in the information steganographic model are updated according to the parameter updating gradient.
Wherein the first loss condition may include that the second loss function value is minimal.
That is to say, the second loss function value determined each time is recorded, after the parameters of the steganographic network and the extraction network are updated, the information steganographic model after the parameters are updated is trained according to a plurality of image samples, that is, the parameters of the steganographic network and the extraction network are updated iteratively, and the steps S101-S102 and S201-S202 are executed until the determined second loss function meets the minimum value of all the second loss functions, and the steganographic network and the extraction network corresponding to the minimum second loss function are determined as the steganographic network and the extraction network of the target information steganographic network.
It should be noted that the second loss function is an error function of the image sample and the steganographic image sample, and the smaller the second loss function is, the better the steganographic effect of the steganographic image sample is, and the more difficult the difference between the steganographic image sample and the image sample is to be found by human eyes.
During model training, an Adam optimizer can be used for optimizing the parameter gradient, the learning rate is set to be 0.001, the iterative parameter gradient value is prevented from being too large through a cutting mode, the threshold value can be set to be 0.25, the critical weight can be set to be between-0.1 and 0.1, and the critical weight refers to a loss function value.
In the information steganography model training method of this embodiment, a first loss function value is calculated according to preset steganography information and recovery steganography information, and a second loss function value is calculated according to the first loss function value and a steganography evaluation value. The quality of the steganographic image sample and the difference between the preset steganographic information and the recovered steganographic information are comprehensively considered to obtain a second loss function value, and when the second loss function value is adopted to update the model parameters, the accuracy of the target information steganographic model is improved.
To the step S202, fig. 3 provides a possible implementation manner, and fig. 3 is a schematic flow chart of a method for training an information steganography model provided in an embodiment of the present application, as shown in fig. 3, including:
s301, calculating the mean square error of the pixel values of the plurality of image samples and the corresponding stego-images in the plurality of color channels.
Wherein, the plurality of color channels may be R, G, B three color channels. For each image sample, pixel values of the image sample and the corresponding stego-image sample on a plurality of color channels may be obtained, and then a mean square error of the pixel values of the image sample and the corresponding stego-image sample on the plurality of color channels is calculated, the mean square error indicating a difference in pixels between the pattern sample and the corresponding stego-image sample, wherein the mean square error is
Figure M_211229101856231_231349001
Comprises the following steps:
Figure M_211229101856262_262635001
wherein the content of the first and second substances,
Figure M_211229101856309_309496001
representing the width of the image sample and the corresponding stego-image sample,
Figure M_211229101856356_356354002
representing the height of the image sample and the corresponding stego-image sample,
Figure M_211229101856516_516517003
a pixel vector representing a sample of the image,
Figure M_211229101856564_564320004
a pixel vector representing a steganographic image sample, wherein the pixel vector is composed of pixel values of three color channels,
Figure M_211229101856815_815813005
the open root of the sum of squares of the components after the pixel vector representing the image sample is differenced with the pixel vector of the steganographic image sample.
S302, calculating a second loss function value according to the mean square error, the first loss function value and the steganography evaluation value.
The second loss function value may be a sum of the mean square error, the first loss function value, and the steganographic evaluation value, or a weighted average of the mean square error, the first loss function value, and the steganographic evaluation value.
As an example, the second loss function value is:
Figure M_211229101857145_145914001
wherein the content of the first and second substances,
Figure M_211229101857437_437896001
represents the mean square error,
Figure M_211229101857547_547266002
The value of the first loss function is expressed,
Figure M_211229101857636_636186003
a value representing a steganographic evaluation value,
Figure M_211229101857714_714282004
to represent
Figure M_211229101857793_793834005
The coefficient of (a) is determined,
Figure M_211229101857903_903741006
this may be empirically chosen, such as 100, and this embodiment is not limited thereto.
It should be noted that each image sample and the corresponding stego-image sample corresponds to a mean square error, and therefore,
Figure M_211229101858031_031640001
the sum of all the mean square errors or the average of all the mean square errors may be used, which is not limited in this embodiment.
And then, updating the steganography network and extracting the parameters of the network according to the second loss function value, training the information steganography model after the parameters are updated according to a plurality of image samples until the second loss function value reaches a first loss condition, and finishing the training of the information steganography model when the sum value is minimum if the second loss function value is the sum value of the mean square error, the first loss function value and the steganography evaluation value.
In the information steganography model training method of this embodiment, the mean square error of the pixel values of the plurality of image samples and the corresponding steganography image samples in the plurality of color channels is calculated, and the second loss function value is calculated according to the mean square error, the first loss function value and the steganography evaluation value. And the quality of the steganographic image sample, the difference of the preset steganographic information and the recovered steganographic information and the pixel difference of the image sample and the steganographic image sample are comprehensively considered to obtain a second loss function value, and when the second loss function value is adopted to update the model parameters, the accuracy of the target information steganographic model is improved.
Before the step S105, the step shown in fig. 4 is further included, and fig. 4 is a schematic flow chart of a method for training an information steganography model provided in an embodiment of the present application, as shown in fig. 4, the method includes:
s401, calculating a third loss function value of the judgment network according to the plurality of image samples and the corresponding steganographic image samples.
Wherein the target loss function value further comprises: a third loss function value to indicate a difference in image pixel data distribution for the image sample and the corresponding stego image sample.
