HEVC transform coefficient video steganalysis method based on attention mechanism

Abstract

The present invention discloses a HEVC transform coefficient video steganalysis method based on an attention mechanism. The method performs data preprocessing on the compressed video, extracts the prediction residual map corresponding to the I frame, and forms a training set. Then, a deep neural network including a convolutional subnetwork, a self-attention subnetwork, and a binary classifier is constructed to extract steganalysis traces and classify the video. Through training and testing, the network can accurately determine whether the video has been steganographically processed. The advantage is that the method combines deep learning technology for steganalysis, can improve the accuracy and robustness of steganalysis, can adapt to a variety of steganographic methods, and has high detection performance in low embedding rate scenarios.

Description

Technical Field

[0001] The present invention relates to a video steganalysis method, and in particular to a HEVC transform coefficient video steganalysis method based on an attention mechanism. Background Art

[0002] With the widespread use of the internet and digital media, the methods and efficiency of information transmission have undergone fundamental changes, and information security is becoming increasingly challenging. Information hiding technology, primarily focused on steganalysis and steganography, is a key research area in information security, with broad application prospects in military, commercial, and industrial sectors. Steganography is an information hiding technique used to conceal communications, embedding secret information within multimedia data without causing any perceptible changes. While steganography can improve the security of covert communications, it can also be easily exploited by criminals or hostile forces, resulting in significant security implications. To combat the harm caused by the illegal use of steganography, steganalysis has emerged. As a countermeasure against steganography, steganalysis analyzes the statistical properties of the carrier to determine whether additional information is hidden within it, and can even estimate the amount of embedded secret information and even reveal its content. Research on steganalysis is crucial for preventing the leakage of confidential information, combating terrorism and criminal activities, and maintaining internet security, and has garnered significant attention from governments and research institutions worldwide.

[0003] Because digital video data is typically large, it is typically transmitted in compressed form over public channels. H.265 / HEVC (High Efficiency Video Coding) is the latest generation of video compression standards. Its core goal is to double the compression efficiency of H.264 / AVC, playing a vital role in high-definition and ultra-high-definition video applications.

[0004] However, compared to steganalysis research in text, image, and audio, video steganalysis methods targeting HEVC transform coefficients are still in their infancy due to the complexity of H.265 / HEVC encoding. Existing research primarily leverages inter-pixel correlation theory to decode compressed video into the spatial domain. By transferring relevant methods from image steganalysis, the team extracts statistical correlation features in both the spatial and temporal domains. Due to the differences between compressed video and images, this manual feature extraction method often struggles to adapt to diverse steganographic algorithms in practical applications and is unable to effectively detect steganographic algorithms in low-embedding scenarios.

[0005] In recent years, the rapid development of deep learning technology has provided new tools for steganalysis. By building more complex and powerful models, deep learning can improve the accuracy and robustness of steganalysis, effectively addressing complex cybersecurity threats. However, there is still a lack of work on HEVC transform coefficient video steganalysis using deep learning. Therefore, it is of great practical significance to develop a method for steganalysis based on the H.265 / HEVC coding standard that combines deep learning technology. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a HEVC transform coefficient video steganalysis method based on the attention mechanism, which combines deep learning technology for steganalysis, can improve the accuracy and robustness of steganalysis, and can adapt to a variety of steganography methods, and has high detection performance in low embedding rate scenarios.

[0007] The technical solution adopted by the present invention to solve the above technical problems is: a HEVC transform coefficient video steganalysis method based on the attention mechanism, characterized by comprising the following steps:

[0008] Step 1: Obtain several compressed videos, where any compressed video is a compressed original video obtained by encoding with the H.265 / HEVC standard encoder or a compressed steganographic video obtained by encoding with the H.265 / HEVC standard encoder and then steganographically. Then, perform data preprocessing on each compressed video. The data preprocessing includes two stages: data selection and data enhancement. The data selection is to extract all I frames during the decoding process of each compressed video and obtain the prediction residual map corresponding to each I frame. The data enhancement is to randomly flip the prediction residual map corresponding to each I frame, where the size of the prediction residual map is H×W×1. Then, the value range of the flipped prediction residual map corresponding to each I frame is normalized to [0,1]. Finally, the prediction residual maps corresponding to the I frames extracted during the decoding process of all compressed videos and whose value range is within [0,1] constitute the training set.

