A method and apparatus for information steganography

Abstract

The application discloses a steganography method and device. The method comprises the following steps: obtaining a target stego sequence corresponding to a target carrier image; obtaining detection data for representing whether secret information is embedded in the target carrier image, and calculating a gradient image corresponding to the target carrier image by using a back propagation algorithm based on the detection data; and modifying the target stego sequence based on the gradient image to obtain a target stego image. The method uses the gradient of a steganalysis analyzer to perform dithering modulation modification on the target carrier image, so that the target stego image has higher anti-steganalysis capability. Compared with existing robust steganography algorithms, the method has improved robustness and security, and can meet the requirements of security and robustness in the information steganography process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to an information steganography method and device, an electronic device and a storage device. BACKGROUND

[0002] Steganography is a technique of hiding specific secret information in a public non-secret digital carrier for transmission, so that the observer or steganalysis cannot detect its existence. It uses the data redundancy characteristics of image, voice and other carrier sources and the insensitivity of human perception organs to hide secret information in digital carriers in a certain encoding, modulation or encryption manner. Steganography has been applied in many fields in today's digital world, such as copyright protection, information addition, tamper-proofing, information tracking, etc. Its implementation principle is: using the redundancy of carrier information in space, time and storage (i.e. information redundancy) to embed secret information into the carrier to generate stego information. This process needs to embed hidden information into the carrier in a special way without damaging the original expression of the carrier information. The purpose is to protect the information that needs to be hidden from being identified by others.

[0003] In order to achieve high robustness, the existing steganography has a large modification amount and modification range of the original image carrier, which leads to a large change in the statistical characteristics of the image after embedding secret information, so that it is easily detected by steganalysis. Or the existing steganography can provide high security performance, i.e. the performance of not being detected by steganalysis, but its robustness is often not guaranteed. SUMMARY

[0004] The present application provides an information steganography method, device, electronic device and computer readable storage medium to solve the problem that the existing steganography technology cannot simultaneously satisfy security and robustness.

[0005] The present application provides an information steganography method, characterized in that it comprises:

[0006] obtaining a target stego sequence corresponding to a target carrier image;

[0007] obtaining detection data for representing whether secret information is embedded in the target carrier image, and calculating a gradient map corresponding to the target carrier image using a back propagation algorithm based on the detection data;

[0008] modifying the target stego sequence based on the gradient map to obtain a target stego image.

[0009] Optionally, the step of modifying the target stego sequence based on the gradient map to obtain a target stego image comprises:

[0010] a direction opposite to the gradient value in the gradient map corresponding to the DCT coefficient of the target stego sequence is determined as a modification direction of the DCT coefficient;

[0011] The target stego sequence is dithered according to the modification direction, and the target carrier image is modified according to the corresponding dithered result of the target stego sequence, to obtain a target stego image.

[0012] Optionally, the obtaining of the target stego sequence corresponding to the target carrier image comprises:

[0013] selecting an embedding region for the target carrier image;

[0014] obtaining a carrier sequence to be embedded with secret information based on the selected embedding region;

[0015] quantizing the carrier sequence according to a preset quantization step;

[0016] performing a permutation on the quantized carrier sequence to obtain a permuted carrier sequence;

[0017] embedding the secret information into the permuted carrier sequence to obtain a permuted stego sequence;

[0018] performing an inverse permutation on the permuted stego sequence to obtain the target stego sequence.

[0019] Optionally, the selecting of the embedding region for the target carrier image comprises:

[0020] in response to a transmission channel quality factor of the target stego image being known, selecting a frequency band region close to an upper left part of the target carrier image as the embedding region; or.

[0021] in response to the transmission channel quality factor of the target stego image being unknown, selecting a frequency band region close to a lower right part of the target carrier image as the embedding region.

[0022] Optionally, the frequency band region close to the upper left part of the target carrier image is E4 and E5 positions in an 8*8 DCT block corresponding to the target carrier image, and the frequency band region close to the lower right part of the target carrier image is E6 and E7 positions in the 8*8 DCT block corresponding to the target carrier image.

[0023] Optionally, the quantizing of the carrier sequence comprises:

[0024] quantizing elements in the embedding region to 0 or 1.

[0025] Optionally, the method further comprises: performing JPEG recompression processing on the target stego image.

[0026] The embodiment of the present application also provides an information steganography device, comprising:

[0027] An object carrier sequence obtaining unit is configured to obtain an object carrier sequence corresponding to the target carrier image;

[0028] A gradient map obtaining unit is configured to obtain detection data for characterizing whether secret information is embedded in the target carrier image, and calculate a gradient map corresponding to the target carrier image by using a back propagation algorithm based on the detection data;

[0029] An object carrier image obtaining unit is configured to modify the object carrier sequence based on the gradient map, and obtain an object carrier image.