Calculating a sample loss function value of each image sample and the corresponding steganographic image sample, and then determining an added value of all the sample loss function values as a third loss function value, or determining an average value of all the sample loss function values as the third loss function value
Figure M_211229101858125_125408001
Comprises the following steps:
Figure M_211229101858187_187897001
wherein the content of the first and second substances,
Figure M_211229101858271_271909001
the image pixel data, which represents an image sample, may be an average of pixel values of pixel points on the image sample,
Figure M_211229101858334_334381002
the image pixel data representing the steganographic image sample may be pixel values of pixels on the steganographic image sampleAverage value.
Correspondingly, step S105, updating parameters of the information steganography model according to the target loss function value to obtain a target information steganography model, including:
and S402, updating and judging the parameters of the network according to the third loss function value.
After the third loss function value is calculated, a parameter update gradient corresponding to the third loss function may be calculated through a back propagation algorithm, and a decision network in the information steganography model is updated according to the parameter update gradient, and optionally, after the parameter of the decision network is updated according to the third loss function value, the method further includes:
and S403, training the information steganography model after the parameters are updated according to the plurality of image samples until the third loss function value reaches a second loss condition.
Wherein the second loss condition may include that the third loss function value is maximum.
That is, the third loss function value determined each time is recorded, after the parameters of the decision network are updated, the information steganography model after the parameters are updated is trained according to a plurality of image samples, that is, the parameters of the decision network are updated iteratively, and steps S101-S102, S401 are executed until the determined third loss function satisfies the maximum value among all the third loss functions, and the decision network corresponding to the maximum third loss function is determined as the decision network of the target information steganography model.
It should be noted that, when the parameters of the decision network are updated iteratively, the steganographic network and the extraction network of the target information steganographic network are trained, that is, after the training of the steganographic network and the extraction network is completed, the decision network is trained again.
In the information steganography model training method of this embodiment, a third loss function value of the decision network is calculated according to the plurality of image samples and the corresponding steganography image samples, a parameter of the decision network is updated according to the third loss function value, and the information steganography model after the parameter update is trained according to the plurality of image samples until the third loss function value reaches a second loss condition. And the third loss function value is obtained by considering the image pixel data distribution difference of the pattern sample and the steganography image sample, and the accuracy of the target information steganography model is improved when the model parameters are updated by adopting the third loss function value.
Fig. 5 is a schematic flow chart of the information steganography method provided in the embodiment of the present application, and an execution main body of the embodiment may be an information steganography device. As shown in fig. 5, the method includes:
s501, processing the file to be processed to obtain an image to be processed.
The file to be processed may be a format file of any format, such as a PDF file and an OFD file, and the file may be any type of file, such as a scan file, a document, an image, and the like.
In the actual use process, in the Windows system, a file operation interception hook can be arranged in a path of processing and circulating files to be processed so as to intercept the files to be processed, and in the linux desktop release, desktop files corresponding to operation software are modified so as to intercept the files to be processed. Therefore, the whole process of opening, reading and outputting the layout file can be covered.
After the to-be-processed file is captured, the whole page image can be directly extracted as the to-be-processed image for the scanned file, the image in the image can be used as the to-be-processed image for the image and text, and the to-be-processed image can be obtained by performing bitmap formation and blocking on the page according to a certain segmentation mode for the official document, which is specifically similar to the acquisition mode of the image sample in the embodiment of fig. 1.
S502, embedding preset steganographic information of the image to be processed into the image to be processed by adopting a steganographic network in a target information steganographic model, and obtaining a steganographic image of the image to be processed.
The preset steganographic information may be binary information, such as binary information obtained by binary encoding a user identity.
The target information steganography model is obtained by training in the embodiments of fig. 1 to 4, and the image to be processed and the preset steganography information are input into a steganography network in the target information steganography model to obtain a steganography image of the image to be processed.
And S503, generating a target file according to the steganographic image.
The hidden image and the file to be processed are packaged and combined to generate a target file, that is, the image to be processed in the file to be processed is replaced by the hidden image, so that the target file can be generated, if the page is subjected to bitmap processing and blocking according to a certain segmentation mode to obtain the image to be processed, the hidden image can be logically synthesized into the file to be processed according to the blocking to generate the target file, and if the file to be processed is a PDF file, the target file is a new PDF file.
After that, the current user can continue to specify file operations, for example, continue to open files, print files, and the like, it should be noted that, by embedding invisible steganographic information into the image to be processed, the image to be processed and the steganographic image have almost no difference in vision, and the steganographic effect is good.
In addition, if the target file is leaked, the target file can be imaged to obtain a target image, such as scanning, screenshot and shooting, and the steganographic information in the target image can be extracted by adopting an extraction network in a target information steganographic model, wherein the steganographic information can be binary information of a user identity, so that the steganographic information can still be extracted from the leaked target file under the condition that the images are mixed or all the images, and the tracing effect is achieved.