[0009] Step 2: Build a deep neural network as the transformation coefficient video steganalysis network: The transformation coefficient video steganalysis network consists of a convolutional sub-network for extracting steganalysis traces within video frames, a self-attention sub-network for enhancing steganalysis frames, and a binary classifier for determining whether the compressed video is the compressed original video or the compressed steganalysis video;

[0010] The convolutional subnetwork adopts an architecture based on the convolutional residual network, which is divided into two stages: feature extraction and feature representation. The feature extraction stage consists of a preprocessing block and three sequentially connected shallow feature extraction blocks, and the feature representation stage consists of two downsampling blocks based on the attention mechanism, two basic feature extraction blocks and a global average pooling. During the decoding process of a compressed video, a prediction residual map corresponding to an I frame extracted in the range of [0,1] and with a size of H×W×1 is obtained after the preprocessing block to obtain a residual feature map of the centered error with a size of H×W×16. The residual feature map of the centered error is obtained after passing through three sequentially connected shallow feature extraction blocks. The fused shallow feature map is as small as H×W×16. After the fused shallow feature map passes through the first downsampling block based on the attention mechanism, the first attention feature map is obtained. After the first basic feature extraction block, the first basic feature map is obtained. The second attention feature map is obtained after the second downsampling block based on the attention mechanism. The second basic feature map is obtained after the second basic feature extraction block. The second attention feature map is obtained. The second basic feature map is obtained after the second basic feature extraction block. The second basic feature map is obtained after global average pooling. The feature vector of size 64×1;

[0011] Assume that N I frames are extracted during the decoding process of a compressed video. After the convolutional sub-network outputs N 64×1 feature vectors corresponding to a compressed video, the self-attention sub-network receives N 64×1 feature vectors corresponding to a compressed video, and the self-attention sub-network outputs a video feature vector of size N×64×1.

[0012] During the decoding process of a compressed video, the prediction residual maps corresponding to N I frames with values ​​in the range [0, 1] are extracted. After the convolution sub-network and the self-attention sub-network, the video feature vector of size N×64×1 is obtained. The result of the splicing operation on the channel dimension is input into the binary classifier, which determines whether the compressed video is the compressed original video or the compressed steganographic video.

[0013] Step 3: Use the training set to train the transformation coefficient video steganalysis network, and obtain the trained transformation coefficient video steganalysis network after the training is completed;

[0014] Step 4: For any test compressed video, all I frames are extracted during the decoding process of the test compressed video, and the prediction residual map corresponding to each I frame is obtained, wherein the test compressed video is a compressed original video obtained by encoding with the H.265 / HEVC standard encoder or a compressed steganographic video obtained by encoding with the H.265 / HEVC standard encoder and then steganographically. The prediction residual maps corresponding to all I frames extracted during the decoding process of the test compressed video are then input into the trained transform coefficient video steganalysis network, and the trained transform coefficient video steganalysis network outputs a determination result of whether the test compressed video is a compressed original video or a compressed steganographic video.

[0015] In step 1, the specific process of data selection is: using the H.265 / HEVC standard decoder to decode each compressed video, extracting all I frames during the decoding process, and saving the prediction residual value corresponding to each pixel point in each I frame to obtain the prediction residual map corresponding to each I frame; flipping is vertical flipping or horizontal flipping.

[0016] In step 2, the implementation process of the preprocessing block is as follows: the prediction residual map with a value range of [0,1] and a size of H×W×1 is extracted after passing through a learnable SRM high-pass filter kernel to obtain the residual connection of the centering error, wherein the convolution kernel size of the convolution in the learnable SRM high-pass filter kernel is 5×5; the residual connection of the centering error is sequentially subjected to a batch normalization, a truncation function TLU and a convolution kernel size of 1×1 to obtain a residual feature map of the centering error with a size of H×W×16.

[0017] In the step 2, the structures of the three shallow feature extraction blocks are the same, and each shallow feature extraction block has two feature extraction branches and a feature fusion branch; the implementation process of the first feature extraction branch is: the input feature map is convolved with a convolution kernel size of 1×1 to obtain a first shallow feature map; the second feature extraction branch uses a convolutional residual network architecture, and the implementation process of the second feature extraction branch is: the input feature map is sequentially convolved with a first convolution kernel size of 3×3, the first batch normalization, the Relu activation function, the second convolution kernel size of 3×3, and the second batch normalization to obtain an intermediate feature map, and the intermediate feature map is then added to the input feature map to obtain a second shallow feature map; the implementation process of the feature fusion branch is: The first shallow feature map and the second shallow feature map are concatenated in the channel dimension and then subjected to a convolution with a convolution kernel size of 1×1 to obtain a fused shallow feature map; among them, the size of the input feature map, the first shallow feature map, the intermediate feature map, the second shallow feature map, and the fused shallow feature map are all H×W×16, the input feature map in the first shallow feature extraction block is the residual feature map of the centering error, the input feature map in the second shallow feature extraction block is the fused shallow feature map output by the first shallow feature extraction block, the input feature map in the third shallow feature extraction block is the fused shallow feature map output by the second shallow feature extraction block, and the fused shallow feature map output by the third shallow feature extraction block is used to input the first downsampling block based on the attention mechanism.

