CN109819260B - Video steganography method and device based on multi-embedded domain fusion - Google Patents
Video steganography method and device based on multi-embedded domain fusion Download PDFInfo
- Publication number
- CN109819260B CN109819260B CN201910102368.1A CN201910102368A CN109819260B CN 109819260 B CN109819260 B CN 109819260B CN 201910102368 A CN201910102368 A CN 201910102368A CN 109819260 B CN109819260 B CN 109819260B
- Authority
- CN
- China
- Prior art keywords
- video
- frame
- secret information
- embedding
- domain
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention relates to a video steganography method and device based on multi-embedded domain fusion. The method is not limited to a single video embedding domain for embedding the secret information, but integrates three embedding domains, namely a motion vector domain, a quantization parameter domain and an intra-frame prediction domain for embedding the secret information. For the motion vector field part, performing steganographic embedding by modifying the motion vectors generated by a motion estimation module in a video compression coding framework; for the quantization parameter domain, steganographic embedding is implemented by modifying the quantization parameters of the video coding units (macroblocks); for the Intra-prediction domain, steganographic embedding is implemented by modifying Intra4x4 Intra prediction modes generated by an Intra prediction module in the video compression coding framework. The method is particularly suitable for digital multimedia security scenes with high requirements on embedding capacity and steganography security.
Description
Technical Field
The invention relates to a video steganography (video steganography) method based on multiple embedded domains under the H.264/AVC video coding standard, in particular to a video steganography method based on a motion vector domain, a quantization parameter domain and an intra-frame prediction domain, which can be used in the digital media security field and belongs to the information hiding field in the technical field of information security.
Background
Steganography is one of the important means of the current covert communication, and the method hides the confidential information in the publicable content for transmission or storage, so that an adversary cannot browse the confidential content and cannot know the existence of the confidential communication or the confidential storage fact.
Generally, when designing a steganographic algorithm, the following factors need to be considered comprehensively: 1) imperceptibility: the carriers before and after steganography are required to be indistinguishable to human senses, namely whether the carriers are subjected to steganography cannot be judged only by means of human hearing and vision; 2) robustness: the steganographic file is required to be capable of recovering the secret information therein even if the steganographic file is subjected to certain disturbance, such as noise interference in a lossy compression or transmission process; 3) embedding capacity: on the premise of ensuring imperceptibility and certain degree of robustness, secret information is required to be embedded into the carrier as much as possible; 4) embedding efficiency: the modification degree of the carrier is required to be reduced as much as possible on the premise that the quantity of the embedded information is certain; 5) safety: the existing steganalysis method can not distinguish the common carrier and the file which is steganographically processed by the steganalysis algorithm with high enough accuracy rate even if the existing steganalysis method has prior knowledge of the steganalysis algorithm.
In recent years, with the rise and wide popularity of high-interaction multimedia applications such as video on demand and streaming media live broadcast, digital video is gradually replacing images with rich visual representation and strong information carrying capacity, and becomes the most influential transmission medium in the entertainment industry today. In addition, with the rapid development of technologies such as video compression coding, computer networking, and high-performance computing, digital video can be rapidly prepared and transmitted over the internet in real time while maintaining high coding efficiency and visual fidelity. Based on the above facts, digital video is currently generally regarded as an ideal covert communication carrier, and video steganography and video steganalysis technologies have attracted extensive attention of researchers in the field of information hiding, and thus become one of the research hotspots in the field.
Generally, video steganography can be divided into two main categories, namely spatial domain video steganography and compressed domain video steganography according to different embedded domains. Most airspace video steganography algorithms draw reference to the design idea of the classical algorithm in the image steganography field, and the original pixel values of video frames are directly modified to embed secret information, and an error correction code technology and a repeated embedding mode are adopted to enhance the robustness of the secret information to video compression coding so as to reduce the error rate.
The spatial domain video steganography algorithm has the advantages that the robustness is relatively strong, and secret information in the spatial domain video steganography algorithm can be recovered even if the spatial domain video steganography algorithm is interfered to a certain degree; the disadvantage is that the kind of algorithm is relatively small and the embedding capacity is limited because the secret information generally needs to be repeatedly embedded.
The method comprises the steps that the steganography embedding operation and video compression coding are tightly coupled through video steganography in a compression domain, key modules in a video compression coding framework are disturbed, and syntax elements of a code stream generated by the key modules are modulated and modified to embed secret information. According to the steganographic embedded domain, the compressed domain video steganography can be further divided into: motion vector based steganography, transform coefficient based steganography, intra prediction mode based steganography, inter prediction mode based steganography, quantization parameter based steganography, entropy coding based steganography.
The motion vector-based video steganography takes the motion vector generated by a motion estimation module in a video compression coding framework as an information carrier, and research results on the steganography are more. The traditional motion vector domain video steganography algorithm usually adopts a simple screening rule, such as motion vector amplitude or phase angle, and selects part of motion vectors to carry out embedding modification. Different from the traditional motion vector domain video steganography algorithm, the second generation motion vector domain video steganography algorithm reduces the modification quantity of motion vectors by adopting steganography codes, thereby effectively improving the embedding efficiency and the steganography safety. The third generation motion vector field video steganography algorithm also keeps the local optimum of the modified motion vector as much as possible on the basis of adopting steganography codes, thereby further enhancing the steganography.
