CN113628090A - Anti-interference message steganography and extraction method and system, computer equipment and terminal - Google Patents
Anti-interference message steganography and extraction method and system, computer equipment and terminal Download PDFInfo
- Publication number
- CN113628090A CN113628090A CN202110793854.XA CN202110793854A CN113628090A CN 113628090 A CN113628090 A CN 113628090A CN 202110793854 A CN202110793854 A CN 202110793854A CN 113628090 A CN113628090 A CN 113628090A
- Authority
- CN
- China
- Prior art keywords
- image
- message
- interference
- steganography
- steganographic
- 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
- 238000000605 extraction Methods 0.000 title claims abstract description 53
- 238000012549 training Methods 0.000 claims abstract description 54
- 238000000034 method Methods 0.000 claims abstract description 42
- 230000000007 visual effect Effects 0.000 claims abstract description 36
- 230000006870 function Effects 0.000 claims abstract description 33
- 230000008569 process Effects 0.000 claims abstract description 20
- 238000004088 simulation Methods 0.000 claims abstract description 11
- 238000013528 artificial neural network Methods 0.000 claims abstract description 8
- 239000011159 matrix material Substances 0.000 claims description 19
- 238000007906 compression Methods 0.000 claims description 9
- 230000006835 compression Effects 0.000 claims description 9
- 238000007639 printing Methods 0.000 claims description 9
- 238000012986 modification Methods 0.000 claims description 7
- 230000004048 modification Effects 0.000 claims description 7
- 230000008447 perception Effects 0.000 claims description 7
- 230000000694 effects Effects 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000013256 Gubra-Amylin NASH model Methods 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 3
- 230000016776 visual perception Effects 0.000 claims description 3
- 238000005336 cracking Methods 0.000 claims 1
- 238000013135 deep learning Methods 0.000 abstract description 14
- 238000009826 distribution Methods 0.000 abstract description 2
- 238000004422 calculation algorithm Methods 0.000 description 17
- 238000010586 diagram Methods 0.000 description 12
- 238000005516 engineering process Methods 0.000 description 6
- 230000009466 transformation Effects 0.000 description 5
- 238000005070 sampling Methods 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 230000006378 damage Effects 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000004438 eyesight Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 235000002566 Capsicum Nutrition 0.000 description 1
- 241000282376 Panthera tigris Species 0.000 description 1
- 239000006002 Pepper Substances 0.000 description 1
- 235000016761 Piper aduncum Nutrition 0.000 description 1
- 235000017804 Piper guineense Nutrition 0.000 description 1
- 244000203593 Piper nigrum Species 0.000 description 1
- 235000008184 Piper nigrum Nutrition 0.000 description 1
- VYPSYNLAJGMNEJ-UHFFFAOYSA-N Silicium dioxide Chemical compound O=[Si]=O VYPSYNLAJGMNEJ-UHFFFAOYSA-N 0.000 description 1
- 230000003213 activating effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000013144 data compression Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000001066 destructive effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 210000001747 pupil Anatomy 0.000 description 1
- 150000003839 salts Chemical class 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000009528 severe injury Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/0021—Image watermarking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
- G06F21/62—Protecting access to data via a platform, e.g. using keys or access control rules
- G06F21/6218—Protecting access to data via a platform, e.g. using keys or access control rules to a system of files or objects, e.g. local or distributed file system or database
- G06F21/6245—Protecting personal data, e.g. for financial or medical purposes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioethics (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Computer Hardware Design (AREA)
- Computer Security & Cryptography (AREA)
- Image Processing (AREA)
Abstract
The invention belongs to the technical field of message steganography, deep learning and neural networks, and discloses an anti-interference message steganography and extraction method, a system, computer equipment and a terminal, wherein the anti-interference message steganography and extraction method comprises the following steps: in the process of training the steganography function, a loss function is used, wherein the loss function comprises items related to the visual scores of the carrier images; in the process of training the extraction function, the steganographic image passes through an interference simulation module so as to train the anti-interference capability of the image; and the user prints the obtained steganographic image on paper or displays the obtained steganographic image on a screen, and uses a network model to extract the message after shooting and collecting by using the camera equipment. The anti-interference message steganography and extraction method provided by the invention is based on a GAN framework in deep learning, and not only can effectively extract the message steganographically written in the steganography image, thereby achieving the purposes of message hiding and distribution, ensuring that the visual quality of the image is not greatly damaged, but also resisting the interference which commonly affects the image quality in reality.
Description
Technical Field
The invention belongs to the technical field of message steganography, deep learning and neural networks, and particularly relates to an anti-interference message steganography and extraction method, a system, computer equipment and a terminal.
