A dual dynamic mapping-based steganography method and device without embedding images

Abstract

The application provides a double dynamic mapping-based steganography method and device for non-embedded images, and belongs to the technical field of information security, which comprises the following steps: constructing a double dynamic mapping relationship between secret information and a random noise vector; obtaining a secret image generation seed corresponding to target secret information according to the double dynamic mapping relationship; the secret image generation seed is a random noise vector corresponding to the target secret information; and inputting the secret image generation seed into a pre-established steganography model for non-embedded images to perform steganography. The double dynamic mapping-based steganography method and device for non-embedded images provided by the application realize double dynamic transformation through one-to-many dynamic mapping between secret information and a random noise vector and dynamic switching of the mapping relationship, thereby alleviating the solidification problem of the existing mapping relationship and having significant flexibility.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information security technology, and in particular to a no-embedding image steganography method and device based on double dynamic mapping. BACKGROUND

[0002] The secure transmission of user information is of great importance to users. At present, whether it is government information or commercial information, the most recognized secure transmission method is to build a special network and set up a special communication system. However, it should be recognized that, due to resource constraints, time constraints and economic cost constraints, this method will only be adopted by users when necessary. Usually, the user information needs to be transmitted through a public network, and in order to ensure the security of the information transmission, the information needs to be steganographed and encrypted. The purpose of steganography is to avoid arousing the suspicion of third parties and reduce the risk of being monitored and intercepted.

[0003] At present, traditional information steganography needs to select a suitable carrier medium, and by slightly embedding and modifying it, the secret information is hidden in the carrier medium. However, with the rapid development of steganalysis technology, various steganalysis methods with high detection accuracy can effectively identify and detect the steganographic embedding traces in the secret-containing carrier, and determine whether secret information is hidden. Based on this, traditional embedded steganography cannot effectively guarantee the concealment, security and reliability of secret information, and there is a risk of exposing secret communication behavior.

[0004] Mapping steganography is a carrier-free steganography technology, also known as no-embedding steganography. By constructing the mapping relationship between secret information and carrier media, the secret-containing object is directly obtained, which can effectively resist existing steganalysis methods and is an important means to protect information security. The role of no-embedding steganography cannot be underestimated. The core of no-embedding steganography is the mapping relationship between secret information and secret-containing objects. The existing mapping relationship is basically a single static mapping relationship, which has the problems of simple structure, fixed relationship and lack of diversity. SUMMARY

[0005] The present application provides a no-embedding image steganography method and device based on double dynamic mapping, which solves the defects of the existing mapping relationship between secret information and secret-containing objects, such as simple structure, fixed relationship and lack of diversity, and realizes a flexible, dynamic and double mapping relationship.

[0006] In a first aspect, the present application provides a steganographic method for embedding-free image based on double dynamic mapping, comprising: constructing a double dynamic mapping relationship between secret information and a random noise vector; the double dynamic mapping relationship comprises a plurality of preset mapping relationships which are automatically switched over time, and each preset mapping relationship is set such that each preset segment of the secret information corresponds to each preset interval of the random noise vector, and each preset interval has a plurality of random noise values corresponding to the same preset segment of the secret information; obtaining a stego image generation seed corresponding to target secret information according to the double dynamic mapping relationship; the stego image generation seed is a random noise vector corresponding to the target secret information; inputting the stego image generation seed into a pre-established embedding-free image steganographic model for embedding-free image steganography.

[0007] According to the embedding-free image steganographic method based on double dynamic mapping provided by the present application, the construction of the double dynamic mapping relationship between the secret information and the random noise vector comprises: constructing an initial mapping relationship between the secret information and the random noise vector; adjusting the initial mapping relationship by adjusting the number of bits of the secret information in each preset segment and the interval between the preset intervals to generate a plurality of preset mapping relationships; and dynamically adjusting the plurality of preset mapping relationships at preset time intervals to generate the double dynamic mapping relationship.