[0018] In the step 2, the structures of the two downsampling blocks based on the attention mechanism are the same, and each downsampling block based on the attention mechanism is divided into two feature representation branches; the implementation process of the first feature representation branch is: the input feature map is sequentially subjected to a convolution with a convolution kernel size of 1×1 and a batch normalization to obtain a first feature map; the implementation process of the second feature representation branch is: the input feature map is sequentially subjected to a convolution with a convolution kernel size of 3×3, the first batch normalization, the Relu activation function, an average pooling, the second batch normalization, and a central error perception block to obtain a second feature map; the first feature map and the second feature map are subjected to The output feature map is obtained by the addition operation; in the first downsampling block based on the attention mechanism, the input feature map is the fused shallow feature map of size H×W×16 output by the third shallow feature extraction block, the sizes of the first feature map, the second feature map, and the output feature map are all H×W×32, and the output feature map is used as the first attention feature map; in the second downsampling block based on the attention mechanism, the input feature map is the first basic feature map of size H×W×32 output by the first basic feature extraction block, the sizes of the first feature map, the second feature map, and the output feature map are all H×W×64, and the output feature map is used as the second attention feature map.

[0019] The central error perception block is divided into two stages: channel attention and spatial attention;

[0020] In the channel attention stage, the input feature map is sequentially subjected to a global average pooling, a multi-layer perceptron, and a softmax function to obtain the channel weight, which is then multiplied by the input feature map to obtain the channel attention feature map. In the first downsampling block based on the attention mechanism, the size of the channel weight is 1×1×16, and the size of the channel attention feature map is H×W×16; in the second downsampling block based on the attention mechanism, the size of the channel weight is 1×1×32, and the size of the channel attention feature map is H×W×32.

[0021] In the spatial attention stage, the channel attention feature map is sequentially subjected to a convolution with a convolution kernel size of 1×1 and a softmax function to obtain a spatial attention weight map, and the spatial attention weight map is then multiplied with the channel attention feature map to obtain a weighted feature map. In the centralization error perception block in the first downsampling block based on the attention mechanism, the size of the spatial attention weight map is H×W×1, and the weighted feature map is used as the second feature map in the first downsampling block based on the attention mechanism, with a size of H×W×32; in the centralization error perception block in the second downsampling block based on the attention mechanism, the size of the spatial attention weight map is H×W×1, and the weighted feature map is used as the second feature map in the second downsampling block based on the attention mechanism, with a size of H×W×64.

[0022] In the step 2, the structures of the two basic feature extraction blocks are the same, and the implementation process of each basic feature extraction block is as follows: the input feature map is sequentially subjected to the first convolution with a convolution kernel size of 3×3, the first batch normalization, the Relu activation function, the second convolution with a convolution kernel size of 3×3, and the second batch normalization to obtain an intermediate feature map, and the intermediate feature map is then added to the input feature map to obtain an output feature map; in the first basic feature extraction block, the input feature map is the first attention feature map of size H×W×32 output by the first downsampling block based on the attention mechanism, the sizes of the intermediate feature map and the output feature map are both H×W×32, and the output feature map is used as the first basic feature map; in the second basic feature extraction block, the input feature map is the second attention feature map of size H×W×64 output by the second downsampling block based on the attention mechanism, the sizes of the intermediate feature map and the output feature map are both H×W×64, and the output feature map is used as the second basic feature map.

[0023] In step 3, cross entropy loss is used during training and the gradient is updated.

[0024] Compared with the prior art, the advantages of the present invention are:

[0025] 1) The method of the present invention takes the prediction residual map as the analysis object and proposes a transformation coefficient video steganalysis network based on the attention mechanism, which solves the problem of low detection performance of the current related transformation coefficient steganalysis.

[0026] 2) To address the problem of sparse stegographic traces within and between frames caused by adaptive steganography, the method of the present invention proposes a centralized error-aware block and a self-attention mechanism to adaptively enhance the stegographic trace area and stegographic frame features, respectively. Therefore, the method of the present invention has higher robustness and accuracy in detection scenarios with multiple steganographic methods and low embedding rates. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a flowchart of the overall implementation of the method of the present invention;

[0028] Figure 2 Schematic diagram of the composition structure of the convolutional sub-network;

[0029] Figure 3 Schematic diagram of the composition structure of the centralized error perception block. DETAILED DESCRIPTION

[0030] The present invention will be described in further detail below with reference to the accompanying drawings and embodiments.

[0031] The present invention proposes a HEVC transform coefficient video steganalysis method based on the attention mechanism, and its overall implementation flow chart is as follows: Figure 1 As shown, it includes the following steps:

[0032] Step 1: Obtain several compressed videos, where any compressed video is a compressed original video obtained by encoding with the H.265 / HEVC standard encoder or a compressed steganographic video obtained by encoding with the H.265 / HEVC standard encoder and then steganographically. Then, perform data preprocessing on each compressed video. The data preprocessing includes two stages: data selection and data enhancement. The data selection is to extract all I frames during the decoding process of each compressed video and obtain the prediction residual map corresponding to each I frame. The data enhancement is to randomly flip the prediction residual map corresponding to each I frame, where the size of the prediction residual map is H×W×1. Then, the value range of the flipped prediction residual map corresponding to each I frame is normalized to [0,1]. Finally, the prediction residual maps corresponding to the I frames extracted during the decoding process of all compressed videos with a value range of [0,1] constitute the training set.