Transform coefficient-based video steganography implements steganography embedding by modifying luminance dct (discrete Cosine transform) coefficients, which algorithms typically have a large embedding capacity. The scholars propose that for a4x4 luminance block (referred to as Intra4x4 Intra-frame predictive coding block for short) in h.264 video which is predictive coded by adopting an Intra4x4 prediction mode, quantized DCT coefficients at certain specific positions in the block can be selected to be modified according to the Intra-frame prediction modes of adjacent blocks so as to implement steganography. Meanwhile, some scholars analyze distortion drift phenomena generated by modification of H.264 quantized DCT coefficients, and on the basis, the quantized DCT coefficients of Intra-frame predictive coding blocks in an Intra4x4 frame in an I frame are taken as carriers, and in combination with STC (Sync trellis codes), a content adaptive DCT coefficient steganography embedding scheme is provided, so that the distortion drift phenomena are effectively relieved, and good steganography safety is achieved.
Video steganography based on intra-frame prediction modes takes coding units adopting intra-frame prediction coding in compressed video as information carriers, and the steganography is implemented by modifying the intra-frame prediction modes of the coding units. For example, the steganography algorithm proposed by scholars reduces the number of intra-frame prediction modes needing to be modified through matrix coding in the embedding process, and improves the embedding efficiency and the steganography security.
The video steganography based on the inter-frame prediction mode takes a coding unit which adopts inter-frame prediction coding in a compressed video as an information carrier, and the inter-frame prediction mode is modified to implement steganography. The existing algorithm generally modulates and modifies the inter-frame prediction mode of a macro block according to bits to be embedded based on a preset mapping rule between the inter-frame prediction mode and binary bits (sequences).
It can be seen that the compressed domain video steganography algorithm has the advantages of multiple embedding selectivity and rich algorithm types. But also has the defects that the available part for modification in the embedding domain is limited, so that the embedding capacity is often not as good as that of the spatial domain video steganography algorithm. Based on the limitation of the existing video steganography technology, the method can fill the blank of the field of multi-embedded-domain fusion to a certain extent, so that the technical completeness of the field is enhanced.
Disclosure of Invention
The invention aims to design a video steganography method based on multi-embedded domain fusion according to the characteristic that embedded information of each embedded domain in an H.264/AVC compressed video does not influence each other.
Compared with other video steganography methods, the method is not limited to a single video embedding domain for embedding the secret information, but integrates three embedding domains, namely a motion vector domain, a quantization parameter domain and an intra-frame prediction domain for embedding the secret information. For the motion vector field part, performing steganographic embedding by modifying the motion vectors generated by a motion estimation module in a video compression coding framework; for the quantization parameter domain, steganographic embedding is implemented by modifying the quantization parameters of the video coding units (macroblocks); for the Intra-prediction domain, steganographic embedding is implemented by modifying Intra4x4 Intra prediction modes generated by an Intra prediction module in the video compression coding framework. Therefore, the method provided by the invention is different from any previous video steganography method, and is particularly suitable for digital multimedia security scenes with higher requirements on embedding capacity and steganography security.
According to research, most of the current video steganography methods fail to effectively balance the embedding capacity and the steganography security, which is shown in the following steps: in order to increase the embedding capacity and sacrifice the steganography security, when the steganography modification is carried out on specific syntactic elements (such as motion vectors, interframe prediction modes and the like) of a compressed video code stream, the steganography security is negatively influenced when the load rate is overlarge; or to reduce the embedding capacity in order to guarantee steganographic security. In addition, most video steganography methods have difficulty in effectively maintaining the properties and encoding parameters of the original carrier video. Aiming at the limitations, under the condition of comprehensively considering factors such as load rate, steganography security, effective maintenance of carrier video attributes and the like, how to design and optimize a high-efficiency video steganography algorithm is one of the problems which are urgently needed to be solved in the field of digital multimedia security at present.
Specifically, the technical scheme adopted by the invention is as follows:
a video steganography method based on multi-embedded domain fusion, as shown in fig. 1, includes the following steps:
1) acquiring a video frame needing steganography embedding;
2) according to the selected steganography strategy, selecting a specified compressed encoding element to be embedded in the frame, and acquiring a binary carrier vector according to a mapping rule;
3) for each appointed compression coding element to be embedded in the video frame, calculating a model according to a preset cost to obtain the steganography embedding cost;
4) establishing a hidden channel by applying a steganographic code, embedding secret information into the hidden channel, and modulating and modifying corresponding compressed encoding elements in the video frame according to a result;
5) repeatedly executing the steps 2) to 4), and encoding the video frame according to the result, namely completing the embedding of the corresponding secret information of the frame;
6) and restarting the execution of the rest each frame from the step 1) according to the frame coding sequence of the video file until all the secret information is embedded, thereby obtaining the final steganographic video file.