[0008] According to the embedding-free image steganographic method based on double dynamic mapping provided by the present application, the obtaining of the stego image generation seed corresponding to the target secret information according to the double dynamic mapping relationship comprises: determining a target preset mapping relationship corresponding to the target secret information at the current time according to the double dynamic mapping relationship; and determining the stego image generation seed corresponding to the target secret information according to the target preset mapping relationship.

[0009] According to the embedding-free image steganographic method based on double dynamic mapping provided by the present application, the inputting of the stego image generation seed into a pre-established embedding-free image steganographic model for embedding-free image steganography comprises: generating a target stego image of the target secret information based on the embedding-free image steganographic model using the stego image generation seed in response to a stego image generation request of a sending end; and extracting the stego image generation seed of the target stego image based on the embedding-free image steganographic model in response to a secret information recovery request of a receiving end, so as to recover the target secret information using the target preset mapping relationship.

[0010] According to the image steganography method without embedding based on double dynamic mapping provided by the application, the image steganography model without embedding includes a generative adversarial network and a decoding network; the generative adversarial network includes a generator and a discriminator; the generator is used to generate a seed according to the stego image, and generate the target stego image; the discriminator is used to perform adversarial training with the generator, and optimize the generation performance of the target stego image of the generator; the decoding network is used to extract the stego image generation seed of the target stego image, and perform inverse mapping on the stego image generation seed by using the target preset mapping relationship, so as to recover the target secret information.

[0011] According to the image steganography method without embedding based on double dynamic mapping provided by the application, the generator includes a plurality of deconvolution layers connected in sequence, and the discriminator and the decoding network each include a plurality of convolution layers connected in sequence.

[0012] According to the image steganography method without embedding based on double dynamic mapping provided by the application, the generative adversarial network takes the target that enables the generator to generate a stego image capable of cheating the discriminator with a greater probability as the training target; and the decoding network takes the target that enables the difference between the stego image and the real image on the random noise vector to be less than a preset value as the training target.

[0013] In a second aspect, the application further provides an image steganography device based on double dynamic mapping without embedding, comprising:

[0014] A mapping relationship construction module is configured to construct a double dynamic mapping relationship between secret information and a random noise vector; the double dynamic mapping relationship includes a plurality of preset mapping relationships that are automatically switched over time, and each preset mapping relationship is set such that each preset segment of the secret information corresponds to each preset interval of the random noise vector one by one, and each preset interval has a plurality of random noise values corresponding to the same preset segment of the secret information.

[0015] A seed generation module is configured to obtain a stego image generation seed corresponding to target secret information according to the double dynamic mapping relationship; the stego image generation seed is a random noise vector corresponding to the target secret information.

[0016] A steganography module is configured to input the stego image generation seed into a pre-established image steganography model without embedding, so as to perform image steganography without embedding.

[0017] In a third aspect, the application provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor; when the processor executes the program, the steps of the image steganography method without embedding based on double dynamic mapping according to any one of the above aspects are implemented.

[0018] In a fourth aspect, the present application also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the steganographic method based on double dynamic mapping without embedding images according to any of the above aspects.

[0019] The steganographic method based on double dynamic mapping without embedding images provided by the present application realizes double dynamic transformation through one-to-many dynamic mapping of secret information and random noise vectors and dynamic switching of mapping relationships, and has the following beneficial effects:

[0020] (1) Flexible mapping relationship: The present application realizes double dynamic transformation through one-to-many dynamic mapping of secret information and random noise vectors and dynamic switching of mapping tables, alleviates the problem of existing mapping relationship solidification, and has significant flexibility.

[0021] (2) High diversity of stego image generation: Since the secret information and the stego image generation seed have a one-to-many dynamic mapping relationship, the same piece of secret information can generate different stego images, effectively ensuring the diversity of generated images and alleviating the problem of stego image generation homogeneity and insufficient diversity.