In the information steganography method of this embodiment, a file to be processed is processed to obtain an image to be processed, a steganography network in a target information steganography model is adopted to embed preset steganography information of the image to be processed into the image to be processed to obtain a steganography image of the image to be processed, and a target file is generated according to the steganography image. The compatibility is strong, and the steganography effect is good.
Fig. 6 is a schematic structural diagram of an information steganography model training apparatus provided in an embodiment of the present application, which may be integrated in an information steganography model training device. As shown in fig. 6, the apparatus includes:
the processing module 601 is configured to use the steganographic network to embed preset steganographic information of a plurality of image samples into the plurality of image samples, respectively, so as to obtain steganographic image samples corresponding to the plurality of image samples;
processing the steganographic image sample by adopting the extraction network to obtain recovery steganographic information corresponding to the steganographic image sample;
processing the plurality of image samples and the corresponding stego-images samples by adopting the judging network to obtain stego-evaluation values, wherein the stego-evaluation values are used for representing the quality of the stego-images samples, the judging network is used for extracting the image characteristics of the plurality of image samples and the image characteristics of the corresponding stego-images samples, and the stego-evaluation values are the characteristic differences of the image characteristics of the plurality of image samples and the image characteristics of the corresponding stego-images samples;
a calculating module 602, configured to calculate a target loss function value according to the preset steganography information, the recovery steganography information, and the steganography evaluation value;
and the updating module 603 is configured to perform parameter updating on the information steganography model according to the target loss function value, so as to obtain a target information steganography model.
Optionally, the calculating module 602 is specifically configured to:
calculating a first loss function value according to the preset steganography information and the recovery steganography information;
calculating a second loss function value according to the first loss function value and the steganography evaluation value; the target loss function value comprises: the second loss function value;
the update module 603 is specifically configured to:
and updating the parameters of the steganographic network and the extraction network according to the second loss function value.
Optionally, the calculating module 602 is specifically configured to:
calculating the mean square error of the pixel values of the plurality of image samples and the corresponding stego-images samples in a plurality of color channels;
and calculating the second loss function value according to the mean square error, the first loss function value and the steganography evaluation value.
Optionally, the method further comprises:
a training module 604, configured to train the information steganography model with updated parameters according to the plurality of image samples until the second loss function value reaches the first loss condition.
Optionally, the calculating module 602 is further configured to:
calculating a third loss function value of the decision network according to the plurality of image samples and the corresponding stego image samples; the target loss function value further comprises: the third loss function value;
the update module 603 is specifically configured to:
and updating the parameters of the decision network according to the third loss function value.
Optionally, the training module 604 is further configured to:
and training the information steganography model after the parameters are updated according to the plurality of image samples until the third loss function value reaches a second loss condition.
The description of the processing flow of each module in the device and the interaction flow between the modules may refer to the related description in the above embodiment of the information steganography model training method, and will not be described in detail here.
Fig. 7 is a schematic structural diagram of an information steganography apparatus provided in an embodiment of the present application, where the apparatus may be integrated in an information steganography device. As shown in fig. 7, the apparatus includes:
the processing module 701 is used for processing the file to be processed to obtain an image to be processed;
training by adopting the training method of any one of the first aspect to obtain a steganographic network in a target information steganographic model, and embedding preset steganographic information of the image to be processed into the image to be processed to obtain a steganographic image of the image to be processed;
a generating module 702, configured to generate a target file according to the steganographic image.
The description of the processing flow of each module in the device and the interaction flow between each module may refer to the related description in the above-mentioned information steganography method embodiment, and will not be described in detail here.
Fig. 8 is a schematic structural diagram of an information steganography model training device provided in an embodiment of the present application, and as shown in fig. 8, the device includes: a processor 801, a memory 802 and a bus 803, wherein the memory 802 stores machine readable instructions executable by the processor 801, when the information steganography model training device operates, the processor 801 communicates with the memory 802 through the bus 803, and the processor 801 executes the machine readable instructions to execute the information steganography model training method.
Fig. 9 is a schematic structural diagram of an information steganography device provided in an embodiment of the present application, and as shown in fig. 9, the device includes: a processor 901, a memory 902 and a bus 903, wherein the memory 902 stores machine readable instructions executable by the processor 901, when the information steganography device operates, the processor 901 communicates with the memory 902 through the bus 903, and the processor 901 executes the machine readable instructions to execute the information steganography method.
Embodiments of the present application further provide a computer-readable storage medium, on which a computer program is stored, where the computer program is executed by a processor to perform the above method embodiments.
It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the system and the apparatus described above may refer to corresponding processes in the method embodiments, and are not described in detail in this application. In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. The above-described apparatus embodiments are merely illustrative, and for example, the division of the modules is merely a logical division, and there may be other divisions in actual implementation, and for example, a plurality of modules or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of devices or modules through some communication interfaces, and may be in an electrical, mechanical or other form.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present application, and shall be covered by the scope of the present application.