[0030] Compared with the prior art, the present application has the following advantages:

[0031] The information steganography method provided by the present application comprises the following steps: obtaining an object carrier sequence corresponding to a target carrier image; obtaining detection data for characterizing whether secret information is embedded in the target carrier image, and calculating a gradient map corresponding to the target carrier image by using a back propagation algorithm based on the detection data; modifying the object carrier sequence based on the gradient map, and obtaining an object carrier image. The method uses the gradient of a steganalysis device to perform dithering modulation modification on the target carrier image. The gradient can indicate how to modify the target carrier image to make the generated object carrier image safer, so that the object carrier image can pass the detection of the steganalysis device, and the ability of the object carrier image to resist steganalysis is higher. Compared with existing robust steganography algorithms, the method has improved robustness and safety, and can meet the requirements of safety and robustness in the information steganography process. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 FIG. 1 is a flowchart of an information steganography method provided by a first embodiment of the present application;

[0033] Figure 1 -A FIG. 2 is a unit block diagram of an information steganography device provided by the first embodiment of the present application;

[0034] Figure 1 -B FIG. 3 is a comparison diagram of the robustness of a method (Proposed-Unaware) provided by the present application and GMAS and DMMR methods;

[0035] Figure 1 -C FIG. 4 is a comparison diagram of the safety of the method (Proposed-Unaware) provided by the present application and GMAS and DMMR methods;

[0036] Figure 2 FIG. 5 is a unit block diagram of an information steganography device provided by a second embodiment of the present application;

[0037] Figure 3 FIG. 1 is a schematic diagram of a logical structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0038] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details set forth in this description. In other instances, well-known methods have not been described in detail in order not to unnecessarily obscure aspects of the present application.

[0039] Digital image watermarking and steganography are both techniques of embedding secret information in digital images, but they have different focuses. Digital image watermarking pays more attention to robustness, i.e., the image containing the watermark can still successfully extract the embedded information after attacks such as scaling, cropping, JPEG compression, etc. In order to meet this requirement, the modification amount and modification amplitude of the original image are often large, thereby resulting in low security. Steganography pays more attention to security, i.e., the image containing the secret information embedded by steganography is difficult to be detected by a detector, and the modification amplitude of the original image is often small. However, traditional steganography assumes that the image is transmitted in an ideal channel, i.e., it will not undergo JPEG compression and other attacks during transmission, so the image containing the information embedded by steganography may not be able to successfully extract the information after the attack. With the increasing application of digital image watermarking, various attack scenarios triggered thereby are also increasing, for example, an attacker will first detect the received digital image to determine whether it contains secret information, and if so, will maliciously attack it, so that it cannot correctly extract the embedded information.

[0040] Robustness refers to stability, i.e., the ability of the embedded carrier information to recover to the original state even after a certain degree of interference and maintain the accuracy of the hidden information. That is, the stability of the hidden information needs to be ensured and it cannot be easily disturbed. Security refers to confidentiality, i.e., the ability of the carrier to maintain the information from being perceived and extracted even after being maliciously analyzed and attacked by an attacker. However, most existing steganography cannot guarantee security while ensuring robustness, for example, image blind watermarking, which adds a watermark to the original image in an invisible form and does not greatly affect the quality of the original image, and can resist various image theft attacks such as cropping and smearing. However, in order to achieve high robustness, the modification amplitude and proportion of the pixel values of the carrier image are large, resulting in a decrease in security performance, i.e., the ability to resist steganalysis detection.

[0041] In order to improve the ability of robustness-based stego image against steganalysis, for example, in order to enable the digital image embedded with secret information to resist malicious detection, and after transmission through the network channel, the corresponding secret information can still be successfully extracted, the present application provides an information steganography method, a device corresponding to the method, an electronic device and a readable storage medium. The following provides embodiments to explain the above method, device, electronic device and computer readable storage medium in detail.

[0042] Firstly, the related terms of the embodiments of the present application are explained as follows:

[0043] Dither modulation: the coefficients d (DCT (Discrete Cosine Transform) coefficients) of the carrier elements are divided into odd and even intervals according to the selected quantization step q, which represent message bits 0 and 1 respectively. In order to embed message bits w with the smallest modification distance h, the coefficients d are reasonably quantized, so that the embedded message bits w can be represented by the interval of the quantized result d'.