[0022] (3) Strong anti-steganalysis capability: By dynamically adjusting the mapping relationship between the secret information and the stego image, and generating stego images that are difficult to distinguish, the stego images generated by the present application are difficult to be detected by third parties, and have strong anti-steganalysis capability. BRIEF DESCRIPTION OF DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0024] Figure 1 is a flowchart of the steganographic method based on double dynamic mapping without embedding images provided by the present application;

[0025] Figure 2 is a schematic diagram of the double dynamic mapping relationship between secret information and random noise vectors provided by the present application;

[0026] Figure 3 is a schematic diagram of the steganographic method based on double dynamic mapping without embedding images provided by the present application;

[0027] Figure 4 is a structural schematic diagram of the generative adversarial network provided by the present application;

[0028] Figure 5 is a structural schematic diagram of a steganographic decoder network provided by the present application;

[0029] Figure 6 is a structural schematic diagram of an electronic device provided by the present application. DETAILED DESCRIPTION

[0030] For the purpose of making the object, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be described clearly and completely below with reference to the drawings in the present application. Obviously, the described embodiments are some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the present application.

[0031] It should be noted that, in the description of the embodiments of the present application, the terms “comprise”, “contain” or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitation, the element defined by the statement “comprises a…” does not exclude the presence of another identical element in the process, method, article or device comprising the element.

[0032] The present application will be described below with reference to the drawings. Figures 1-6 The steganographic method and device based on double dynamic mapping and without embedding image provided by the embodiments of the present application are described.

[0033] Figure 1 is a flowchart of the steganographic method based on double dynamic mapping and without embedding image provided by the present application, as shown in Figure 1 including but not limited to the following steps:

[0034] Step 101: constructing a double dynamic mapping relationship between secret information and a random noise vector.

[0035] The secret information in the present application is pre-processed into binary code form, which lays the foundation for the establishment of the mapping relationship. Specifically, the secret data can be encoded into a byte sequence in “UTF-8” format, and then its binary representation form is calculated to obtain binary code to generate secret information.

[0036] The double dynamic mapping relationship includes a plurality of preset mapping relationships that automatically switch over time, and each preset mapping relationship is set such that each preset segment of the secret information corresponds to each preset interval of the random noise vector one by one, and each preset interval has a plurality of random noise values corresponding to the same preset segment of the secret information.

[0037] The random noise vector is normal distribution data of [-1, 1], in order to conceal the transmission of secret information, the application carries certain secret information for each bit of the random noise vector, inputs the random noise vector into the no-embedding image steganography model (which can be a generative adversarial network), and generates a secret-containing image.

[0038] Figure 2 It is a schematic diagram of the double dynamic mapping relationship between secret information and a random noise vector provided by the application, as shown in Figure 2 The secret information is segmented, the random noise vector is correspondingly partitioned, different intervals correspond to different segments of secret information, and each interval has multiple values corresponding to the same segment of secret information.

[0039] Specifically, by constructing a one-to-many mapping relationship between secret information and a random noise vector, the first layer of dynamic change of mapping steganography is realized, and the image generation homogenization problem is alleviated. Figure 2 In the application, the secret information S is divided into multiple segments, the random noise vector Z is divided into multiple intervals, and each segment of secret information is mapped to a different interval of the random noise vector, for example, the random noise interval Z1={Z 11 ,Z 12 ,...,Z 1k} carries secret information S1, that is, any random noise value in the interval represents secret information S1.

[0040] In addition, by changing the mapping relationship between secret information and a random noise vector, multiple mapping relationship tables (i.e., representing multiple preset mapping relationships) are generated, such as Figure 2 Table 1 to Table M in the application. That is, the application also constructs multiple mapping tables for dynamic switching, realizes the second layer of dynamic change of mapping steganography, and further alleviates the mapping relationship solidification problem.

[0041] Step 102: According to the double dynamic mapping relationship, a secret-containing image generation seed corresponding to the target secret information is obtained.