Background
Steganography, which is a technique for hiding secret messages after normal messages, is, like this, a message encryption technique, and offers a viable alternative to encryption in special circumstances, since in such cases the use of encryption techniques may be of unnecessary concern. Classical steganography is a secret communication means used by ancient people and mainly comprises invisible ink, carten grids, Tibetan poetry and the like. Modern steganography refers to the hiding of messages into digital media using electronic communication and digital techniques. Each modern steganographic system contains two basic components: an embedding algorithm and an extraction algorithm. The embedding algorithm accepts three inputs: a secret message, a secret key and a carrier object, which are used to convey a message. The output of the embedding algorithm is called a steganographic object. Steganographic objects are also used as input to an extraction algorithm to generate a secret message. The most common form in steganography is image steganography, and meanwhile, character steganography or video steganography also exists, and the two-dimensional code common in real life can be regarded as image steganography to some extent.
Traditional image steganography algorithms can be divided into two categories: a spatial domain steganography algorithm and a frequency domain steganography algorithm. The spatial domain steganography algorithm embeds secret information by modifying image pixels, such as LSB replacement and matching algorithm; the frequency domain steganography algorithm carries out the embedding of secret information by modifying certain specified frequency domain coefficients in the host signal, such as a Discrete Cosine Transform (DCT) algorithm, a Discrete Fourier Transform (DFT) algorithm, a Discrete Wavelet Transform (DWT) algorithm and the like. However, these traditional steganography algorithms lack robustness, and when the secret information is transmitted in a damaged channel such as a social network and wireless communication, the secret information cannot be correctly extracted due to slight interference.
Based on the defects, a robust lossless information hiding algorithm is proposed, robustness is achieved through cyclic interpretation of double mapping conversion by using a patchwork theory and modulo 256 addition operation, but salt and pepper noise is easily generated by the method; an algorithm based on integer wavelet transform has also been proposed to embed information by changing the average of HL1 and LH1 coefficients. However, this method requires the embedding side and the extracting side to share side information such as a threshold value, and has a low capacity.
With the development of deep learning, many scholars use deep learning to realize robust steganography, and a hiddenn model is proposed, which is divided into four parts: encoder E, decoder D, noise layer N and discriminator a. The encoder E inputs the image and the secret information and outputs the image and the secret information as an image; the decoder inputs the image and outputs the image as secret information; the discriminator a is responsible for discriminating the difference between the encoder generated image and the input image. The HiDDeN model has good performance in the aspect of secret information embedding amount, the proposed end-to-end steganography framework allows new noise to be added to a noise layer on the basis of the original framework, so that robustness to the new noise is realized, and the framework has certain expandability but is insufficient in steganography image quality. Still another person has proposed a stegasamp model on the basis of the hiddenn model, and image processing operations such as perspective transformation, color transformation, blurring and the like are added to a noise layer, so that the change existing in the process of obtaining a new image by printing and photographing a steganographic image is simulated, the problem that the secret information of the hiddenn cannot be decrypted after physical transmission is solved, and the method is an improvement on the application aspect of the hiddenn. However, images generated by the Stegasamp model have obvious traces which influence the visual perception of the images, and the image is more and more damaged by the traces as the embedding amount of the messages increases.
The difficulty in solving the above problems and drawbacks is that the trace generated by hiding the message causes significant and human-perceptible damage to the image, and in order to ensure that the image can still be analyzed after physical transmission, such as printing and shooting, the message cannot be hidden with subtle changes, i.e., the generated trace must be present. The footprint also severely impacts the steganographic capacity ceiling of the technology, as the larger the capacity, the heavier the footprint.
The significance of solving the above problems and defects is that severe damage of image quality cannot be tolerated in some usage scenarios, such as digital image watermarking, thus severely limiting the application of stegasamp in such occasions, solving the problem of severe footprint can expand the usage scenarios of the technology, and at the same time can also improve the available steganographic capacity of the technology.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides an anti-interference message steganography and extraction method, a system, computer equipment and a terminal, in particular to a robust anti-interference message steganography and extraction method, a system, computer equipment and a terminal based on a deep learning technology, aiming at solving the problem of robust message steganography under various complex conditions in reality.
The invention is realized in such a way that an anti-interference message steganography and extraction method comprises the following steps:
in the process of training the steganography function, a loss function is used, wherein the loss function comprises items related to the visual scores of the carrier images; in the process of training the extraction function, the steganographic image passes through an interference simulation module so as to train the anti-interference capability of the image; and the user prints the obtained steganographic image on paper or displays the obtained steganographic image on a screen, and uses a network model to extract the message after shooting and collecting by using the camera equipment.