[0044] Steganalysis: steganalysis refers to the technology of detecting and judging whether there is secret information in the observed data under the condition of known or unknown embedding algorithm, which is based on detecting the changes of the characteristics of the digital carrier before and after steganography to judge whether the digital carrier contains secret information embedded by steganography. The steganalysis refers to a binary classification network model that detects the input digital carrier by steganalysis technology and outputs the detection result of whether the digital carrier is embedded with secret information.

[0045] Robust steganography: traditional steganography algorithm assumes that the image is transmitted in a lossless channel, i.e. without lossy image processing operation. The image embedded with information by this kind of method cannot successfully extract the corresponding information after some attacks such as compression, scaling, etc. Robust steganography algorithm requires both security, i.e. anti-steganalysis detection ability, and robustness, i.e. still being able to correctly extract secret information after lossy image operation processing.

[0046] The first embodiment of the present application provides an information steganography method, the execution subject of which can be a computing device application for generating stego image, which can run on a terminal or a server. The following will be described in combination with Figure 1 .

[0047] As shown in Figure 1 , the information steganography method provided by the present embodiment includes the following steps:

[0048] S101, obtaining a target stego sequence corresponding to a target carrier image.

[0049] This step is used to obtain a target stego sequence corresponding to a target carrier image, which can specifically include the following contents:

[0050] First, an embedding region is selected for the target carrier image. In this embodiment, the process of selecting the embedding region can be specifically: in response to the transmission channel quality factor of the target stego image being known, selecting a frequency band region close to the upper left part of the target carrier image as the embedding region; or, in response to the transmission channel quality factor of the target stego image being unknown, selecting a frequency band region close to the lower right part of the target carrier image as the embedding region, the basic principle being: the transmission channel is known, the ability to resist attacks is strong, and the robustness is strong enough, and on the premise that the robustness is strong enough, the security should be considered more, that is, the security requirement is higher, the more secure the better, and relatively speaking, the frequency of the frequency band region close to the upper left part of the target carrier image is lower than that of the frequency band region close to the lower right part of the target carrier image, and the security is better; on the contrary, in the case where the transmission channel is unknown, the robustness requirement is higher. In this embodiment, the frequency band region close to the upper left part of the target carrier image is the E4 and E5 positions in the 8*8 DCT block corresponding to the target carrier image; and the frequency band region close to the lower right part of the target carrier image is the E6 and E7 positions in the 8*8 DCT block corresponding to the target carrier image.

[0051] Secondly, based on the selected embedding region, a carrier sequence to be embedded with secret information is obtained, and the carrier sequence is quantized according to a preset quantization step, for example, the elements in the embedding region are quantized to 0 or 1.

[0052] Then, the carrier sequence after quantization processing is subjected to a scrambling process to obtain a scrambled carrier sequence.

[0053] Finally, the secret information is embedded into the scrambled carrier sequence (embedding bits are selected for the secret information), a stego sequence after scrambling is obtained, and the stego sequence after scrambling is subjected to an anti-scrambling process to obtain the target stego sequence.

[0054] S102, obtain detection data for characterizing whether secret information is embedded in the target carrier image, and based on the detection data, use a back propagation algorithm to calculate a gradient graph corresponding to the target carrier image.

[0055] This step is used to obtain detection data for characterizing whether secret information is embedded in the target carrier image, and based on the detection data, use a back propagation algorithm to calculate a gradient graph corresponding to the target carrier image, for example, input the above target carrier image into a pre-trained steganalysis analyzer, and based on the output result output by the steganalysis analyzer for characterizing whether secret information is embedded in the target carrier image, use a back propagation algorithm to calculate a gradient graph corresponding to the target carrier image.

[0056] The back propagation algorithm (BP) is a short form of "error back propagation". The method calculates the gradient of the loss function for all weights in the network, which is used to update the weights to minimize the loss function. Back propagation requires the known output for each input value to calculate the gradient of the loss function, that is, the back propagation algorithm is based on the gradient descent method, and the direction of the gradient indicates the direction of error expansion, so the weight needs to be taken in the opposite direction when updating the weight, so as to reduce the error caused by the weight.

[0057] It should be noted that the implementation order of steps S101 and S102 is not limited, that is, the target steganographic sequence corresponding to the target carrier image can also be obtained after the gradient map corresponding to the target carrier image is calculated through step S102, or both can be implemented simultaneously.

[0058] S103, modifying the target steganographic sequence based on the gradient map to obtain a target steganographic image.