[0033] In the present invention, the specific process of data selection is: use the H.265 / HEVC standard decoder to decode each compressed video, extract all I frames during the decoding process, and save the prediction residual value corresponding to each pixel point in each I frame to obtain the prediction residual map corresponding to each I frame; flipping is vertical flipping or horizontal flipping.

[0034] Step 2: Build a deep neural network as the transformation coefficient video steganalysis network: Figure 1 As shown in the figure, the transform coefficient video steganalysis network consists of a convolutional subnetwork (CESNet) for extracting steganalysis traces within video frames, a self-attention subnetwork (Self Attention) for enhancing steganalysis frames, and a binary classifier (Classifier) ​​for determining whether the compressed video is the compressed original video (Cover) or the compressed stego video (Stego).

[0035] like Figure 2 As shown in the figure, the convolutional subnetwork adopts an architecture based on the convolutional residual network, which is divided into two stages: feature extraction and feature representation. The feature extraction stage consists of a preprocessing block (PreprocessBlock) and three shallow feature extraction blocks (ShallowBlock) connected in sequence. The feature representation stage consists of two downsampling blocks (AttentionBlock) based on the attention mechanism, two basic feature extraction blocks (BasicBock) and a global average pooling. The prediction residual map (Prediction Residual Map) corresponding to an I frame extracted during the decoding process of a compressed video has a value range of [0,1] and a size of H×W×1. Map) after the preprocessing block obtains the residual feature map of the centering error with a size of H×W×16. The residual feature map of the centering error passes through three sequentially connected shallow feature extraction blocks to obtain a fused shallow feature map with a size of H×W×16. The fused shallow feature map passes through the first downsampling block based on the attention mechanism to obtain the first attention feature map with a size of H×W×32. The first attention feature map passes through the first basic feature extraction block to obtain the first basic feature map with a size of H×W×32. The first basic feature map passes through the second downsampling block based on the attention mechanism to obtain the second attention feature map with a size of H×W×64. The second attention feature map passes through the second basic feature extraction block to obtain the second basic feature map with a size of H×W×64. The second basic feature map is subjected to global average pooling to obtain a feature vector of size 64×1.

[0036] Further limit, such as Figure 2As shown in the figure, the preprocessing block is implemented as follows: the prediction residual map with a value range of [0, 1] and a size of H×W×1 is passed through a learnable SRM high-pass filter kernel (Conv 5×5) to extract the residual relationship of the centering error. The convolution kernel size of the learnable SRM high-pass filter kernel is 5×5. The residual relationship of the centering error is then subjected to batch normalization (Norm), a truncation function (TLU), and a convolution kernel size of 1×1 (Conv 1×1) to obtain a residual feature map of the centering error of size H×W×16. The learnable SRM high-pass filter kernel is a state-of-the-art one. It contains 30 convolution kernels of 5×5, with a total of first-order to third-order residual filter kernels. It has a rich residual feature representation, which can effectively put the network in a good initial state.

[0037] Further limit, such as Figure 2 As shown in the figure, the structures of the three shallow feature extraction blocks are the same. Each shallow feature extraction block has two feature extraction branches and a feature fusion branch. The implementation process of the first feature extraction branch is as follows: the input feature map is convolved with a convolution kernel size of 1×1 (Conv 1×1) to obtain the first shallow feature map. The second feature extraction branch uses the classic convolutional residual network architecture. The implementation process of the second feature extraction branch is as follows: the input feature map is sequentially convolved with the first convolution kernel size of 3×3 (Conv 3×3), the first batch normalization (Norm), the Relu activation function, the second convolution kernel size of 3×3 (Conv 3×3), the second batch normalization (Norm) is performed to obtain the intermediate feature map, and the intermediate feature map is then added to the input feature map to obtain the second shallow feature map; the implementation process of the feature fusion branch is as follows: the first shallow feature map and the second shallow feature map are concatenated (concat operation) in the channel dimension and then passed through a convolution with a convolution kernel size of 1×1 (Conv 1×1) to obtain the fused shallow feature map; among them, the size of the input feature map, the first shallow feature map, the intermediate feature map, the second shallow feature map, and the fused shallow feature map are all H×W×16, the input feature map in the first shallow feature extraction block is the residual feature map of the centralization error, the input feature map in the second shallow feature extraction block is the fused shallow feature map output by the first shallow feature extraction block, the input feature map in the third shallow feature extraction block is the fused shallow feature map output by the second shallow feature extraction block, and the fused shallow feature map output by the third shallow feature extraction block is used to input the first downsampling block based on the attention mechanism.