Claims (10)

1. An information steganography model training method, wherein the information steganography model comprises: a steganographic network, an extraction network, and a decision network, the method comprising:
embedding preset steganography information of a plurality of image samples into the plurality of image samples respectively by adopting the steganography network to obtain steganography image samples corresponding to the plurality of image samples;
processing the steganographic image sample by adopting the extraction network to obtain recovery steganographic information corresponding to the steganographic image sample;
processing the plurality of image samples and the corresponding stego-images samples by adopting the judging network to obtain stego-evaluation values, wherein the stego-evaluation values are used for representing the quality of the stego-images samples, the judging network is used for extracting the image characteristics of the plurality of image samples and the image characteristics of the corresponding stego-images samples, and the stego-evaluation values are the characteristic differences of the image characteristics of the plurality of image samples and the image characteristics of the corresponding stego-images samples;
calculating a target loss function value according to the preset steganography information, the recovery steganography information and the steganography evaluation value;
and updating parameters of the information steganography model according to the target loss function value to obtain a target information steganography model.
2. A method as claimed in claim 1, wherein said calculating a target loss function value from said preset steganographic information, said recovery steganographic information and said steganographic evaluation value comprises:
calculating a first loss function value according to the preset steganography information and the recovery steganography information;
calculating a second loss function value according to the first loss function value and the steganography evaluation value; the target loss function value comprises: the second loss function value;
the updating the parameters of the information steganography model according to the target loss function value to obtain a target information steganography model comprises the following steps:
and updating the parameters of the steganographic network and the extraction network according to the second loss function value.
3. The method of claim 2, wherein said calculating a second loss function value based on said first loss function value and said steganographic evaluation value comprises:
calculating the mean square error of the pixel values of the plurality of image samples and the corresponding stego-images samples in a plurality of color channels;
and calculating the second loss function value according to the mean square error, the first loss function value and the steganography evaluation value.
4. The method of claim 2, wherein after updating the parameters of the steganographic network and the extraction network according to the second loss function value, the method further comprises:
and training the information steganography model after the parameters are updated according to the plurality of image samples until the second loss function value reaches a first loss condition.
5. The method of claim 2, wherein before updating the information steganography model according to the target loss function value to obtain a target information steganography model, the method further comprises:
calculating a third loss function value of the decision network according to the plurality of image samples and the corresponding stego image samples; the target loss function value further comprises: the third loss function value;
the updating the parameters of the information steganography model according to the target loss function value to obtain a target information steganography model comprises the following steps:
and updating the parameters of the decision network according to the third loss function value.
6. The method of claim 5, wherein after updating the parameters of the decision network based on the third loss function value, the method further comprises:
and training the information steganography model after the parameters are updated according to the plurality of image samples until the third loss function value reaches a second loss condition.
7. A method of steganography of information, comprising:
processing the file to be processed to obtain an image to be processed;
training by adopting the training method of any one of claims 1 to 6 to obtain a steganographic network in a target information steganographic model, and embedding preset steganographic information of the image to be processed into the image to be processed to obtain a steganographic image of the image to be processed;
and generating a target file according to the steganographic image.
8. An information steganography model training apparatus, comprising: a processor, a memory and a bus, the memory storing machine readable instructions executable by the processor, the processor and the memory communicating over the bus when the information steganography model training apparatus is operated, the processor executing the machine readable instructions to perform the method of any one of claims 1 to 6.
9. An information steganography device, comprising: a processor, a memory and a bus, the memory storing machine-readable instructions executable by the processor, the processor and the memory communicating over the bus when the information steganography device is operated, the processor executing the machine-readable instructions to perform the method of claim 7.
10. A computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium, which computer program, when being executed by a processor, performs the method of any one of claims 1 to 7.
CN202111637371.7A 2021-12-30 2021-12-30 Information steganography model training method, information steganography device and storage medium Active CN113989097B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111637371.7A CN113989097B (en) 2021-12-30 2021-12-30 Information steganography model training method, information steganography device and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111637371.7A CN113989097B (en) 2021-12-30 2021-12-30 Information steganography model training method, information steganography device and storage medium