[0042] The secret-containing image generation seed is a random noise vector corresponding to the target secret information.

[0043] Step 103: Input the secret-containing image generation seed into the pre-established no-embedding image steganography model for no-embedding image steganography.

[0044] The no-embedding image steganography method based on double dynamic mapping provided by the application realizes double dynamic transformation through one-to-many dynamic mapping of secret information and a random noise vector and dynamic switching of the mapping relationship, alleviates the existing mapping relationship solidification problem, and has significant flexibility.

[0045] Further, since the secret information has a one-to-many dynamic mapping relationship with the stego image generation seed, the same piece of secret information can generate different stego images, effectively ensuring the diversity of the generated images and alleviating the problem of stego image generation homogenization and lack of diversity.

[0046] Based on the content of the above embodiments, as an optional embodiment, the application further provides a double dynamic mapping based stego image embedding-free steganography method, the double dynamic mapping relationship between the secret information and the random noise vector is constructed, including: constructing an initial mapping relationship between the secret information and the random noise vector; adjusting the initial mapping relationship by adjusting the number of bits of the secret information in each preset segment and the interval between the preset intervals to generate a plurality of preset mapping relationships; dynamically adjusting a plurality of preset mapping relationships according to a preset time interval to generate the double dynamic mapping relationship.

[0047] Optionally, the number of bits of the secret information in each preset segment is the same, and the interval between the preset intervals is the same.

[0048] Optionally, the initial mapping relationship is constructed as follows:

[0049] The one-to-many mapping function between the secret information and the stego image is specifically shown in formula (1):

[0050]

[0051] wherein, S i is the i-th piece of secret information, Z i represents the mappable noise vector in the i-th random noise interval, random() represents a random number in the interval, sigma represents how many bits of secret information a random noise carries, and delta controls the interval of the random noise vector partition.

[0052] Further, the initial mapping relationship is adjusted to generate a plurality of preset mapping relationships as follows:

[0053] The mapping parameters sigma and delta are adjusted, a plurality of mapping tables are constructed, and the mapping parameters are automatically changed at a certain time interval. Specifically, assuming that a random noise carries 2 bits (2 bits) of secret information (i.e., the number of bits of the secret information in each preset segment), the interval of the preset partition of the random noise vector is 0.1, i.e., the mapping parameters sigma = 2 and delta = 0.1 are set. Subsequently, the random noise is partitioned into 4 intervals corresponding to different preset segments of secret information "00", "01", "10", "11", respectively, and each noise interval has a plurality of values representing secret information, as shown in Table 1.

[0054] Table 1 mapping relationship between secret information and random noise vector (σ = 2, δ = 0.1)

[0055] Secret information Noise adoption interval 00 random(-0.9,-0.6) 01 random(-0.4,-0.1) 10 random(+0.1,+0.4) 11 random(+0.6,+0.9)

[0056] If a random noise carries 3-bit secret information, the interval of random noise vector partition is 0.001, that is, the mapping parameters σ = 3 and δ = 0.001 are set. Then, the random noise is partitioned into 8 intervals corresponding to secret information "000", "001", "010", "011", "100", "101", "110" and "111", respectively, and each noise interval has multiple values representing secret information, as shown in Table 2.

[0057] Table 2 mapping relationship between secret information and random noise vector (σ = 3, δ = 0.001)

[0058] Secret information Noise adoption interval 000 random(-0.999,-0.751) 001 random(-0.749,-0.501) 010 random(-0.499,-0.251) 011 random(-0.249,-0.001) 100 random(0.001,0.249) 101 random(0.251,0.499) 110 random(0.501,0.749) 111 random(0.751,0.999)

[0059] In addition, the time interval T (1 minute / 2 minutes / 3 minutes) is set to adjust the mapping parameters, change the secret information carried by the random noise and the partition interval, dynamically construct multiple mapping relationship tables, and realize the second layer dynamic mapping.