Further, the method further includes a secret information extraction process, as shown in fig. 2, including the following steps:
1) acquiring a video frame needing to be subjected to secret information extraction;
2) and determining the specified compression encoding elements to be embedded in the frame according to the selected steganographic strategy. Extracting secret information bits in each appointed embedded coding element according to a rule agreed by both communication parties in advance;
3) and repeating the steps 1) to 2) for the rest of each frame according to the frame decoding sequence of the video file until all the secret information is extracted.
On the basis of the scheme, the invention further improves that when the invention is adopted to carry out steganographic embedding of the secret information, each single embedded domain is modified as follows for a single video frame.
[1] Video steganography algorithm based on motion vector field
Motion vector domain based video steganography algorithms implement steganography embedding by modifying motion vectors generated by a motion estimation module in a video compression coding framework. In particular, when the algorithm is applied to motion vectors MV0,0＝(mvx,mvy) (where 0,0 denotes the current macroblock, mvx,mvyRepresenting motion vectors in the horizontal and vertical directions), the following steps are performed in order to select the most suitable locally optimal motion vector from all candidate motion vectors as the final modification result:
a) and (4) preprocessing. Respectively to MV0,0Four motion vectors MV in the neighborhood range-1,0＝(mvx-1,mvy)、MV1,0＝(mvx+1,mvy)、MV0,-1＝(mvx,mvy-1) and MV0,1＝(mvx,mvy+1) performing local optimality determination;
b) and constructing an optimal candidate set. Mixing MV-1,0，MV1,0，MV0,-1And MV0,1Forming a candidate motion vector set psi by the local optimal motion vector in the motion vector set, if psi is not null, skipping to step e), otherwise, continuing to execute step c);
c) a search area is delineated. Defining a search area for a motion vector for which any candidate motion vector MV' ═ (MV)x′,mvy') must satisfy LSB (mv)x+mvy)≠LSB(mvx′+mvy') to a host; wherein, LSB represents the least significant bit embedding algorithm;
d) and constructing an optimal candidate set. Respectively carrying out local optimality judgment on each candidate motion vector in the search area, determining all locally optimal motion vectors in the candidate motion vectors, and combining the locally optimal motion vectors into a set Ψ;
e) and selecting a local optimal motion vector. For each locally optimal motion vector in Ψ, its rate-distortion cost is calculated, and the locally optimal motion vector with the smallest rate-distortion cost is selected as the final modification result.
[2] Video steganography algorithm based on quantization parameters
Quantization parameter based video steganography algorithms implement steganographic embedding by modifying the quantization parameters of video coding units (macroblocks). Specifically, when steganography embedding is performed on quantization parameters of a macro block in a certain video frame, the algorithm sequentially executes the following steps:
a) and determining a carrier vector. Coding the current frame to obtain the quantization parameter of the current macro block, judging whether the quantization parameter is 0, if so, jumping to the next macro block, otherwise, continuously executing the step b);
b) and (4) preprocessing. And judging whether the last bit of the quantization parameter of the current macro block is matched with the current bit of the secret information to be embedded. If matching, keeping the quantization parameter unchanged, otherwise, executing the step c);
c) embedding is performed. And (4) performing an operation of adding and subtracting one to the quantization parameter of the current macroblock, and simultaneously ensuring that the value of the quantization parameter is always less than 51.
d) And subsequently coding the video frame. And continuing to execute the subsequent encoding step of the video frame, and outputting the obtained steganographic video frame to a code stream.
[3] Video steganography algorithm based on intra-frame prediction mode
Video steganography algorithms based on Intra-prediction modes implement steganography embedding by modifying Intra4x4 Intra-prediction modes generated by Intra-prediction modules in the video compression coding framework. In the embedding process, by selecting a suboptimal Intra-frame prediction mode of Intra4x4, the method reduces the negative influence of Intra-frame prediction mode disturbance on the video compression coding efficiency as much as possible.
The algorithm divides 9 Intra4x4 Intra prediction modes Mode _ i (i ═ 0,1, …,8) in the h.264 video coding standard into two groups a and B, where the Intra prediction modes in the group a represent bit 0 and the Intra prediction modes in the group B represent bit 1.
A＝{Mode_0,Mode_2,Mode_4,Mode_6,Mode_8}
B＝{Mode_1,Mode_3,Mode_5,Mode_7}
The embedded flow of the algorithm is as shown. Specifically, when the Intra4x4 Intra prediction Mode _ i is subjected to steganographic modification, the algorithm sequentially performs the following steps to select the most suitable suboptimal Intra4x4 Intra prediction Mode from all candidate Intra4x4 Intra prediction modes as the final modification result.