[0059] After obtaining the target steganographic sequence corresponding to the target carrier image and calculating the gradient map corresponding to the target carrier image using the back propagation algorithm, this step is used to modify the target steganographic sequence based on the gradient map to obtain a target steganographic image.

[0060] For a JPEG image, secret information is embedded based on a JPEG image frequency domain steganography algorithm, and the secret information is embedded by modifying the DCT coefficient. The modification of the target steganographic sequence based on the gradient map to obtain a target steganographic image can be specifically: determining the opposite direction of the gradient corresponding to the DCT coefficient of the target steganographic sequence in the gradient map as the modification direction of the DCT coefficient; and performing dither modulation on the target steganographic sequence according to the modification direction, and modifying the target carrier image according to the modulation result corresponding to the target steganographic sequence to obtain a target steganographic image. In this process, the gradient corresponding to the DCT coefficient can indicate how to modify the generated steganographic image to be more secure, that is, to make the steganographic image more close to the carrier image without embedding secret information, so as to pass the detection of the steganalysis analyzer.

[0061] It should be noted that when the application scenario is known, the target steganographic image can also be subjected to JPEG recompression processing to improve the robustness.

[0062] The specific process of obtaining the target steganographic image is shown in the following example:

[0063] Step 0: For a given target carrier image X, extract its quantization table Q C, obtain the dequantized DCT coefficients of the target carrier image, and calculate the cost function p of the carrier image using an existing adaptive steganography method (such as the J-UNIWARD algorithm).

[0064] Step 1: Extract the corresponding robust embedding region according to different application scenarios, such as Figure 1 -A If the transmission channel quality factor is known, the selected robust embedding region is the E4 and E5 positions in each 8*8 DCT block; if the transmission channel quality factor is unknown, the selected robust embedding region is the E6 and E7 positions in each 8*8 DCT block; let the set of robust elements (DCT coefficients) be X emb .

[0065] Step 2: Use a generalized quantization modulation algorithm to quantize and modulate the robust elements in X emb , obtain the carrier sequence C, and calculate the corresponding modification distances d + and d - based on the preset quantization step.

[0066] Step 3: Calculate the cost of modifying the carrier sequence C based on the quantization table Qc of the carrier image and the cost function p, as well as the modification distances obtained in Step 2. The calculation method is shown in formula (1).

[0067]

[0068] Step 4: Randomly shuffle the carrier sequence c, and use an existing encoding algorithm (such as the STC-CRC encoding method) to embed the secret information m into C, obtaining the target stego sequence S.

[0069] Step 5: Reverse the target stego sequence S. Place the image X into the trained steganalysis network N T , and use the backpropagation algorithm (formula 2) to obtain the gradient G X of the image X, where setting t = 0 represents the label of the carrier image.

[0070]

[0071] L(X, t; N T ) = -t log(N T (X)) - (1-t) log(1-N T (X)) cross-entropy function

[0072] The meaning of formula 2 is: X is the input of the steganalysis network N T , and L(X, t; N T ) is the output of the steganalysis network N TThe loss function loss between (X) and the target label t is used, and then the gradient G is obtained by differentiating the loss with respect to X (calculating the partial derivative of the loss with respect to X). X .

[0073] For input X, the output N of the steganalysis network T (X) is a number in the interval [0, 1], where 0 represents a carrier image tag and 1 represents a secret image tag. The number in the interval [0, 1] represents the probability that the steganalysis network determines X to have embedded secret information or not. Assume that the output of the steganalysis network is N. T (X) = 0.4, and the desired result is that the output of the steganalysis network is the target label, i.e., t = 0 (the target label t is set to 0), then the output of the steganalysis network is N. T There is a gap (loss) between (X) = 0.4 and the target label t = 0, and their cross-entropy distance is 0.4. Using the cross-entropy function in Equation 2, we can calculate the loss = log(0.6), and then backpropagate this loss to obtain the gradient G. X .

[0074] Then, using anti-jitter modulation, based on gradient G... X Modify the dequantized DCT coefficients in the target data-carrying sequence S in the opposite direction to obtain the target data-carrying image Y. A .

[0075]

[0076] Formula (3) represents the meaning of: when the gradient When less than 0 (negative), according to the... The opposite positive direction is opposite to the DCT block X. ij Modify the dequantized DCT coefficients; when the gradient When greater than 0 (positive), follow the same rules as... The opposite negative direction for DCT block X ij The dequantized DCT coefficients are modified.