Further, the anti-interference message steganography and extraction method comprises the following steps:
calculating the visual score of each image pixel point according to a training image set to obtain a score matrix of the image pixels;
adding a program capable of simulating similar effects according to the interference to be simulated;
step three, training a proper network model according to the scoring matrix obtained by calculation in the step one by combining a training image set;
step four, selecting a specific carrier image and a specific steganographic message according to the network model obtained in the step three, and generating a required steganographic image;
fifthly, printing or displaying the steganographic image acquired in the fourth step on a carrier, and acquiring a shot image by using camera equipment;
and step six, extracting the message according to the network model by using desktop or mobile terminal software according to the shot image acquired in the step five.
Further, in the first step, a scoring model, namely a JND model, considering the strength of the human visual system on the perception capability of different image features is used for calculating the visual score of each pixel point of the image, the obtained score represents the influence of the pixel on the human visual perception, the score restrains the position of trace generation in training, the trace is allowed to exist in the high-scoring place, and the opposite is realized in the low-scoring place.
Further, in the second step, the interferences to be simulated include various common interferences such as color interference, reflection interference, blur interference, deformation, noise, JPEG image compression and rotation scaling, and the listed interferences are commonly encountered by real images in order to improve the capability of the decoder to extract messages from the interfered images.
Further, in the third step, the network is designed based on a GAN model, the encoder refers to a U-Net network, and two items of LPIPS scores and JND scores considering visual scores are added in the loss function; the LPIPS is an evaluation index for the images, is used for measuring the perception distance between the images, and is represented by the difference between characteristic values of the two images calculated and obtained through a deep neural network, and the lower the score is, the smaller the difference between the characteristic values of the two images is, namely the two images are closer; the JND means just a perceptual difference, representing a minimum threshold for a certain image modification that can be perceived by the human eye's visual system, and a modification smaller than this value is hard to attract the attention of the human eye.
Further, in step four, the steganographic image is very close to the carrier image and is only slightly noisy, because the loss function used in training is enhanced, the quality score of the generated steganographic image is higher before the contrast of the steganographic image is improved, and the subjective feeling of human eyes is also more excellent.
Further, in the sixth step, the image obtained by shooting is bound with the model generating the image, only the model generating the steganographic image can extract the message therein, other models cannot acquire correct messages, and the hidden messages cannot be cracked and acquired.
Another object of the present invention is to provide an anti-interference message steganography and extraction system using the anti-interference message steganography and extraction method, the anti-interference message steganography and extraction system including:
the scoring matrix acquisition module is used for calculating the visual scoring of each image pixel point according to the training image set to obtain a scoring matrix of the image pixels;
the program adding module is used for adding a program capable of simulating similar effects according to the interference needing to be simulated;
the network model building module is used for training a proper network model according to the calculated grading matrix and by combining the training image set;
the steganographic image generation module is used for selecting a specific carrier image and a specific steganographic message according to the acquired network model to generate a required steganographic image;
the image acquisition shooting module is used for printing or displaying the obtained steganographic image on a carrier and acquiring a shot image by using camera equipment;
and the message extraction module is used for extracting the message according to the acquired shot image by using desktop or mobile terminal software and a network model.
It is a further object of the invention to provide a computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of:
in the process of training the steganography function, a loss function is used, wherein the loss function comprises items related to the visual scores of the carrier images; in the process of training the extraction function, the steganographic image passes through an interference simulation module so as to train the anti-interference capability of the image; and the user prints the obtained steganographic image on paper or displays the obtained steganographic image on a screen, and uses a network model to extract the message after shooting and collecting by using the camera equipment.
Another object of the present invention is to provide an information data processing terminal, where the information data processing terminal is configured to implement the anti-interference message steganography and extraction system.
By combining all the technical schemes, the invention has the advantages and positive effects that: the anti-interference message steganography and extraction method provided by the invention is based on a GAN framework in deep learning, the human eye visual features of a carrier image and the common image interference in reality are considered, the image visual scoring is used for guiding the trace position generated by the network, the interference simulation link is added to improve the anti-interference capability of a network model, and the problem of steganography and extraction of messages in various environments with interference on image quality is solved by a reliable and robust method.
The invention provides a robust image steganography based on image visual grading and image interference simulation, which can not only effectively extract the information steganographically written in the steganography image so as to achieve the purposes of hiding and distributing the information, but also ensure that the visual quality of the image is not greatly damaged and simultaneously resist the interference which commonly affects the image quality in reality. The invention can steganographically write the text message into one image to obtain the steganographic image, so that the text message is invisible to human eyes, but can be extracted from the steganographic image by using the method of the invention, thereby realizing steganography and extraction of the message.