a) Candidate intra prediction mode groupings are determined. According to the H.264/AVC video coding standard, determining the index number of the Mode _ i, and determining a candidate Intra4X4 Intra prediction Mode set X according to the lowest bit of the index number: if the lowest bit of the Mode _ i index number is 0, X is B; if the lowest bit of the Mode _ i index number is 1, X is A;
b) and determining the best candidate intra-prediction mode. For any candidate Intra4X4 Intra prediction mode IMODE in X, its corresponding Lagrangian cost is calculated
J(IMODE)＝D+λR
Wherein D represents the visual distortion introduced by encoding the corresponding 4 × 4 block using IMODE; r represents the bit number required by encoding IMODE and the bit number required by the corresponding 4 multiplied by 4 block luminance transformation coefficient; λ is a lagrangian parameter used to control the balance between D and R, which is determined by the encoder. Determining an Intra-prediction mode of Intra4X4 with the minimum Lagrangian cost in X as the best modification result;
c) and (5) subsequent encoding. According to the result of the step b), adopting the best candidate Intra4x4 Intra prediction mode to carry out Intra prediction coding on the corresponding 4x4 block, and outputting the subsequent coding result to the code stream.
Based on the above description, the present invention provides a video steganography method based on multi-embedded domain fusion, wherein the embedding process of secret information includes the following steps (if there is no special description, the following steps are executed by a computer):
1) acquiring a video frame needing steganography embedding;
2) embedding the secret information bit by bit from the first macro block of the current video frame and from the quantization parameter domain, the motion vector domain and the intra-frame prediction domain;
3) and if the current video frame type is a P frame and the macroblock type is 8x8 blocks, the embedding algorithm selects a video steganography algorithm based on quantization parameters for embedding. Determining the quantization parameter of the current macro block, and if the quantization parameter of the current macro block is not 0, embedding secret information according to a preset steganography algorithm;
4) if the current video frame type is a P frame and the macroblock type is 16x16 blocks, 16x8 blocks or 8x16 blocks, the embedding algorithm selects a motion vector-based video steganography algorithm for embedding. Determining a motion vector MV of a current macroblock0,0＝(mvx,mvy) Embedding secret information according to a preset steganography algorithm;
5) if the current video frame type is an I frame and the macro block type is Intra4x4, the embedding algorithm selects a video steganography algorithm based on Intra-frame prediction for embedding, and the steganography embedding is implemented by modifying an Intra-frame prediction mode generated by an Intra-frame prediction module in a video compression coding frame;
6) repeating steps 3) to 5) for the rest of the macroblocks of the frame until all the macroblocks are traversed;
7) and (4) repeatedly executing the steps 2) to 6) on the rest frames according to the frame coding sequence of the video file until all the secret information is embedded, and obtaining the final steganographic video file.
Further, the multi-embedded domain based secret information extraction process comprises the following steps (unless otherwise specified, the following steps are all executed by a computer):
1) acquiring a video frame needing to be subjected to secret information extraction;
2) starting from a first macro block of a current video frame, carrying out bit-by-bit extraction on the information according to the sequence of a quantization parameter domain, a motion vector domain and an intra-frame prediction domain;
3) if the current video frame type is a P frame and the macro block type is 8x8 blocks, secret information is embedded in a quantization parameter domain, and secret information is extracted according to a preset steganography algorithm;
4) if the current video frame type is a P frame and the macroblock type is 16x16 blocks, 16x8 blocks or 8x16 blocks, embedding secret information in a motion vector field, and extracting the secret information according to a preset steganographic algorithm;
5) if the current video frame type is an I frame and the macro block type is Intra4x4, embedding secret information in an Intra-frame prediction domain, and extracting the secret information according to a preset steganography algorithm;
6) repeating steps 3) to 5) for the rest of the macroblocks of the frame until all the macroblocks are traversed;
7) and repeating the steps 2) to 6) for the rest of each frame according to the frame coding sequence of the video file until all the secret information is extracted.
Correspondingly to the above method, the present invention further provides a video steganography apparatus based on multi-embedded domain fusion, which includes:
the video frame acquisition module is responsible for acquiring a video frame needing steganography embedding;
the multi-embedded domain fusion embedding module is responsible for embedding the secret information bit by bit from the first macro block of the current video frame according to the quantization parameter domain, the motion vector domain and the intra-frame prediction domain until all macro blocks of the video frame are traversed; and then embedding the secret information into each remaining video frame according to the frame coding sequence of the video file until the secret information is completely embedded, thereby obtaining the final steganographic video file.
Furthermore, the device also comprises a secret information extraction module which is responsible for extracting the secret information from the steganographic video; the secret information extraction module acquires a video frame needing secret information extraction; starting from a first macro block of a current video frame, bit-by-bit extracting the secret information according to a quantization parameter domain, a motion vector domain and an intra-frame prediction domain until all macro blocks of the video frame are traversed; then according to the frame coding sequence of the video file, the same processing is carried out on each of the rest video frames until the secret information is completely extracted.
The invention also provides a computer device comprising a memory and a processor, the memory storing a computer program configured to be executed by the processor, the computer program comprising instructions for carrying out the steps of the method described above.
The video steganography method of the invention has the following beneficial effects in the relevant technical field:
1) the embedding capacity is large. Most of the existing video steganography algorithms only perform secret steganography on a single compression domain, and because the available part of a compression embedding domain is limited, the embedding capacity is inferior to that of a spatial domain video steganography algorithm. The compressed domain has the other characteristics of more selectivity of the embedded domain, rich algorithm types and larger space in the aspect of improving the steganography load rate. Aiming at the limitation, the invention increases the embeddable capacity and further improves the algorithm load performance by simultaneously embedding secret information in a plurality of embedded domains.