[0077] The target label t is set to 0 (0 represents the carrier image label, 1 represents the carrier image label), and the gradient G is... X The meaning is: to guide how to modify the image to achieve the desired result; that is, if the carrier image is to be modified, the goal is to make the modified carrier image more detectable as a carrier image by the steganalysis network. Since the target carrier image itself is a modified version of the target carrier image, how can the carrier image be modified so that the modified carrier image is more easily identified as a carrier image by the steganalysis network? This embodiment uses the inverse gradient algorithm. For example, the output N of the steganalysis network... T(X) = 0.4, set the target label t = 0, calculate the gradient G X , according to the opposite direction of the gradient G X modify X, that is, the opposite direction of the gradient corresponding to the DCT coefficient of the target steganographic sequence in the gradient map is determined as the modification direction of the DCT coefficient, the target steganographic sequence is dither modulated according to the modification direction, and the target carrier image is modified according to the modulation result corresponding to the target steganographic sequence, to obtain the target steganographic image, which can make the output N T (X) of the steganalysis network closer to the target label 0 (representing the carrier image), thereby effectively improving the ability of the target steganographic image Y A resisting steganalysis detection.

[0078] In the embodiment, the steganographic image extraction process is as follows:

[0079] The STC-CRC decoding process is as follows: the steganographic sequence S is shuffled; the shuffled steganographic sequence is divided into three segments S1, S2 and S3; the check code of S2 is extracted from the steganographic sequence segment S3 using STC decoding; the error-corrected steganographic sequence segment S2 is obtained by decoding S2 and the check code of S2 using RS decoding; the check code of S1 is extracted from the steganographic sequence segment S2 using STC decoding; the error-corrected steganographic sequence segment S1 is obtained by decoding S1 and the check code of S1 using RS decoding; and the secret information is extracted from the steganographic sequence segment S1 using STC decoding.

[0080] The information steganography method provided in the embodiment obtains a target steganographic sequence corresponding to a target carrier image; obtains detection data for representing whether secret information is embedded in the target carrier image, and calculates a gradient map corresponding to the target carrier image using a back propagation algorithm based on the detection data; and modifies the target steganographic sequence based on the gradient map to obtain a target steganographic image. Compared with existing solutions, the information steganography method provided in the present application determines the modification direction of the DCT coefficient as the opposite direction of the gradient corresponding to the DCT coefficient in the gradient map, dither modulates the target steganographic sequence according to the modification direction, and modifies the target carrier image according to the modulation result corresponding to the target steganographic sequence to obtain the target steganographic image. By modifying the target carrier image in this way, the output of the steganalyzer can be made closer to the target label 0 (representing the carrier image), thereby effectively improving the ability of the target steganographic image to resist steganalysis detection (including traditional steganalyzers designed manually and steganalyzers based on neural networks), while ensuring robustness and high security.

[0081] The implementation effects of the present application are illustrated by the following experiments:

[0082] Experimental setup:

[0083] Dataset: 20000 images

[0084] Data split: training set: 14000 images; validation set: 1000 images; test set: 5000 images

[0085] Comparison method: existing GMAS algorithm and DMMR algorithm

[0086] Steganalysis: DCTR, CovNet, J-YeNet

[0087] Comparison index: robustness, using 20000 images to average extraction error rate R error As a measure index; security, using detection accuracy Acc = 1 / 2(P FA + P MD ) as a measure index, wherein P FA is the false alarm rate, P MD is the missed detection rate, and the lower the Acc, the safer the algorithm.

[0088] Experimental results: robustness is shown in Figure 1 -B , and security is shown in Figure 1 -C .

[0089] Figure 1 -B The comparison schematic diagram of the average extraction error rate R error (characterizing robustness) of the method provided by the embodiment (Proposed-Unaware) and the existing GMAS and DMMR information steganography methods under different payloads in the case of known or unknown transmission channel is shown.

[0090] Figure 1 -C The comparison schematic diagram of the detection accuracy (characterizing security) of the method provided by the embodiment (Proposed-Unaware) and the existing GMAS and DMMR methods when various different steganalysis is used for detection is shown.

[0091] It can be known from the above comparison schematic diagram that the information steganography method provided by the embodiment has better robustness and security compared with the existing GMAS, DMMR and other information steganography methods.