[0038] Further limit, such as Figure 2As shown in the figure, the structures of the two downsampling blocks based on the attention mechanism are the same. Each downsampling block based on the attention mechanism is divided into two feature representation branches. The implementation process of the first feature representation branch is as follows: the input feature map is sequentially subjected to a convolution with a convolution kernel size of 1×1 (Conv 1×1) and a batch normalization (Norm) to obtain the first feature map; the implementation process of the second feature representation branch is as follows: the input feature map is sequentially subjected to a convolution with a convolution kernel size of 3×3 (Conv 3×3), the first batch normalization (Norm), Relu activation function, an average pooling (AvgPool), the second batch normalization (Norm), and a centering error perception block (CEA) to obtain the second feature map; the first feature map and the second feature map are added to obtain the output feature map; wherein, in the first downsampling block based on the attention mechanism, the input feature map is the fused shallow feature map of size H×W×16 output by the third shallow feature extraction block, the sizes of the first feature map, the second feature map, and the output feature map are all H×W×32, and the output feature map is used as the first attention feature map; in the second downsampling block based on the attention mechanism, the input feature map is the first basic feature map of size H×W×32 output by the first basic feature extraction block, the sizes of the first feature map, the second feature map, and the output feature map are all H×W×64, and the output feature map is used as the second attention feature map.

[0039] like Figure 3As shown in Figure 1, the centralized error perception block is divided into two stages: channel attention and spatial attention. In the channel attention stage, the input feature map undergoes a global average pooling (GAP), a multi-layer perceptron (MLP), and a softmax function to obtain channel weights. The channel weights are then multiplied with the input feature map to strengthen the effective centralized error feature channels and obtain a channel attention feature map. Global average pooling is then performed on the input feature map channels to obtain the importance of each channel. The weight of each channel is calculated using a shared multi-layer perceptron. In the centralized error perception block in the first attention-based downsampling block, the channel weights are 1×1×16, and the channel attention feature map is H×W×16. In the centralized error perception block in the second attention-based downsampling block, the channel weights are 1×1×32, and the channel attention feature map is H×W×32. In the spatial attention stage, the channel attention feature map is sequentially subjected to a convolution with a convolution kernel size of 1×1 (Conv 1×1) and a softmax function to obtain a spatial attention weight map. The spatial attention weight map is then multiplied with the channel attention feature map to obtain a weighted feature map. Here, a convolution with a convolution kernel size of 1×1 is used to obtain the importance of each spatial position, and the softmax function is used to obtain the corresponding spatial attention weight map. In the centralization error perception block in the first downsampling block based on the attention mechanism, the size of the spatial attention weight map is H×W×1, and the weighted feature map is used as the second feature map in the first downsampling block based on the attention mechanism, with a size of H×W×32; in the centralization error perception block in the second downsampling block based on the attention mechanism, the size of the spatial attention weight map is H×W×1, and the weighted feature map is used as the second feature map in the second downsampling block based on the attention mechanism, with a size of H×W×64. Here, a centralized error perception block is designed to address the sparsity of steganographic traces within video frames. The multi-layer perceptron is an existing technology with a two-layer structure by default. Through the fusion of channel attention and spatial attention, the centralized error perception block can further extract the steganographic features of the centralized error.

[0040] It is further defined that the structures of the two basic feature extraction blocks are the same, and the implementation process of each basic feature extraction block is as follows: the input feature map is sequentially subjected to the first convolution with a convolution kernel size of 3×3 (Conv 3×3), the first batch normalization (Norm), the Relu activation function, the second convolution with a convolution kernel size of 3×3 (Conv 3×3), and the second batch normalization (Norm) to obtain an intermediate feature map, and the intermediate feature map is then added to the input feature map to obtain an output feature map; in the first basic feature extraction block, the input feature map is the first attention feature map of size H×W×32 output by the first downsampling block based on the attention mechanism, the sizes of the intermediate feature map and the output feature map are both H×W×32, and the output feature map is used as the first basic feature map; in the second basic feature extraction block, the input feature map is the second attention feature map of size H×W×64 output by the second downsampling block based on the attention mechanism, the sizes of the intermediate feature map and the output feature map are both H×W×64, and the output feature map is used as the second basic feature map.

[0041] like Figure 1 As shown in the figure, assume that N I-frames are extracted during the decoding process of a compressed video. After the convolutional subnetwork outputs N 64×1 feature vectors corresponding to the compressed video, the self-attention subnetwork receives N 64×1 feature vectors corresponding to the compressed video and outputs a video feature vector of size N×64×1. Here, N represents the total number of I-frames extracted during the decoding process of a compressed video, which is generally 4.