2) The safety is high. The existing video steganography algorithm almost adopts a targeted steganography analysis method, and the safety is low in practical application. Meanwhile, when the steganography modification is carried out on specific syntax elements (such as motion vectors, inter-frame prediction modes and the like) of the compressed video code stream, the steganography safety is negatively influenced when the load rate is overlarge. Aiming at the limitation, the secret information is dispersedly embedded into different compressed embedded domains, so that the load rate of each embedded domain is ensured to be under a certain low threshold value, a higher total load rate is met, and certain steganography safety is ensured.
3) And maintaining the properties and coding parameters of the original carrier video before and after steganography. The existing video steganography algorithm does not usually consider the basic attribute parameters (such as frame rate, code rate and resolution) of the original carrier video, so that the visual and auditory quality of the original carrier video can be damaged to a certain extent, and phenomena such as picture distortion or sound and picture asynchronization are caused. Aiming at the limitation, before steganography embedding, the basic attribute parameters of the original carrier video are quickly identified or pre-estimated, and the compression coding parameters of the steganography video are reasonably set according to the detection result, so that the steganography video can effectively keep the visual and auditory quality and the audio parameters of the original carrier video and maintain the sound and picture synchronization, and the steganography safety is improved to a certain extent.
4) No intermediate file is generated. When the existing video steganography algorithm is used for steganography embedding, an intermediate file (such as a YUV file) is generally generated, the space overhead of software operation is greatly increased, the operation efficiency is reduced, and the application range of the software is greatly restricted. Aiming at the limitations, the invention improves the embedding process and framework of the existing video steganography algorithm, and controls the generation, use, release and other operations of the intermediate file in the memory when the steganography is carried out, thereby effectively avoiding the generation of the intermediate file, lightening the IO load, greatly reducing the software operation space overhead and improving the software operation efficiency and the application range.
5) High resolution video is adapted. The carrier video resolution applicable to the existing video steganography algorithm is usually CIF and the resolution is lower. At present, with the continuous updating and upgrading of shooting equipment and the continuous improvement of processor performance, the resolution code rate of a mainstream video is higher and higher, and the phenomena of longer processing time, even breakdown and the like often occur when the existing video steganography algorithm carries out steganography on a video with higher resolution. In view of the above limitations, the present invention embeds the secret information while encoding and decoding the video, and can process steganography embedding of higher resolution video such as 1080P.
Drawings
Fig. 1 is a general flow diagram of the secret information embedding of the present invention.
Fig. 2 is a general flow diagram of the secret information extraction of the present invention.
Fig. 3 is a schematic diagram of the steganography process based on motion vectors in the present invention.
Fig. 4 is a schematic diagram of a steganography process based on quantization parameters in the present invention.
Fig. 5 is a schematic diagram of the intra-prediction based steganography process in the present invention.
Fig. 6 is a flow chart of secret information embedding based on multi-embedding domain fusion of the present invention.
Fig. 7 is a flow chart of secret information extraction based on multi-embedded domain fusion according to the present invention.
Detailed Description
The invention will now be further described by way of example with reference to the accompanying figures 6 to 7.
Before the invention is adopted to carry out secret information steganography embedding, the data to be embedded can be encrypted to obtain random binary data stream. The secret information embedding process of the video steganography algorithm based on multi-embedding domain fusion is shown in fig. 6, and the specific operation details are as follows:
1) acquiring a video frame needing steganography embedding;
2) from the first macro block of the current video frame, embedding the secret information bit by bit according to the sequence of a quantization parameter domain, a motion vector domain and an intra-frame prediction domain;
3) and if the current video frame type is a P frame and the macroblock type is 8x8 blocks, the embedding algorithm selects a video steganography algorithm based on quantization parameters for embedding. And determining the quantization parameter of the current macro block, and judging whether the last bit of the quantization parameter of the current macro block is matched with the current bit of the secret information to be embedded if the quantization parameter of the current macro block is not 0. And if the quantization parameters are matched with the macro blocks, keeping the quantization parameters unchanged, otherwise, performing an operation of adding or subtracting one to the quantization parameters of the current macro blocks, and simultaneously ensuring that the numerical values of the quantization parameters are always smaller than 51. The steganographic process based on quantization parameters is shown in fig. 4.
4) If the current video frame type is a P frame and the macroblock type is 16x16 blocks, 16x8 blocks or 8x16 blocks, the embedding algorithm selects a motion vector-based video steganography algorithm for embedding. Determining a motion vector MV of a current macroblock0,0＝(mvx,mvy) Simultaneously, a search area is defined, local optimality determination is performed for each candidate motion vector in the search area, all locally optimal motion vectors are determined, and the local optimality determination is combined into a set Ψ. For each locally optimal motion vector in Ψ and for calculating its rate-distortion cost, the locally optimal motion vector with the smallest rate-distortion cost is selected as the final modification result. The motion vector based steganography process is illustrated in fig. 3.