[0092] The information steganography method provided by the embodiments of the present application is applicable to both known and unknown transmission channel application scenarios. The scheme calculates the modification cost of a carrier image by using a steganography algorithm and constructs a corresponding robust element, modulates the robust element by using a generalized dither modulation algorithm to obtain a corresponding modification distance, embeds secret information to be embedded into a target carrier image by using STC-CRC encoding technology, inputs the target carrier image into a pre-trained steganalysis analyzer, and based on an output result output by the steganalysis analyzer and used to represent whether secret information is embedded in the target carrier image, calculates and obtains a gradient image corresponding to the target carrier image by using a back propagation algorithm, and modifies the target carrier image by using the gradient, thereby generating a target stego image with higher security. For example, the target carrier image is modified by using an adversarial dither modulation method, that is, the target stego sequence is dither modulated in the opposite direction of the gradient, and the target carrier image is modified according to the modulation result of the target stego sequence, to obtain the target stego image, so that the target stego image has higher adversarial steganalysis capability. Compared with existing robust steganography algorithms, the method has improved robustness and security.

[0093] Corresponding to the data processing method provided by the first embodiment of the present application, the second embodiment of the present application provides a data processing device.

[0094] As shown in Figure 2 The information steganography device includes:

[0095] A target stego sequence obtaining unit 201 is configured to obtain a target stego sequence corresponding to a target carrier image.

[0096] A gradient image obtaining unit 202 is configured to obtain detection data used to represent whether secret information is embedded in the target carrier image, and calculate and obtain a gradient image corresponding to the target carrier image by using a back propagation algorithm based on the detection data.

[0097] A target carrier image obtaining unit 203 is configured to modify the target stego sequence based on the gradient image to obtain a target stego image.

[0098] Optionally, the modification of the target stego sequence based on the gradient image to obtain the target stego image includes:

[0099] The opposite direction of a gradient value in the gradient image corresponding to a DCT coefficient of the target stego sequence is determined as a modification direction of the DCT coefficient.

[0100] The target stego sequence is dither modulated in the modification direction, and the target carrier image is modified according to the modulation result of the target stego sequence to obtain the target stego image.

[0101] Optionally, the obtaining the target stego sequence corresponding to the target stego image comprises:

[0102] selecting an embedding region for the target stego image;

[0103] obtaining a carrier sequence to be embedded with secret information based on the selected embedding region;

[0104] performing quantization processing on the carrier sequence according to a preset quantization step;

[0105] performing shuffling processing on the carrier sequence after the quantization processing to obtain a shuffled carrier sequence;

[0106] embedding the secret information into the shuffled carrier sequence to obtain a shuffled stego sequence;

[0107] performing anti-shuffling processing on the shuffled stego sequence to obtain the target stego sequence.

[0108] Optionally, the selecting the embedding region for the target stego image comprises:

[0109] in response to a transmission channel quality factor of the target stego image being known, selecting a frequency band region close to an upper left part of the target stego image as the embedding region; or.

[0110] in response to the transmission channel quality factor of the target stego image being unknown, selecting a frequency band region close to a lower right part of the target stego image as the embedding region.

[0111] Optionally, the frequency band region close to the upper left part of the target stego image is E4 and E5 positions in an 8*8 DCT block corresponding to the target stego image; and the frequency band region close to the lower right part of the target stego image is E6 and E7 positions in the 8*8 DCT block corresponding to the target stego image.

[0112] Optionally, the performing quantization processing on the carrier sequence comprises:

[0113] quantizing elements in the embedding region to 0 or 1.

[0114] Optionally, the method further comprises: performing JPEG recompression processing on the target stego image.

[0115] Corresponding to the information steganography method provided in the first embodiment of the present application, the fourth embodiment of the present application further provides an electronic device.

[0116] As Figure 3As shown, the electronic device comprises: a processor 301; and a memory 302 for storing a program of a data processing method, after the device is powered on and the processor runs the program of the data processing method, the following steps are performed:

[0117] obtaining a target stego sequence corresponding to the target carrier image;

[0118] obtaining detection data for characterizing whether secret information is embedded in the target carrier image, and based on the detection data, using a back propagation algorithm to calculate a gradient map corresponding to the target carrier image;

[0119] based on the gradient map, modifying the target stego sequence to obtain a target stego image.

[0120] Optionally, the step of modifying the target stego sequence based on the gradient map to obtain a target stego image comprises:

[0121] determining the opposite direction of the gradient value in the gradient map corresponding to the DCT coefficient of the target stego sequence as the modification direction of the DCT coefficient;

[0122] dithering the target stego sequence according to the modification direction, and modifying the target carrier image according to the modulation result corresponding to the target stego sequence to obtain a target stego image.