Publications (2)

Publication Number Publication Date
CN113989097A true CN113989097A (en) 2022-01-28
CN113989097B CN113989097B (en) 2022-05-06

Family

ID=79734870

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111637371.7A Active CN113989097B (en) 2021-12-30 2021-12-30 Information steganography model training method, information steganography device and storage medium

Country Status (1)

Country Link
CN (1) CN113989097B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115883260A (en) * 2023-02-27 2023-03-31 安徽深迪科技有限公司 Digital Tibetan traceability system based on steganography technology

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100191565A1 (en) * 2009-01-23 2010-07-29 Hudson Marine Management Services System and Method for Complying With TMSA Guidelines
CN109492416A (en) * 2019-01-07 2019-03-19 南京信息工程大学 A kind of guard method of big data image and system based on safety zone
CN109587372A (en) * 2018-12-11 2019-04-05 北京邮电大学 A kind of invisible image latent writing art based on generation confrontation network

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100191565A1 (en) * 2009-01-23 2010-07-29 Hudson Marine Management Services System and Method for Complying With TMSA Guidelines
CN109587372A (en) * 2018-12-11 2019-04-05 北京邮电大学 A kind of invisible image latent writing art based on generation confrontation network
CN109492416A (en) * 2019-01-07 2019-03-19 南京信息工程大学 A kind of guard method of big data image and system based on safety zone

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115883260A (en) * 2023-02-27 2023-03-31 安徽深迪科技有限公司 Digital Tibetan traceability system based on steganography technology

Also Published As

Publication number Publication date
CN113989097B (en) 2022-05-06

Similar Documents

Publication Publication Date Title
Sedighi et al. Content-adaptive steganography by minimizing statistical detectability
CN110516201B (en) Image processing method, image processing device, electronic equipment and storage medium
KR20190109732A (en) Method and apparatus for synthesizing kanji printed characters images
JP6139396B2 (en) Method and program for compressing binary image representing document
CN101924851B (en) Image processing apparatus and image processing method
US20150228045A1 (en) Methods for embedding and extracting a watermark in a text document and devices thereof
CN110400362B (en) ABAQUS two-dimensional crack modeling method and system based on image and computer readable storage medium
EP2410727A1 (en) Marker generation program, restoration program, marker generation device, restoration device, and marker generation method
CN101122995A (en) Binary image digit water mark embedding, extraction method and device
JP4595014B2 (en) Digital watermark embedding device and detection device
CN113989097B (en) Information steganography model training method, information steganography device and storage medium
CN112088378A (en) Image hidden information detector
JP4364809B2 (en) Image processing apparatus, image processing method, program, and recording medium
JP2007158725A (en) Image processor, image processing method, and program
JP5893379B2 (en) Image compression apparatus, image compression method, and computer program
US20090110313A1 (en) Device for performing image processing based on image attribute
JP5038231B2 (en) Image processing apparatus, image processing method, program for executing image processing method, and recording medium
JP2009004990A (en) Image forming apparatus and image forming method
US8810877B2 (en) Image processing apparatus, image processing method, and storage medium for eliminating blurring of scanned image
JP4414907B2 (en) Image processing apparatus, image processing method, and image processing program
JP7379876B2 (en) Character recognition device, document file generation method, document file generation program
CN113065407B (en) Financial bill seal erasing method based on attention mechanism and generation countermeasure network
US9053359B2 (en) Method and system for document authentication using Krawtchouk decomposition of image patches for image comparison
CN111127288B (en) Reversible image watermarking method, reversible image watermarking device and computer readable storage medium
CN114241493A (en) Training method and training device for training data of amplification document analysis model

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