According to the invention, a training network obtains the capability of converting the message into noise similar to a carrier image and extracting the original message from the noise according to a training set image and a randomly generated message; in the whole training process, under the constraint of a loss function, the network can add noise to an area insensitive to human eyes in a carrier image to obtain a steganographic image; before the network is trained in the message extraction capability, the steganographic image needs to be subjected to a series of interference simulation processes including illumination, color transformation, blurring, deformation, JPEG (joint photographic experts group) compression, rotating scaling and the like to obtain a simulated distorted image; the extraction process of the network is trained by using the distorted image, so that the robustness of the system for extracting the message can be greatly improved. The anti-interference message steganography and extraction system based on the deep learning technology comprises: the system comprises a desktop end system which can be used for steganography, extraction and training and a mobile end system which can be used for extraction. The invention uses the deep learning network, adopts the loss function considering the image visual grading, ensures the whole visual quality of the final steganographic image, and simultaneously adds a module capable of simulating various different interferences in a real scene, improves the adaptability and the robustness of the whole system, thereby effectively and widely applying to the field of image-based message steganography.
The deep learning network provided by the invention is based on a GAN architecture, and needs generators and discriminators to perform staggered training as other GAN networks. The parameters of the discriminator are fixed when the optimal generation network needs to be trained, and similarly, the parameters of the discriminator are fixed when the optimal discrimination network needs to be trained. In the alternate execution of the two parts of training, the whole neural network can gradually obtain the capability of generating the hidden image with fine traces and difficult vision and extracting the message from the hidden image subjected to interference. After the training of the whole network is converged, the generator and the discriminator reach a Nash balance state, the discriminator cannot distinguish the image generated by the generator from the original image, the image generated by the encoder successfully learns a method of adding invisible traces to the carrier image to hide the message under the guidance of the whole loss function, and the decoder learns a method of extracting the complete message from the distorted image under the interference of the image distortion model.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments of the present invention will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a flowchart of an anti-interference message steganography and extraction method according to an embodiment of the present invention.
Fig. 2 is a schematic diagram of an anti-interference message steganography and extraction method according to an embodiment of the present invention.
Fig. 3 is a block diagram of an anti-interference message steganography and extraction system provided in an embodiment of the present invention;
in the figure: 1. a scoring matrix acquisition module; 2. a program adding module; 3. a network model construction module; 4. a steganographic image generation module; 5. an image acquisition shooting module; 6. and a message extraction module.
Fig. 4 is a schematic diagram illustrating a sample of the corresponding visual scores calculated by the training images according to the embodiment of the present invention.
Fig. 5 is a schematic diagram of a simulation process of image disturbance according to an embodiment of the present invention.
Fig. 6 is a schematic diagram of the overall process of the method of the present invention provided by the embodiment of the present invention.
Fig. 7 is a schematic diagram of a comparison between an original carrier image and a network-generated steganographic image provided by an embodiment of the present invention.
Fig. 7(a) is a carrier image provided by an embodiment of the present invention.
Fig. 7(b) is a steganographic image provided by an embodiment of the present invention.
Fig. 8 is a schematic diagram for comparing the details of original image enlargement, stegasamp detail enlargement and the details of the invention enlargement provided by the embodiment of the invention.
Fig. 8(a) is a schematic diagram of an enlarged image of details of an original image according to an embodiment of the present invention.
Fig. 8(b) is a schematic diagram of a stegasamp detail magnified image according to an embodiment of the present invention.
Fig. 8(c) is a schematic diagram of a detail enlarged image according to the present invention provided by the embodiment of the present invention.
Fig. 9 is a diagram of intermediate results obtained during the training process provided by the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Aiming at the problems in the prior art, the invention provides an anti-interference message steganography and extraction method, a system, computer equipment and a terminal, and the invention is described in detail below with reference to the attached drawings.
As shown in fig. 1, the method for steganography and extraction of an anti-interference message provided in the embodiment of the present invention includes the following steps:
s101, calculating the visual score of each image pixel point according to a training image set to obtain a score matrix of the image pixels;
s102, adding a program capable of simulating similar effects according to the interference needing to be simulated;
s103, training a proper network model according to the scoring matrix obtained by calculation in the S101 and by combining a training image set;
s104, selecting a specific carrier image and a specific steganographic message according to the network model obtained in the S103 to generate a required steganographic image;
s105, printing or displaying the steganographic image acquired in the S104 on a carrier, and acquiring a shot image by using camera equipment;
and S106, extracting the message according to the network model by using desktop or mobile terminal software according to the shot image acquired in the S105.
A schematic diagram of an anti-interference message steganography and extraction method provided by the embodiment of the present invention is shown in fig. 2.