[0060] Figure 3 is the framework schematic diagram of the double dynamic mapping based non-embedded image steganography method provided by the application, and the feasible technical solutions of the application will be further described below with reference to Figure 3 and specific embodiments.

[0061] Optionally, according to the double dynamic mapping relationship, a target secret information corresponding to a secret-containing image generation seed is obtained, including: according to the double dynamic mapping relationship, a target preset mapping relationship corresponding to the target secret information at the current time is determined; and according to the target preset mapping relationship, a target secret information corresponding to a secret-containing image generation seed is determined.

[0062] It can be understood that the preset mapping relationship is dynamically switched, and the application can determine the corresponding target preset mapping relationship according to the time parameter to generate the secret-containing image generation seed.

[0063] Optionally, the application inputs the stego image generation seed into a pre-established stego image embedding-free model for stego image embedding-free, including: in response to a stego image generation request of a sending end, generating a target stego image of the target secret information based on the stego image embedding-free model and using the stego image generation seed; and in response to a secret information recovery request of a receiving end, extracting a stego image generation seed of the target stego image based on the stego image embedding-free model, and recovering the target secret information by using the target preset mapping relationship.

[0064] The stego image embedding-free model includes a generative adversarial network and a decoding network; the generative adversarial network includes a generator and a discriminator; the generator is configured to generate the target stego image according to the stego image generation seed; the discriminator is configured to perform adversarial training with the generator to optimize the generation performance of the target stego image of the generator; and the decoding network (i.e., a stego decoding network) is configured to extract the stego image generation seed of the target stego image and inversely map the stego image generation seed by using the target preset mapping relationship to recover the target secret information.

[0065] Specifically, as shown in Figure 3 The stego image embedding-free method provided by the application starts from target secret information, constructs a double dynamic mapping relationship between the target secret information and a stego image generation seed by referring to the dynamic change idea, and constructs and iteratively trains a stego deep convolutional generative adversarial network (i.e., a generative adversarial network) and a stego decoding network based on the double dynamic mapping relationship.

[0066] Specifically, at the sending end, first, the target secret data to be stegoed can be binary encoded to generate target secret information. Then, mapping parameters σ, δ and T are set to obtain a stego image generation seed. Finally, the stego image generation seed is input into the generative adversarial network of the stego image embedding-free model to generate a target stego image that is realistic and not easy to be detected by a stego analyzer.

[0067] At the receiving end, the target stego image is received and input into the decoding network of the stego image embedding-free model to generate a stego image generation seed, and then the corresponding mapping relationship table is queried according to the mapping parameters σ, δ and T to inversely map and recover the secret information.

[0068] More specifically, the generator generates a stego stego image that is difficult to distinguish from a real image starting from a random noise vector; the discriminator judges whether the input sample is a real image and feeds back the result to the generator. Through the game training of the two, the generator and the discriminator dynamically adjust the network weights to optimize the network generation ability and the discrimination ability.

[0069] Figure 4is a structural schematic diagram of the generative adversarial network provided by the present application, as shown in Figure 4 The generator network has 5 layers: all are deconvolution layers, the convolution kernel size is 4*4, the step size of the 1st layer is 1, and the step size of the 2nd-5th layers is 2; the activation function of the first 4 layers is ReLu, and the activation function of the 5th layer is Tanh. The discriminator network has 5 layers: all are convolution layers, the convolution kernel size is 4*4, the step size of the 1st-4th layers is 2, and the step size of the 5th layer is 1; the activation function of the first 4 layers is LeakReLU, and the activation function of the 5th layer is Sigmoid.

[0070] Figure 5 is a structural schematic diagram of the steganographic decoder network provided by the present application, as shown in Figure 5 The steganographic decoding network has 5 layers: all are convolution layers, the convolution kernel size of the 1st layer is 5*5, the convolution kernel size of the 2nd-5th layers is 4*4, the convolution kernel size of the 6th layer is 1*1, the step size is 1, 2, 1, 2, 1, 1 respectively, and the activation function is all Leaky ReLU.