5) If the current video frame type is an I frame and the macro block type is Intra4x4, the video steganography algorithm based on Intra prediction is selected by the embedding algorithm for embedding. Firstly, determining a candidate set of Intra-prediction modes of Intra4x4 according to the current bit of the secret information to be embedded, and calculating the Lagrangian cost of all modes in the candidate set, wherein the Intra-prediction mode of Intra4x4 with the minimum Lagrangian cost is used as the best modification result. The process of steganography based on intra prediction is illustrated in fig. 5.
6) Repeating steps 3) to 5) for the rest of the macroblocks of the frame until all the macroblocks are traversed;
7) and (4) repeatedly executing the steps 2) to 6) on the rest frames according to the frame coding sequence of the video file, knowing that all the secret information is embedded, and obtaining the final steganographic video file.
The process of extracting the secret information based on the video steganography algorithm with multi-embedded domain fusion provided by the invention is shown in fig. 7, and the specific operation details are as follows:
1) acquiring a video frame needing secret information extraction;
2) from the beginning of the first macro block of the current video frame, carrying out bit-by-bit extraction on the secret information according to the sequence of a quantization parameter domain, a motion vector domain and an intra-frame prediction domain;
3) and if the current video frame type is a P frame and the macroblock type is 8x8 blocks, selecting a video steganography algorithm based on quantization parameters for extraction. Determining the quantization parameter of the current macro block, and if the quantization parameter of the current macro block is not 0, extracting the last bit of the quantization parameter of the current macro block as an extracted secret information bit;
4) and if the current video frame type is a P frame and the macroblock type is 16x16 blocks, 16x8 blocks or 8x16 blocks, selecting a motion vector-based video steganography algorithm for extraction. Determining a motion vector MV of a current macroblock0,0＝(mvx,mvy) The least significant bit, LSB (mv), of the sum of its horizontal and vertical components is obtainedx+mvy) And using the extracted secret information bit as the extracted secret information bit;
5) and if the current video frame type is an I frame and the macroblock type is an Intra4x4 Intra prediction mode, selecting a video steganography algorithm based on Intra prediction for extraction. According to the Intra4x4 Intra prediction mode as the extracted secret information bits;
6) repeating steps 3) to 5) for the rest of the macroblocks of the frame until all the macroblocks are traversed;
7) and repeating the steps 2) to 6) for the rest of each frame according to the frame coding sequence of the video file, and knowing that all the secret information is extracted.
In order to highlight the invention, the invention provides a high-concealment video steganography method, which is to embed the secret information of H.264/AVC video with the MP4 format and the resolution of 1080p, and then carry out steganography analysis detection by an ADB _ S method. Specifically, a video sample set is prepared and subjected to steganalysis experiments by adopting the following configuration:
selecting 40 MP4 format videos with the resolution of 1080p, embedding the information, and then performing steganalysis detection by an ADB _ S method, wherein 29 pairs of training and 11 pairs of detection are performed, and the accuracy is counted, and the result is shown in Table 1.
Table 1 detection and analysis of video steganography method based on multi-embedded domain fusion by ADB _ S method
From table 1, it can be seen that the detection accuracy still does not exceed 60% for the ADB _ S method when the average load rate exceeds 5%.
Another embodiment of the present invention provides a video steganography apparatus based on multi-embedded domain fusion, which includes: the video frame acquisition module is responsible for acquiring a video frame needing steganography embedding; the multi-embedded domain fusion embedding module is responsible for embedding the secret information bit by bit from the first macro block of the current video frame according to the quantization parameter domain, the motion vector domain and the intra-frame prediction domain until all macro blocks of the video frame are traversed; and then embedding the secret information into each remaining video frame according to the frame coding sequence of the video file until the secret information is completely embedded, thereby obtaining the final steganographic video file.
The device also comprises a secret information extraction module which is responsible for extracting the secret information from the steganographic video; the secret information extraction module acquires a video frame needing secret information extraction; starting from a first macro block of a current video frame, bit-by-bit extracting the secret information according to a quantization parameter domain, a motion vector domain and an intra-frame prediction domain until all macro blocks of the video frame are traversed; then according to the frame coding sequence of the video file, the same processing is carried out on each of the rest video frames until the secret information is completely extracted.
Another embodiment of the invention provides a computer device comprising a memory and a processor, the memory storing a computer program configured to be executed by the processor, the computer program comprising instructions for carrying out the steps of the method described above.
The above embodiments are only intended to illustrate the technical solution of the present invention and not to limit the same, and a person skilled in the art can modify the technical solution of the present invention or substitute the same without departing from the spirit and scope of the present invention, and the scope of the present invention should be determined by the claims.