As shown in fig. 3, the anti-interference message steganography and extraction system provided in the embodiment of the present invention includes:
the scoring matrix acquisition module 1 is used for calculating the visual scoring of each image pixel point according to the training image set to obtain a scoring matrix of the image pixels;
the program adding module 2 is used for adding programs capable of simulating similar effects according to the interference needing to be simulated;
the network model building module 3 is used for training a proper network model according to the calculated grading matrix and by combining the training image set;
the steganographic image generation module 4 is used for selecting a specific carrier image and a specific steganographic message according to the acquired network model to generate a required steganographic image;
the image acquisition shooting module 5 is used for printing or displaying the obtained steganographic image on a carrier and acquiring a shot image by using camera equipment;
and the message extraction module 6 is used for extracting the message according to the acquired shot image and the network model by using desktop or mobile terminal software.
The technical solution of the present invention will be further described with reference to the following explanation of terms.
Generating a confrontation Network (GAN) is a type of neural Network originally proposed by Ian Goodfellow, samples are taken from a complex probability distribution by training a Discriminator (Discriminator) and a Generator (Generator) in turns to confront each other, and has been widely applied to various application scenarios.
U-Net network model: a network structure is used that includes down-sampling and up-sampling. The down-sampling is used to gradually present the environment information, and the up-sampling is a process of restoring detail information by combining the down-sampled layer information and the up-sampled input information, and gradually restoring the image precision.
PSNR is an engineering term that represents the ratio of the maximum possible power of a signal to the power of destructive noise that affects the fidelity of the signal. PSNR is usually evaluated logarithmically, in decibels, due to the presence of a very wide dynamic range signal. PSNR is most commonly used to evaluate the output reconstruction quality of lossy compression codecs, e.g. for image compression.
SSIM is a perceptual index used to quantify the degradation of image quality caused by processing such as data compression or loss in data transmission. This is a complete reference index, i.e. two similar images (a reference image and a processed image) from the same image are needed, wherein the processed image is usually a compressed image.
The technical solution of the present invention is further described below with reference to specific examples.
According to the robust anti-interference message steganography and extraction method based on the deep learning technology, in the process of training the steganography function, a loss function uses items related to carrier image visual scores; in the process of training the extraction function, the steganographic image passes through an interference simulation module so as to train the anti-interference capability of the image; the user can print the obtained steganographic image on paper or display the steganographic image on a screen, and the information is extracted by using a network model after the steganographic image is shot and collected by using a camera device.
As shown in fig. 2, the robust anti-interference message steganography and extraction method based on the deep learning technique according to the embodiment of the present invention includes the following steps:
calculating the visual score of each image pixel point according to a training image set to obtain a score matrix of the image pixels;
secondly, adding interference simulation components including color transformation, noise, blur, deformation, JPEG compression, rotation scaling and the like according to the interference to be simulated;
step three, training according to the scoring matrix obtained by calculation in the step one and combining a training image set to obtain specific parameters of the network model;
step four, selecting a specific carrier image and a specific steganographic message according to the network model obtained in the step three, and generating a required steganographic image;
and step five, displaying the steganographic image acquired in the step four on a proper carrier, and acquiring an interfered image result by using camera equipment.
And step six, extracting the message by using the previously trained model parameters by using desktop or mobile terminal software according to the shot image acquired in the step five.
In the first step, a scoring model, which is also called a JND model, considering the strength of the human eye visual system in the perception capability of different image features is used for calculating the visual score of each pixel point of the image, and the obtained score represents the influence of the pixel on the human eye perception.
As shown in fig. 4, in a training set image according to the embodiment of the present invention, the visual score of each pixel of the training set image is calculated, and a score matrix of the image is obtained, where the score will guide the generation of traces in training.
In the second step, various common interferences such as color interference, reflection interference, fuzzy interference, deformation, noise, image compression rotation and the like can be added to enhance the robustness of the system.
As shown in fig. 5, a training image according to the embodiment of the present invention undergoes various common interferences such as deformation, blurring, noise, color transformation, JPEG compression, etc. before it is transmitted to a decoder to extract a message, which is to enhance the robustness of the steganographic image.
In step three, the network is designed based on a GAN model, the encoder refers to a U-Net network, and two items of LPIPS scores and JND scores considering the visual scores are added in the loss function. LPIPS is an evaluation index for images, and is used for measuring the perception distance between images. The difference between the characteristic values of the two images calculated and obtained through the deep neural network is used for representing, and the lower the score of the characteristic value is, the smaller the difference between the characteristic values of the two images is, namely, the closer the two images are represented. JND means just a perceptual difference, representing the minimum threshold for a modification of an image that the human visual system can perceive, and modifications smaller than this value are hardly noticeable to the human eye. The two items are scored to obtain a focus image totality and a focus image detail, so that the visual quality of the final steganographic image is improved from different angles.