[0042] For the self-attention sub-network, the existing self-attention network structure is adopted. In the self-attention network structure, for the current I frame, the query vector Q and the value vector V are used to calculate the correlation weight between the 64×1 feature vector corresponding to the current I frame and the 64×1 feature vectors corresponding to the remaining I frames. The calculated correlation weight is then weighted and summed with the key vector K to obtain the video feature vector after self-attention enhancement.

[0043] like Figure 1 As shown in the figure, during the decoding process of a compressed video, the predicted residual maps corresponding to N I frames with values ​​in the range of [0,1] are extracted. After the convolution sub-network and the self-attention sub-network, the video feature vector of size N×64×1 is obtained. The result of the splicing operation on the channel dimension is input into the binary classifier, which determines whether the compressed video is a compressed original video or a compressed steganographic video.

[0044] Step 3: Use the training set to train the transformation coefficient video steganalysis network. After the training, a trained transformation coefficient video steganalysis network is obtained. The cross entropy loss is used in the training process, and the gradient is updated.

[0045] Step 4: For any test compressed video, all I frames are extracted during the decoding process of the test compressed video, and the prediction residual map corresponding to each I frame is obtained, wherein the test compressed video is a compressed original video obtained by encoding with the H.265 / HEVC standard encoder or a compressed steganographic video obtained by encoding with the H.265 / HEVC standard encoder and then steganographically. The prediction residual maps corresponding to all I frames extracted during the decoding process of the test compressed video are then input into the trained transform coefficient video steganalysis network, and the trained transform coefficient video steganalysis network outputs a determination result of whether the test compressed video is a compressed original video or a compressed steganographic video.

[0046] In order to further illustrate the feasibility and effectiveness of the method of the present invention, a simulation experiment was carried out on the method of the present invention.

[0047] The video sequences used in the simulation experiments are from the internet https: / / media.xiph.org / video / derf / . They contain 116 YUV video sequences with a resolution of 288×352 and a 4:2:0 format. Each YUV video sequence contains an average of 261 frames. The original YUV video sequences are encoded using the HEVC reference software HM 16.15, with an IPPP GOP structure and the encoding configuration file encoder_lowdelay_P_main.cfg.

[0048] In the simulation experiment, three different steganography methods were used to steganalyze the encoded video sequence to obtain compressed stego videos, and then three different steganalysis methods were used to detect the compressed stego videos. The three different steganography methods are respectively from the literature 1 (Y.Liu, S.Liu, H.Zhao, and S.Liu, “A new data hiding method for H.265 / HEVC video streams without intra-frame distortion drift,” Multimedia Tools and Applications, vol. 78, no. 6, pp. 6459-6486, 2019. (A new data hiding method for H.265 / HEVC video steganography without intra-frame distortion drift, Multimedia Tools and Applications, vol. 78, no. 6, pp. 6459-6486, 2019.)) and the stegoscopy method from the literature 2 (L.Yang, D.Xu, R.Wang, and S.He, “Adaptive HEVC video steganography based on distortion compensation optimization,” Journal of Information Security andApplications,vol.73,2023.(Adaptive HEVC video steganography method based on distortion compensation optimization, Journal of Information Security and Applications)) (abbreviated as Tar2), the steganography method from document 3 (L.Yang,R.Wang,D.Xu,L.Dong andS.He,“Centralized Error Distribution-Preserving Adaptive Steganography forHEVC,”IEEE Transactions on Multimedia,vol.26,pp.4255-4270,2024(Adaptive steganography method based on centralized error preservation, IEEE Journal of Multimedia)) (abbreviated as Tar3).The three different steganalysis methods are the steganalysis method based on the central error under the HEVC platform from document 2 (abbreviated as MCER), the spatiotemporal feature steganalysis method based on accordion unfolding (AU) and special frames extraction (SFE) from document 4 (H.Shi, T.Sun, X.Jiang, et al, "AHEVC Video Steganalysis Against DCT / DST-Based Steganography," International Journal of Digital Crime and Forensics, vol.13, no.3, pp.19-33, 2021. (A HEVC video steganalysis method for transform coefficients, International Journal of Digital Crime and Forensics)), and the method of the present invention.

[0049] In the simulation experiment, the accuracy Acc is used as the evaluation index. The higher the Acc value, the stronger the detection performance.

[0050] Table 1 shows the comparison of the accuracy rates Acc of detecting the steganographic compressed videos obtained by Tar1, Tar2 and Tar3 steganography using MCER, Shi and the method of the present invention respectively.

[0051] Table 1 Accuracy of detecting the steganographic compressed videos obtained by Tar1, Tar2 and Tar3 steganography using MCER, Shi and the method of the present invention

[0052]

[0053] As shown in Table 1, the proposed method not only effectively detects Tar1, Tar2, and Tar3 steganography methods, but also significantly outperforms other comparison methods in detecting these three steganography methods. Therefore, the proposed method offers superior performance for compressed steganographic videos and is more robust in low-embedding-rate scenarios.