Claims (8)
1. A video steganography method based on multi-embedded domain fusion is characterized by comprising the following steps:
1) acquiring a video frame needing steganography embedding;
2) embedding the secret information bit by bit from the first macro block of the current video frame according to a quantization parameter domain, a motion vector domain and an intra-frame prediction domain until all macro blocks of the video frame are traversed;
3) repeating the step 2) for each remaining video frame according to the frame coding sequence of the video file until the secret information is completely embedded to obtain a final steganographic video file;
wherein, step 2) includes:
2.1) if the current video frame type is a P frame and the macroblock type is 8x8 blocks, the embedding algorithm selects a video steganography algorithm based on quantization parameters, and the video steganography algorithm is embedded by modifying the quantization parameters of a video coding unit;
2.2) if the current video frame type is a P frame and the macroblock type is 16x16 blocks, 16x8 blocks or 8x16 blocks, the embedding algorithm selects a video steganography algorithm based on a motion vector field, and the motion vector generated by a motion estimation module in a video compression coding frame is modified for embedding;
2.3) if the current video frame type is an I frame and the macro block type is Intra4x4, the embedding algorithm selects a video steganography algorithm based on Intra prediction, and the video is embedded by modifying an Intra prediction mode of Intra4x4 generated by an Intra prediction module in a video compression coding frame.
2. The method according to claim 1, wherein the quantization parameter based video steganography algorithm comprises the steps of:
a) coding the current frame to obtain the quantization parameter of the current macro block, judging whether the quantization parameter is 0, if so, jumping to the next macro block, otherwise, continuously executing the step b);
b) judging whether the last bit of the quantization parameter of the current macro block is matched with the current bit of the secret information to be embedded, if so, keeping the quantization parameter unchanged, otherwise, executing the step c);
c) embedding, namely performing an operation of adding and subtracting one to the quantization parameter of the current macroblock, and simultaneously ensuring that the numerical value of the quantization parameter is always less than 51;
d) and continuing to execute the subsequent encoding step of the video frame, and outputting the obtained steganographic video frame to a code stream.
3. The method of claim 1, wherein the motion vector field based video steganography algorithm is applied to motion vectors MV0,0＝(mvx,mvy) When modification is carried out, the following steps are sequentially executed to select the most suitable local optimal motion vector from all candidate motion vectors as a final modification result, so that the motion vector MV is enabled0,0Is equal to the current bit of the secret information to be embedded:
a) respectively to MV0,0Four motion vectors MV in the neighborhood range-1,0＝(mvx-1,mvy)、MV1,0＝(mvx+1,mvy)、MV0,-1＝(mvx,mvy-1) and MV0,1＝(mvx,mvy+1) performing local optimality determination;
b) mixing MV-1,0，MV1,0，MV0,-1And MV0,1Forming a candidate motion vector set psi by the local optimal motion vector in the motion vector set, if psi is not null, skipping to step e), otherwise, continuing to execute step c);
c) defining a search area for a motion vector for which any candidate motion vector MV' ═ (MV)x′,mvy') must satisfy LSB (mv)x+mvy)≠LSB(mvx′+mvy′)；
d) Respectively carrying out local optimality judgment on each candidate motion vector in the search area, determining all locally optimal motion vectors in the candidate motion vectors, and combining the locally optimal motion vectors into a set Ψ;
e) for each locally optimal motion vector in Ψ, its rate-distortion cost is calculated, and the locally optimal motion vector with the smallest rate-distortion cost is selected as the final modification result.
4. The method of claim 1, wherein the video steganography algorithm based on Intra-prediction Mode sequentially performs the following steps to select the most suitable next-best Intra-prediction Mode of Intra4x4 from all candidate Intra-prediction modes of Intra4x4 as the final modification result when performing the steganography modification on the Intra4x4 Intra-prediction Mode _ i, wherein the selected Intra4x4 Intra-prediction Mode is equal to the current bit of the secret information to be embedded:
a) according to the H.264/AVC video coding standard, determining the index number of the Mode _ i, and determining a candidate Intra4X4 Intra prediction Mode set X according to the lowest bit of the index number: dividing Intra4X4 Intra prediction modes Mode _ i into two groups a and B, wherein the Intra prediction modes in the group a represent bit 0, the Intra prediction modes in the group B represent bit 1, and if the lowest bit of the Mode _ i index number is 0, X is equal to B; if the lowest bit of the ModOri index number is 1, X is A;
b) for any candidate Intra4X4 Intra prediction mode IMODE in X, its corresponding lagrangian cost is computed:
J(IMODE)＝D+λR
wherein D represents the visual distortion introduced by encoding the corresponding 4 × 4 block using IMODE; r represents the bit number required by encoding IMODE and the bit number required by the corresponding 4 multiplied by 4 block luminance transformation coefficient; λ is a lagrange parameter used to control the balance between D and R, which is determined by the encoder; determining an Intra4X4 Intra prediction mode with the minimum Lagrangian cost in X, and taking the Intra prediction mode as the best candidate Intra prediction mode;
c) according to the result of the step b), adopting the best candidate Intra4x4 Intra prediction mode to carry out Intra prediction coding on the corresponding 4x4 block, and outputting the subsequent coding result to the code stream.