As shown in fig. 6, a carrier image according to the embodiment of the present invention is combined with a binary code obtained by converting an input message according to Unicode code, and then transmitted to an encoder, so as to obtain a steganographic image through network model operation. The steganographic image may then be printed on a carrier or directly displayed on a display, captured images captured using a camera device such as a cell phone, and finally messages extracted using the same network model. The message extracted by the model operation is also represented in a binary form and can be converted into the original text message according to the Unicode.
In step four, the steganographic image is very close to the carrier image, and is noisy only at a slight point.
As shown in fig. 7, the difference between the carrier image and the steganographic image is small, but the steganographic image has hidden available messages, and the messages hidden therein can be extracted using a specific model that generates the steganographic image. Meanwhile, the steganographic image can be printed on paper or displayed in a display, and the steganographic image acquired by using common photographic equipment does not lose the capability of extracting messages, so that the robustness of anti-interference steganography and the possible use range of the steganographic image are greatly improved.
As shown in fig. 8, the tiger eye clearly shows that the trace generated by the invention is less noticeable than the trace generated by stegasamp, and does not affect the details perceived by human eyes. The trace generated around the tiger pupil by the method is obviously lighter, and the Stegasamp trace is very heavy, so that the perception definition of human eyes is damaged.
As shown in fig. 9, it can be seen that the steganographic traces are very obvious at the beginning of training, causing serious damage to the image quality, but as the number of training sessions increases, the macroscopic traces slowly disappear, and the quality of the entire image returns to an acceptable level.
In the sixth step, the shot image is bound with the model generating the image, only the model generating the steganographic image can extract the message therein, and other models cannot acquire the correct message, so that the security of the steganography of the image is ensured, and the message hidden therein cannot be cracked and acquired.
The technical solution of the present invention is further described below with reference to the working principle.
The structure of the deep learning network provided by the invention is a production line, and can be divided into four modules: an encoder, a discriminator, a distortion model and a decoder. The encoder reconstructs an image with the original image and the hidden binary string by means of a discriminator based on the GAN concept, then uses a distortion model for attacking the output of the encoder. The decoder receives these attacked images and is trained to recover the binary strings hidden therein. The design ideas of the distortion model and decoder come from the study of the related adversarial attacks.
The goal of the encoder is to reconstruct the steganographic image using the original carrier image and the hidden binary string, thereby ensuring a high quality of the reconstructed image. The encoder uses a U-Net structure, inputs the RGB image and the binary bit string, and outputs the RGB residual image with the binary bit string hidden therein, wherein the hidden image is the combination of the residual image and the carrier image.
The decoder receives the steganographic image after a distortion attack and is trained to extract hidden binary strings. The decoder consists of a series of convolutional layers followed by a fully-connected layer of the same length as the incoming message. This is followed by a Sigmoid function for activating the fully-connected layer and outputting the decoded binary string.
The discriminator uses the WGAN to evaluate the output of the network supervised encoder.
The distortion model uses various approximation methods to simulate the damage to the image quality caused by the printing, displaying, shooting and other processes commonly seen in reality, including but not limited to image deformation, image blurring, image color change, noise, image compression and rotation scaling.
The deep learning network provided by the invention is based on a GAN architecture, and needs generators and discriminators to perform staggered training as other GAN networks. The parameters of the discriminator are fixed when the optimal generation network needs to be trained, and similarly, the parameters of the discriminator are fixed when the optimal discrimination network needs to be trained. In the alternate execution of the two parts of training, the whole neural network can gradually obtain the capability of generating the hidden image with fine traces and difficult vision and extracting the message from the hidden image subjected to interference. After the training of the whole network is converged, the generator and the discriminator reach a Nash balance state, the discriminator cannot distinguish the image generated by the generator from the original image, the image generated by the encoder successfully learns the method of adding invisible traces to the carrier image to hide the message under the guidance of the whole loss function, and the decoder learns the method of extracting the complete message from the distorted image under the interference of the image distortion model.
As shown in table 1, the image obtained by the present invention is superior to stegasamp in both PSNR and SSIM visual scores, and has a higher extraction success rate.
TABLE 1
| PSNR | SSIM | Success rate of extraction | |
| StegaStamp | 0.9241 | 27.31 | 90％ |
| The invention | 0.9317 | 28.13 | 95％ |
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When used in whole or in part, can be implemented in a computer program product that includes one or more computer instructions. When loaded or executed on a computer, cause the flow or functions according to embodiments of the invention to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, the computer instructions may be transmitted from one website site, computer, server, or data center to another website site, computer, server, or data center via wire (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL), or wireless (e.g., infrared, wireless, microwave, etc.)). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that includes one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.