Claims

1. A HEVC transform coefficient video steganalysis method based on attention mechanism, characterized by The following steps are involved: Step 1: Obtain several compressed videos, where any compressed video is a compressed original video obtained by encoding with the H.265 / HEVC standard encoder or a compressed steganographic video obtained by encoding with the H.265 / HEVC standard encoder and then steganographically encoded; then perform data preprocessing on each compressed video, which includes two stages: data selection and data enhancement. Data selection involves extracting all I frames during the decoding process of each compressed video and obtaining the prediction residual map corresponding to each I frame. Data enhancement involves randomly flipping the prediction residual map corresponding to each I frame; then normalizing the value range of the flipped prediction residual map corresponding to each I frame to [0, 1]; finally, the prediction residual maps corresponding to all I frames with a value range in [0, 1] constitute a training set; Step 2: Build a deep neural network as the transformation coefficient video steganalysis network: The transformation coefficient video steganalysis network consists of a convolutional sub-network for extracting steganalysis traces within video frames, a self-attention sub-network for enhancing steganalysis frames, and a binary classifier for determining whether the compressed video is the compressed original video or the compressed steganalysis video; The convolutional subnetwork adopts an architecture based on a convolutional residual network, which is divided into two stages: feature extraction and feature representation. The feature extraction stage consists of a preprocessing block and three sequentially connected shallow feature extraction blocks. The feature representation stage consists of two downsampling blocks based on the attention mechanism, two basic feature extraction blocks, and a global average pooling. Assume that N I frames are extracted during the decoding process of a compressed video. The convolutional sub-network outputs N feature vectors corresponding to the compressed video. The self-attention sub-network receives the N feature vectors and outputs a video feature vector. The result of the splicing operation on the channel dimension of the video feature vector is input into a binary classifier to determine whether the compressed video is a compressed original video or a compressed steganographic video; Step 3: Use the training set to train the transformation coefficient video steganalysis network to obtain a trained transformation coefficient video steganalysis network; Step 4: For any test compressed video, all I frames are extracted during the decoding process of the test compressed video, and the prediction residual map corresponding to each I frame is obtained. The map is input into the trained transform coefficient video steganalysis network, and the result of determining whether the test compressed video is a compressed original video or a compressed stego video is output.

2. The HEVC transform coefficient video steganalysis method based on the attention mechanism according to claim 1 is characterized in that In step 1, the specific process of data selection is: using the H.265 / HEVC standard decoder to decode each compressed video, extracting all I frames during the decoding process, and saving the prediction residual value corresponding to each pixel point in each I frame to obtain the prediction residual map corresponding to each I frame; flipping is vertical flipping or horizontal flipping.

3. The HEVC transform coefficient video steganalysis method based on the attention mechanism according to claim 1 is characterized in that In the step 2, a prediction residual map corresponding to an I frame extracted during the decoding process of a compressed video, whose value range is within [0,1] and whose size is H×W×1, is obtained after passing through a preprocessing block. A residual feature map of a centered error of size H×W×16 is obtained after passing through three sequentially connected shallow feature extraction blocks. A fused shallow feature map of size H×W×16 is obtained after passing through the first downsampling block based on the attention mechanism. The fused shallow feature map is obtained after passing through the first downsampling block based on the attention mechanism. A first attention feature map of size H×W×32 is obtained after passing through the first basic feature extraction block. The first basic feature map is obtained after passing through the second downsampling block based on the attention mechanism. A second attention feature map of size H×W×64 is obtained after passing through the second basic feature extraction block. The second basic feature map is obtained after passing through the global average pooling.

4. The HEVC transform coefficient video steganalysis method based on the attention mechanism according to claim 3 is characterized in that In step 2, the implementation process of the preprocessing block is as follows: the prediction residual map with a value range of [0,1] and a size of H×W×1 is extracted after passing through a learnable SRM high-pass filter kernel to obtain the residual connection of the centering error, wherein the convolution kernel size of the convolution in the learnable SRM high-pass filter kernel is 5×5; the residual connection of the centering error is sequentially subjected to a batch normalization, a truncation function TLU and a convolution kernel size of 1×1 to obtain a residual feature map of the centering error with a size of H×W×16.