[0071] Based on the content of the above embodiment, as an optional embodiment, the steganographic method for non-embedded images based on double dynamic mapping provided by the present application, the training target of the generative adversarial network is to enable the generator to generate stego images capable of cheating the discriminator with a greater probability; the training target of the decoding network is to enable the difference between the stego image and the real image on the random noise vector to be less than a preset value. The training process of the generative adversarial network and the decoding network is described below.

[0072] Among them, the adversarial network simultaneously trains the generator G which captures the data distribution, and the discriminator D which estimates the probability that the sample comes from the real data. In the training process, the generator expects to generate a realistic stego image from the steganographic seed to cheat the discriminator, and the discriminator aims to distinguish the input sample true or false and feed back the result to the generator. Through the game between the two, the generator after training can generate a false stego image to cheat the discriminator. Assuming that the label "1" corresponds to the real image, and the label "0" corresponds to the generated stego image, the adversarial network training objective function is specifically shown in formula (2):

[0073]

[0074] Among them, D is the discriminator, G is the generator, x is the real image, Z is the random noise vector, P data represents the distribution of the real image, represents the expectation of the real image, P Z represents the distribution of the random noise vector, The expectation of generating stego images is shown. In the process of training, the discriminator hopes that the output on the real image is close to 1 and the discrimination result on the stego image is close to 0; and the generator hopes to generate a stego image that can deceive the discriminator and make the discrimination result wrong.

[0075] The loss function of the stego decoding network needs to consider the difference between the stego image and the real image in the random noise vector. When the loss value is small enough, the decoding network can accurately recover the secret information:

[0076]

[0077] wherein, is the random noise vector mapped by the stego stego image, Z is the random noise vector mapped by the real image, and ||·|| is the L1 norm distance. 1

[0078] Based on the above-mentioned loss function, the network weight is dynamically updated, and the generation ability and prediction ability of the deep neural network are optimized. Moreover, the training batch and the learning rate are dynamically updated according to the training effect. The training process can follow the following principles: ① If the sample space difference is small, the sample number of the training batch is appropriately increased; ② If the sample space difference is large, the sample number of the training batch is appropriately reduced; ③ The learning rate is dynamically adjusted according to the loss values of the generator, the discriminator and the decoding network. After the training is completed, the mapping stego network model, the network weight and the mapping parameter are encapsulated as stego metadata and physically delivered to both parties.

[0079] In summary, the stego method without embedding image based on double dynamic mapping has the following beneficial effects:

[0080] (1) Flexible mapping relationship: the present application realizes double dynamic transformation through one-to-many dynamic mapping of secret information and random noise vector and dynamic switching of the mapping table, thereby relieving the problem of fixed mapping relationship and having significant flexibility.

[0081] (2) High diversity of stego image generation: since the secret information and the stego image generation seed have a one-to-many dynamic mapping relationship, the same piece of secret information can generate different stego images, thereby effectively ensuring the diversity of the generated images and relieving the problems of stego image generation homogenization and insufficient diversity.

[0082] (3) Strong steganalysis resistance: by dynamically adjusting the mapping relationship between the secret information and the stego image and generating stego images that are difficult to distinguish, the stego images generated by the present application are difficult to be detected by a third party, and have strong steganalysis resistance.

[0083] The present application also provides a stego device without embedding image based on double dynamic mapping, comprising: ​

[0084] The mapping relationship construction module is configured to construct a double dynamic mapping relationship between the secret information and the random noise vector, wherein the double dynamic mapping relationship comprises a plurality of preset mapping relationships that are automatically switched over time, and each preset mapping relationship is set such that each preset segment of the secret information corresponds to each preset interval of the random noise vector one-to-one, and each preset interval has a plurality of random noise values corresponding to the same preset segment of the secret information.