The above description is only for the purpose of illustrating the present invention and the appended claims are not to be construed as limiting the scope of the invention, which is intended to cover all modifications, equivalents and improvements that are within the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. An anti-interference message steganography and extraction method is characterized by comprising the following steps: in the process of training the steganography function, a loss function is used, wherein the loss function comprises items related to the visual scores of the carrier images; in the process of training the extraction function, the steganographic image passes through an interference simulation module so as to train the anti-interference capability of the image; and the user prints the obtained steganographic image on paper or displays the obtained steganographic image on a screen, and uses a network model to extract the message after shooting and collecting by using the camera equipment.
2. The tamper-resistant message steganography and extraction method of claim 1, wherein said tamper-resistant message steganography and extraction method comprises the steps of:
calculating the visual score of each image pixel point according to a training image set to obtain a score matrix of the image pixels;
adding a program capable of simulating similar effects according to the interference to be simulated;
step three, training a proper network model according to the scoring matrix obtained by calculation in the step one by combining a training image set;
step four, selecting a specific carrier image and a specific steganographic message according to the network model obtained in the step three, and generating a required steganographic image;
fifthly, printing or displaying the steganographic image acquired in the fourth step on a carrier, and acquiring a shot image by using camera equipment;
and step six, extracting the message according to the network model by using desktop or mobile terminal software according to the shot image acquired in the step five.
3. The method according to claim 2, wherein in the first step, a JND model, which is a scoring model considering the strength of human visual system's perceptibility to different image features, is used to calculate the visual score of each pixel point of the image, and the obtained score represents the influence of the pixel on human visual perception.
4. The method according to claim 2, wherein in step two, the interference to be simulated includes various common interferences including color interference, reflection interference, blur interference, deformation, noise, JPEG image compression and rotation scaling.
5. The method according to claim 2, wherein in step three, the network is designed based on GAN model, the encoder refers to U-Net network, and adds two LPIPS scores and JND scores considering visual scores to the loss function; the LPIPS is an evaluation index for the images, is used for measuring the perception distance between the images, and is represented by the difference between characteristic values of the two images calculated and obtained through a deep neural network, and the lower the score is, the smaller the difference between the characteristic values of the two images is, namely the two images are closer; the JND means just a perceptual difference, representing a minimum threshold for a certain image modification that can be perceived by the human eye's visual system, and a modification smaller than this value is hard to attract the attention of the human eye.
6. The method of claim 2, wherein in step four, the steganographic image is in close proximity to the carrier image and is noisy only in subtle ways.
7. The anti-interference message steganography and extraction method according to claim 2, wherein in the sixth step, the captured image is bound to the model generating the image, only the model generating the steganography image can extract the message therein, and other models cannot acquire the correct message, and the message hidden therein cannot be obtained by cracking.
8. An anti-interference message steganography and extraction system implementing the anti-interference message steganography and extraction method of any one of claims 1-7, the anti-interference message steganography and extraction system comprising:
the scoring matrix acquisition module is used for calculating the visual scoring of each image pixel point according to the training image set to obtain a scoring matrix of the image pixels;
the program adding module is used for adding a program capable of simulating similar effects according to the interference needing to be simulated;
the network model building module is used for training a proper network model according to the calculated grading matrix and by combining the training image set;
the steganographic image generation module is used for selecting a specific carrier image and a specific steganographic message according to the acquired network model to generate a required steganographic image;
the image acquisition shooting module is used for printing or displaying the obtained steganographic image on a carrier and acquiring a shot image by using camera equipment;
and the message extraction module is used for extracting the message according to the acquired shot image by using desktop or mobile terminal software and a network model.
9. A computer device, characterized in that the computer device comprises a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to carry out the steps of:
in the process of training the steganography function, a loss function is used, wherein the loss function comprises items related to the visual scores of the carrier images; in the process of training the extraction function, the steganographic image passes through an interference simulation module so as to train the anti-interference capability of the image; and the user prints the obtained steganographic image on paper or displays the obtained steganographic image on a screen, and uses a network model to extract the message after shooting and collecting by using the camera equipment.