5. The HEVC transform coefficient video steganalysis method based on the attention mechanism according to claim 3 is characterized in that In the step 2, the structures of the three shallow feature extraction blocks are the same, and each shallow feature extraction block has two feature extraction branches and a feature fusion branch; the implementation process of the first feature extraction branch is: the input feature map is convolved with a convolution kernel size of 1×1 to obtain a first shallow feature map; the second feature extraction branch uses a convolutional residual network architecture, and the implementation process of the second feature extraction branch is: the input feature map is sequentially convolved with a first convolution kernel size of 3×3, the first batch normalization, the Relu activation function, the second convolution kernel size of 3×3, and the second batch normalization to obtain an intermediate feature map, and the intermediate feature map is then added to the input feature map to obtain a second shallow feature map; the implementation process of the feature fusion branch is: The first shallow feature map and the second shallow feature map are concatenated in the channel dimension and then subjected to a convolution with a convolution kernel size of 1×1 to obtain a fused shallow feature map; among them, the size of the input feature map, the first shallow feature map, the intermediate feature map, the second shallow feature map, and the fused shallow feature map are all H×W×16, the input feature map in the first shallow feature extraction block is the residual feature map of the centering error, the input feature map in the second shallow feature extraction block is the fused shallow feature map output by the first shallow feature extraction block, the input feature map in the third shallow feature extraction block is the fused shallow feature map output by the second shallow feature extraction block, and the fused shallow feature map output by the third shallow feature extraction block is used to input the first downsampling block based on the attention mechanism.

6. The HEVC transform coefficient video steganalysis method based on the attention mechanism according to claim 3 is characterized in that In step 2, the two downsampling blocks based on the attention mechanism have the same structure, and each downsampling block based on the attention mechanism is divided into two feature representation branches; The implementation process of the first feature representation branch is as follows: the input feature map is sequentially subjected to a convolution with a convolution kernel size of 1×1 and a batch normalization to obtain the first feature map; the implementation process of the second feature representation branch is as follows: the input feature map is sequentially subjected to a convolution with a convolution kernel size of 3×3, the first batch normalization, the Relu activation function, an average pooling, the second batch normalization, and a central error perception block to obtain the second feature map; the first feature map and the second feature map are added to obtain the output feature map; wherein, in the first feature map based on the attention mechanism In the downsampling block, the input feature map is the fused shallow feature map of size H×W×16 output by the third shallow feature extraction block, the sizes of the first feature map, the second feature map, and the output feature map are all H×W×32, and the output feature map is used as the first attention feature map; in the second downsampling block based on the attention mechanism, the input feature map is the first basic feature map of size H×W×32 output by the first basic feature extraction block, the sizes of the first feature map, the second feature map, and the output feature map are all H×W×64, and the output feature map is used as the second attention feature map.

7. The HEVC transform coefficient video steganalysis method based on the attention mechanism according to claim 6 is characterized in that The central error perception block is divided into two stages: channel attention and spatial attention; In the channel attention stage, the input feature map is sequentially subjected to a global average pooling, a multi-layer perceptron, and a softmax function to obtain the channel weight, which is then multiplied by the input feature map to obtain the channel attention feature map. In the first downsampling block based on the attention mechanism, the size of the channel weight is 1×1×16, and the size of the channel attention feature map is H×W×16; in the second downsampling block based on the attention mechanism, the size of the channel weight is 1×1×32, and the size of the channel attention feature map is H×W×32. In the spatial attention stage, the channel attention feature map is sequentially subjected to a convolution with a convolution kernel size of 1×1 and a softmax function to obtain a spatial attention weight map, and the spatial attention weight map is then multiplied with the channel attention feature map to obtain a weighted feature map. In the centralization error perception block in the first downsampling block based on the attention mechanism, the size of the spatial attention weight map is H×W×1, and the weighted feature map is used as the second feature map in the first downsampling block based on the attention mechanism, with a size of H×W×32; in the centralization error perception block in the second downsampling block based on the attention mechanism, the size of the spatial attention weight map is H×W×1, and the weighted feature map is used as the second feature map in the second downsampling block based on the attention mechanism, with a size of H×W×64.

8. The HEVC transform coefficient video steganalysis method based on the attention mechanism according to claim 3 is characterized in that In the step 2, the structures of the two basic feature extraction blocks are the same, and the implementation process of each basic feature extraction block is as follows: the input feature map is sequentially subjected to the first convolution with a convolution kernel size of 3×3, the first batch normalization, the Relu activation function, the second convolution with a convolution kernel size of 3×3, and the second batch normalization to obtain an intermediate feature map, and the intermediate feature map is then added to the input feature map to obtain an output feature map; in the first basic feature extraction block, the input feature map is the first attention feature map of size H×W×32 output by the first downsampling block based on the attention mechanism, the sizes of the intermediate feature map and the output feature map are both H×W×32, and the output feature map is used as the first basic feature map; in the second basic feature extraction block, the input feature map is the second attention feature map of size H×W×64 output by the second downsampling block based on the attention mechanism, the sizes of the intermediate feature map and the output feature map are both H×W×64, and the output feature map is used as the second basic feature map.

9. The HEVC transform coefficient video steganalysis method based on the attention mechanism according to claim 1 is characterized in that In step 3, cross entropy loss is used during training and the gradient is updated.

Images