[0085] The seed generation module is configured to obtain a stego image generation seed corresponding to target secret information according to the double dynamic mapping relationship, wherein the stego image generation seed is a random noise vector corresponding to the target secret information.

[0086] The steganography module is configured to input the stego image generation seed into a pre-established non-embedded image steganography model to perform non-embedded image steganography.

[0087] It should be noted that the non-embedded image steganography device based on double dynamic mapping provided by the embodiment of the present application can perform the non-embedded image steganography method based on double dynamic mapping described in any of the above embodiments when it is actually running, and the embodiment will not be repeated here.

[0088] Figure 6 is a structural schematic diagram of an electronic device provided by the present application, as Figure 6 shown, the electronic device can include a processor 610, a communications interface 620, a memory 630, and a communications bus 640, wherein the processor 610, the communications interface 620, and the memory 630 communicate with each other through the communications bus 640. The processor 610 can invoke the logic instructions in the memory 630 to execute the non-embedded image steganography method based on double dynamic mapping, which comprises: constructing a double dynamic mapping relationship between secret information and a random noise vector; the double dynamic mapping relationship comprises a plurality of preset mapping relationships that are automatically switched over time, and each preset mapping relationship is set such that each preset segment of the secret information corresponds to each preset interval of the random noise vector one-to-one, and each preset interval has a plurality of random noise values corresponding to the same preset segment of the secret information; obtaining a stego image generation seed corresponding to target secret information according to the double dynamic mapping relationship; the stego image generation seed is a random noise vector corresponding to the target secret information; inputting the stego image generation seed into a pre-established non-embedded image steganography model to perform non-embedded image steganography.

[0089] Moreover, the logic instructions in the memory 630 described above can be implemented in the form of software functional units and sold or used as independent products, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0090] In another aspect, the present application also provides a computer program product, which comprises a computer program stored on a non-transitory computer readable storage medium, and the computer program comprises program instructions, when the program instructions are executed by a computer, the computer can execute the steganographic method of non-embedded image based on double dynamic mapping provided by the above-mentioned embodiments, which comprises: constructing a double dynamic mapping relationship between secret information and a random noise vector; the double dynamic mapping relationship comprises a plurality of preset mapping relationships that are automatically switched over time, and each preset mapping relationship is set such that each preset segment of the secret information corresponds to each preset interval of the random noise vector one by one, and each preset interval has a plurality of random noise values corresponding to the same preset segment of the secret information; obtaining a target secret information corresponding to a stego image generation seed according to the double dynamic mapping relationship; the stego image generation seed is a random noise vector corresponding to the target secret information; inputting the stego image generation seed into a pre-established steganographic model of non-embedded image to perform steganography of non-embedded image.

[0091] In yet another aspect, the present application also provides a non-transitory computer readable storage medium having stored thereon a computer program, which, when executed by a processor, implements a dual dynamic mapping based steganography method without embedding images as provided by the above embodiments, the method comprising: constructing a dual dynamic mapping relationship between secret information and a random noise vector; the dual dynamic mapping relationship comprises a plurality of preset mapping relationships automatically switched over time, and each preset mapping relationship is set such that each preset segment of the secret information corresponds to each preset interval of the random noise vector one-to-one, and each preset interval has a plurality of random noise values corresponding to the same preset segment of the secret information; obtaining a stego image generation seed corresponding to target secret information according to the dual dynamic mapping relationship; the stego image generation seed is a random noise vector corresponding to the target secret information; and inputting the stego image generation seed into a pre-established steganography model without embedding images to perform steganography without embedding images.