[0123] Optionally, the step of obtaining a target stego sequence corresponding to the target carrier image comprises:

[0124] selecting an embedding region for the target carrier image;

[0125] based on the selected embedding region, obtaining a carrier sequence to be embedded with secret information;

[0126] quantizing the carrier sequence according to a preset quantization step;

[0127] performing shuffling processing on the quantized carrier sequence to obtain a shuffled carrier sequence;

[0128] embedding the secret information into the shuffled carrier sequence to obtain a shuffled stego sequence;

[0129] performing anti-shuffling processing on the shuffled stego sequence to obtain the target stego sequence.

[0130] Optionally, the step of selecting an embedding region for the target carrier image comprises:

[0131] In response to the transmission channel quality factor of the target stego image being known, a frequency band region close to the upper left part of the target carrier image is selected as the embedding region; or.

[0132] In response to the transmission channel quality factor of the target stego image being unknown, a frequency band region close to the lower right part of the target carrier image is selected as the embedding region.

[0133] Optionally, the frequency band region close to the upper left part of the target carrier image is E4 and E5 positions in an 8*8 DCT block corresponding to the target carrier image; and the frequency band region close to the lower right part of the target carrier image is E6 and E7 positions in the 8*8 DCT block corresponding to the target carrier image.

[0134] Optionally, the quantization processing of the carrier sequence comprises:

[0135] Quantizing elements in the embedding region to 0 or 1.

[0136] Optionally, the method further comprises: performing JPEG recompression processing on the target stego image.

[0137] Corresponding to the information steganography method provided by the first embodiment of the present application, the fourth embodiment of the present application provides a storage device storing a program of a data processing method, the program being run by a processor to execute the following steps: obtaining a target stego sequence corresponding to a target carrier image;

[0138] Obtaining detection data for characterizing whether secret information is embedded in the target carrier image, and based on the detection data, using a back propagation algorithm to calculate a gradient map corresponding to the target carrier image;

[0139] Based on the gradient map, modifying the target stego sequence to obtain a target stego image.

[0140] Optionally, the modification of the target stego sequence based on the gradient map to obtain a target stego image comprises:

[0141] Determining the opposite direction of the gradient value in the gradient map corresponding to the DCT coefficient of the target stego sequence as the modification direction of the DCT coefficient;

[0142] According to the modification direction, performing dither modulation on the target stego sequence, and according to the modulation result corresponding to the target stego sequence, modifying the target carrier image to obtain a target stego image.

[0143] Optionally, the obtaining of the target stego sequence corresponding to the target carrier image comprises:

[0144] Selecting an embedding region for the target carrier image;

[0145] based on the selected embedding region, obtaining a carrier sequence to be embedded with secret information;

[0146] quantizing the carrier sequence according to a preset quantization step;

[0147] performing a scrambling process on the quantized carrier sequence to obtain a scrambled carrier sequence;

[0148] embedding the secret information into the scrambled carrier sequence to obtain a scrambled stego sequence;

[0149] performing an anti-scrambling process on the scrambled stego sequence to obtain the target stego sequence.

[0150] Optionally, the selecting an embedding region for the target carrier image comprises:

[0151] in response to a transmission channel quality factor of the target stego image being known, selecting a frequency band region close to an upper left part of the target carrier image as the embedding region; or.

[0152] in response to the transmission channel quality factor of the target stego image being unknown, selecting a frequency band region close to a lower right part of the target carrier image as the embedding region.

[0153] Optionally, the frequency band region close to the upper left part of the target carrier image is E4 and E5 positions in an 8*8 DCT block corresponding to the target carrier image; and the frequency band region close to the lower right part of the target carrier image is E6 and E7 positions in the 8*8 DCT block corresponding to the target carrier image.

[0154] Optionally, the quantizing the carrier sequence comprises:

[0155] quantizing elements in the embedding region to 0 or 1.

[0156] Optionally, the method further comprises: performing a JPEG recompression process on the target stego image.

[0157] It should be noted that the detailed description of the storage device provided in the fourth embodiment of the present application can refer to the related description of the first embodiment of the present application, which will not be repeated here.

[0158] The present application is disclosed as above with reference to the preferred embodiments, but is not intended to limit the present application, and any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of the present application, therefore the protection scope of the present application should be subject to the scope defined by the claims of the present application.

[0159] In one typical arrangement, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0160] The memory can include non-persistent memory and / or volatile memory, such as random access memory (RAM) and / or cache memory, non-volatile memory, such as read-only memory (ROM), EPROM, and / or flash memory, etc. The memory is an example of computer readable media.

[0161] 1. Computer readable media includes both permanent and non-permanent, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile discs (DVDs) or other optical storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. According to the definition herein, computer readable media does not include transitory media, such as modulated data signals and carrier waves.