10. An information data processing terminal, characterized in that the information data processing terminal is configured to implement the anti-interference message steganography and extraction system of claim 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110793854.XA CN113628090B (en) | 2021-07-14 | 2021-07-14 | Anti-interference message steganography and extraction method, system, computer equipment and terminal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110793854.XA CN113628090B (en) | 2021-07-14 | 2021-07-14 | Anti-interference message steganography and extraction method, system, computer equipment and terminal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113628090A true CN113628090A (en) | 2021-11-09 |
| CN113628090B CN113628090B (en) | 2024-04-02 |
Family
ID=78379717
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110793854.XA Active CN113628090B (en) | 2021-07-14 | 2021-07-14 | Anti-interference message steganography and extraction method, system, computer equipment and terminal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113628090B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114037596A (en) * | 2022-01-07 | 2022-02-11 | 湖南菠萝互娱网络信息有限公司 | End-to-end image steganography method capable of resisting physical transmission deformation |
| CN116049840A (en) * | 2022-07-25 | 2023-05-02 | 荣耀终端有限公司 | Data protection method, device, related equipment and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140357312A1 (en) * | 2010-11-04 | 2014-12-04 | Digimarc Corporation | Smartphone-based methods and systems |
| US20180068429A1 (en) * | 2015-04-15 | 2018-03-08 | Institute Of Automation Chinese Academy Of Sciences | Image Steganalysis Based on Deep Learning |
| CN111951149A (en) * | 2020-08-14 | 2020-11-17 | 中国人民武装警察部队工程大学 | Image information steganography method based on neural network |
-
2021
- 2021-07-14 CN CN202110793854.XA patent/CN113628090B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140357312A1 (en) * | 2010-11-04 | 2014-12-04 | Digimarc Corporation | Smartphone-based methods and systems |
| US20180068429A1 (en) * | 2015-04-15 | 2018-03-08 | Institute Of Automation Chinese Academy Of Sciences | Image Steganalysis Based on Deep Learning |
| CN111951149A (en) * | 2020-08-14 | 2020-11-17 | 中国人民武装警察部队工程大学 | Image information steganography method based on neural network |
Non-Patent Citations (1)
| Title |
|---|
| 付章杰;王帆;孙星明;王彦;: "基于深度学习的图像隐写方法研究", 计算机学报, no. 09 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114037596A (en) * | 2022-01-07 | 2022-02-11 | 湖南菠萝互娱网络信息有限公司 | End-to-end image steganography method capable of resisting physical transmission deformation |
| CN116049840A (en) * | 2022-07-25 | 2023-05-02 | 荣耀终端有限公司 | Data protection method, device, related equipment and system |
| CN116049840B (en) * | 2022-07-25 | 2023-10-20 | 荣耀终端有限公司 | Data protection method, device, related equipment and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113628090B (en) | 2024-04-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111598761B (en) | Anti-printing shooting image digital watermarking method based on image noise reduction | |
| Chang | Adversarial learning for invertible steganography | |
| Li et al. | Image steganography based on style transfer and quaternion exponent moments | |
| CN110232650B (en) | Color image watermark embedding method, detection method and system | |
| CN113628090B (en) | Anti-interference message steganography and extraction method, system, computer equipment and terminal | |
| CN111491170A (en) | Method for embedding watermark and watermark embedding device | |
| CN112132737B (en) | Image robust steganography method without reference generation | |
| Li et al. | Anti-compression JPEG steganography over repetitive compression networks | |
| Bhattacharyya et al. | Data hiding in images in discrete wavelet domain using PMM | |
| Chen et al. | JSNet: a simulation network of JPEG lossy compression and restoration for robust image watermarking against JPEG attack | |
| CN118158329A (en) | Image steganography and extraction information processing method and electronic equipment | |
| Liu et al. | A data hiding scheme based on U-Net and wavelet transform | |
| Zhu et al. | Destroying robust steganography in online social networks | |
| Lu et al. | Wavelet-based CNN for robust and high-capacity image watermarking | |
| CN115345768A (en) | Robust watermark attack method and system based on neural network | |
| CN114827381A (en) | Strong robustness image steganography method and system based on condition standardization flow model | |
| CN112132738A (en) | Image robust steganography method with reference generation | |
| CN114157773B (en) | Image steganography method based on convolutional neural network and frequency domain attention | |
| Bi et al. | High‐Capacity Image Steganography Algorithm Based on Image Style Transfer | |
| Wu et al. | An image authentication and recovery system based on discrete wavelet transform and convolutional neural networks | |
| Soualmi et al. | A blind watermarking approach based on hybrid Imperialistic Competitive Algorithm and SURF points for color Images’ authentication | |
| Rahim et al. | Exploiting de-noising convolutional neural networks DnCNNs for an efficient watermarking scheme: a case for information retrieval | |
| CN116112685A (en) | Image steganography method based on diffusion probability model | |
| CN110766597A (en) | Picture camouflaging method, picture camouflaging device, electronic equipment and storage medium | |
| Bagade et al. | A high quality steganographic method using morphing |
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 |