[0092] The device embodiments described above are merely illustrative, wherein the units illustrated as separate components can or can not be physically separated, and the components illustrated as units can or can not be physical units, i.e., can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0093] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software plus necessary universal hardware platforms, and of course can also be realized by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0094] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An embedding-free image steganography method based on dual dynamic mapping, characterized in that, include: Construct a dual dynamic mapping relationship between secret information and random noise vectors; The dual dynamic mapping relationship includes a variety of preset mapping relationships that automatically switch over time. Each preset mapping relationship is set such that each preset segment of the secret information corresponds one-to-one with each preset interval of the random noise vector, and each preset interval has multiple random noise values ​​corresponding to the same preset segment of the secret information. Based on the dual dynamic mapping relationship, a seed for generating a secret image corresponding to the target secret information is obtained; the seed for generating a secret image is a random noise vector corresponding to the target secret information. The dense image generation seed is input into a pre-established non-embedded image steganography model to perform non-embedded image steganography. The construction of the dual dynamic mapping relationship between secret information and random noise vectors includes: Construct an initial mapping relationship between the secret information and the random noise vector; By adjusting the number of bits of secret information within each preset segment and the interval between the preset intervals, the initial mapping relationship is adjusted to generate multiple preset mapping relationships; According to a preset time interval, multiple preset mapping relationships are dynamically adjusted to generate the dual dynamic mapping relationship; The step of obtaining the seed image for generating the secret image corresponding to the target secret information based on the dual dynamic mapping relationship includes: Based on the dual dynamic mapping relationship, determine the target preset mapping relationship corresponding to the target secret information at the current moment; Based on the target preset mapping relationship, determine the secret image generation seed corresponding to the target secret information.

2. The embedding-free image steganography method based on dual dynamic mapping according to claim 1, characterized in that, The step of inputting the dense image generation seed into a pre-established embedding-free image steganography model to perform embedding-free image steganography includes: In response to the request to generate a cryptic image from the sending end, a target cryptic image containing the target secret information is generated using the cryptic image generation seed based on the embedding-free image steganography model. In response to a request for recovery of private information from the receiving end, based on the non-embedded image steganography model, a seed image for generating the target secret image is extracted, so as to recover the target secret information by utilizing the target preset mapping relationship.

3. The embedding-free image steganography method based on dual dynamic mapping according to claim 2, characterized in that, The embedding-free image steganography model includes a generative adversarial network and a decoding network; The generative adversarial network includes a generator and a discriminator; wherein, the generator is used to generate a seed based on the dense image to generate the target dense image; the discriminator is used to perform adversarial training with the generator to optimize the generator's performance in generating the target dense image; The decoding network is used to extract the seed image of the target secret image and to perform inverse mapping on the seed image using the target preset mapping relationship to recover the target secret information.

4. The embedding-free image steganography method based on dual dynamic mapping according to claim 3, characterized in that, The generator includes multiple deconvolutional layers connected in sequence, and both the discriminator and the decoding network include multiple convolutional layers connected in sequence.

5. The embedding-free image steganography method based on dual dynamic mapping according to claim 3, characterized in that, The generative adversarial network is trained with the goal of enabling the generator to generate dense images that can deceive the discriminator with a higher probability. The decoding network is trained with the objective of making the difference between the dense image and the real image in the random noise vector less than a preset value.

6. An embedding-free image steganography device based on dual dynamic mapping, characterized in that, An image steganography method without embedding as described in any one of claims 1 to 5 includes: A mapping relationship construction module is used to construct a dual dynamic mapping relationship between secret information and random noise vectors. The dual dynamic mapping relationship includes a variety of preset mapping relationships that automatically switch over time. Each preset mapping relationship is set such that each preset segment of the secret information corresponds one-to-one with each preset interval of the random noise vector, and each preset interval has multiple random noise values ​​corresponding to the same preset segment of the secret information. The seed generation module is used to obtain a seed from a cryptic image corresponding to the target secret information based on the dual dynamic mapping relationship; the seed from the cryptic image is a random noise vector corresponding to the target secret information. The steganography module is used to input the seed of the dense image into a pre-established non-embedded image steganography model to perform non-embedded image steganography.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the embedding-free image steganography method based on dual dynamic mapping as described in any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the embedding-free image steganography method based on dual dynamic mapping as described in any one of claims 1 to 5.

Images