[0162] 2. Those skilled in the art will appreciate that embodiments of the present application can be provided as a method, system or computer program product. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product embodied in one or more computer usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage etc.) having computer usable program code embodied thereon.

[0163] Although the present application has been disclosed in its preferred embodiments with reference to the accompanying drawings, it is not intended to limit the present application to the preferred embodiments. It will be apparent to those skilled in the art that various changes and modifications can be made thereto without departing from the spirit and scope of the present application. Therefore, the scope of the present application should be defined by the appended claims.

Claims

1. A method of information steganography, characterized by, The method comprises the following steps: obtaining a target stego sequence corresponding to a target carrier image; obtaining detection data for characterizing whether secret information is embedded in the target carrier image, and based on the detection data, using a back propagation algorithm to calculate a gradient map corresponding to the target carrier image, comprising: inputting the target carrier image into a pre-trained steganalysis analyzer, and based on the output result of the steganalysis analyzer for characterizing whether secret information is embedded in the target carrier image, using a back propagation algorithm to calculate a gradient map corresponding to the target carrier image, wherein the back propagation algorithm is used to calculate the gradient of the loss function for the weight in the steganalysis analyzer, and the gradient is used to update the weight of the steganalysis analyzer to minimize the loss function; based on the gradient map, modifying the target stego sequence to obtain a target stego image.

2. The method of claim 1, wherein, The method of modifying the target stego sequence based on the gradient map to obtain a target stego image comprises: determining the opposite direction of the gradient value in the gradient map corresponding to the DCT coefficient of the target stego sequence as the modification direction of the DCT coefficient; modulating the target stego sequence according to the modification direction, and modifying the target carrier image according to the modulation result corresponding to the target stego sequence to obtain a target stego image.

3. The method of claim 1, wherein, The method of obtaining a target stego sequence corresponding to a target carrier image comprises: selecting an embedding region for the target carrier image; based on the selected embedding region, obtaining a carrier sequence to be embedded with secret information; quantizing the carrier sequence according to a preset quantization step; performing shuffling processing on the quantized carrier sequence to obtain a shuffled carrier sequence; embedding the secret information into the shuffled carrier sequence to obtain a shuffled stego sequence; performing anti-shuffling processing on the shuffled stego sequence to obtain the target stego sequence.

4. The method of claim 3, wherein, The method of selecting an embedding region for the target carrier image comprises: in response to the transmission channel quality factor of the target stego image being known, selecting a frequency band region close to the upper left part of the target carrier image as the embedding region; or in response to the transmission channel quality factor of the target stego image being unknown, selecting a frequency band region close to the lower right part of the target carrier image as the embedding region.

5. The method of claim 4, wherein, The frequency band region close to the upper left part of the target carrier image is the E4 and E5 positions in the 8*8 DCT block corresponding to the target carrier image; and the frequency band region close to the lower right part of the target carrier image is the E6 and E7 positions in the 8*8 DCT block corresponding to the target carrier image.

6. The method of claim 3, wherein, The method of quantizing the carrier sequence comprises: quantizing the elements in the embedding region to 0 or 1.

7. The method of claim 1, wherein, The method further comprises: performing JPEG recompression processing on the target stego image.

8. An information steganography apparatus, characterized by comprising: The method comprises the following steps: a target stego sequence obtaining unit is configured to obtain a target stego sequence corresponding to a target carrier image; The gradient image obtaining unit is configured to obtain detection data for characterizing whether secret information is embedded in the target carrier image, and calculate a gradient image corresponding to the target carrier image based on the detection data using a back propagation algorithm. The gradient image obtaining unit comprises: inputting the target carrier image into a steganalysis analyzer that is pre-trained, and calculating a gradient image corresponding to the target carrier image based on an output result output by the steganalysis analyzer and using a back propagation algorithm, wherein the output result is used to characterize whether secret information is embedded in the target carrier image, and the back propagation algorithm is used to calculate a gradient of a loss function for weights in the steganalysis analyzer, and the gradient is used to update the weights of the steganalysis analyzer to minimize the loss function. The target carrier image obtaining unit is configured to modify the target carrier sequence based on the gradient image to obtain a target stego image.

9. An electronic device, comprising: The device comprises: a processor; and a memory configured to store a program of a data processing method, and the device is configured to execute the method according to any one of claims 1-7 after the program of the data processing method is run by the processor. The program of the data processing method is stored in a memory and is run by a processor to execute the method according to any one of claims 1-7.

10. A storage device, comprising: ​

Images