5. The method according to claim 1, characterized in that the extraction of secret information is performed by the following steps:
[1] acquiring a video frame needing to be subjected to secret information extraction;
[2] starting from a first macro block of a current video frame, bit-by-bit extracting the secret information according to a quantization parameter domain, a motion vector domain and an intra-frame prediction domain until all macro blocks of the video frame are traversed;
[3] repeating the step [2] for each remaining video frame according to the frame coding sequence of the video file until all secret information is extracted;
the step [2] comprises the following steps:
[2.1] if the current video frame type is a P frame and the macro block type is 8x8 blocks, secret information is embedded in a quantization parameter domain, and secret information is extracted according to a preset steganography algorithm;
[2.2] if the current video frame type is a P frame and the macroblock type is 16x16 blocks, 16x8 blocks or 8x16 blocks, embedding secret information in a motion vector field, and extracting the secret information according to a preset steganography algorithm;
[2.3] if the current video frame type is an I frame and the macro block type is Intra4x4, secret information is embedded in an Intra-frame prediction domain, and the secret information is extracted according to a preset steganography algorithm.
6. A video steganography apparatus based on multi-embedded domain fusion, comprising:
the video frame acquisition module is responsible for acquiring a video frame needing steganography embedding;
the multi-embedded domain fusion embedding module is responsible for embedding the secret information bit by bit from the first macro block of the current video frame according to the quantization parameter domain, the motion vector domain and the intra-frame prediction domain until all macro blocks of the video frame are traversed; then embedding secret information into each remaining video frame according to the frame coding sequence of the video file until the secret information is completely embedded, and obtaining a final steganographic video file;
the bit-by-bit embedding of the secret information according to the quantization parameter domain, the motion vector domain and the intra-frame prediction domain comprises the following steps:
if the current video frame type is a P frame and the macro block type is 8x8 blocks, the embedding algorithm selects a video steganography algorithm based on quantization parameters, and the video steganography algorithm is embedded by modifying the quantization parameters of a video coding unit;
if the current video frame type is a P frame and the macroblock type is 16x16 blocks, 16x8 blocks or 8x16 blocks, the embedding algorithm selects a motion vector-based video steganography algorithm, and the motion vector generated by a motion estimation module in a video compression coding frame is modified for embedding;
if the current video frame type is an I frame and the macro block type is Intra4x4, the embedding algorithm selects a video steganography algorithm based on Intra prediction, and the video is embedded by modifying an Intra prediction mode of Intra4x4 generated by an Intra prediction module in a video compression coding frame.
7. The apparatus according to claim 6, further comprising a secret information extraction module responsible for extracting secret information from the steganographic video; the secret information extraction module acquires a video frame needing secret information extraction; starting from a first macro block of a current video frame, bit-by-bit extracting the secret information according to a quantization parameter domain, a motion vector domain and an intra-frame prediction domain until all macro blocks of the video frame are traversed; then, according to the frame coding sequence of the video file, the same processing is carried out on each of the rest video frames until the secret information is completely extracted;
the method for extracting the secret information bit by bit from the first macro block of the current video frame according to the quantization parameter domain, the motion vector domain and the intra-frame prediction domain until all macro blocks of the video frame are traversed comprises the following steps:
if the current video frame type is a P frame and the macro block type is 8x8 blocks, secret information is embedded in a quantization parameter domain, and secret information is extracted according to a preset steganography algorithm;
if the current video frame type is a P frame and the macroblock type is 16x16 blocks, 16x8 blocks or 8x16 blocks, embedding secret information in a motion vector field, and extracting the secret information according to a preset steganographic algorithm;
if the current video frame type is an I frame and the macro block type is Intra4x4, secret information is embedded in the Intra-frame prediction domain, and the secret information is extracted according to a preset steganography algorithm.
8. A computer device comprising a memory and a processor, the memory storing a computer program configured to be executed by the processor, the computer program comprising instructions for carrying out the steps of the method according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910102368.1A CN109819260B (en) | 2019-02-01 | 2019-02-01 | Video steganography method and device based on multi-embedded domain fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910102368.1A CN109819260B (en) | 2019-02-01 | 2019-02-01 | Video steganography method and device based on multi-embedded domain fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109819260A CN109819260A (en) | 2019-05-28 |
| CN109819260B true CN109819260B (en) | 2020-11-06 |
Family
ID=66606303
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910102368.1A Expired - Fee Related CN109819260B (en) | 2019-02-01 | 2019-02-01 | Video steganography method and device based on multi-embedded domain fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109819260B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112118450B (en) * | 2019-06-21 | 2022-03-29 | 杭州海康威视数字技术股份有限公司 | Method and device for decoding and encoding prediction mode |
| CN111263157A (en) * | 2020-02-27 | 2020-06-09 | 武汉大学 | Video multi-domain steganalysis method based on motion vector consistency |
| CN112738355B (en) * | 2020-12-24 | 2021-10-26 | 华南理工大学 | Non-degraded HEVC video steganography method capable of resisting deep learning network detection |
| CN113840146B (en) * | 2021-09-30 | 2022-10-21 | 中南大学 | Video information hiding method and system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7965838B2 (en) * | 2007-11-30 | 2011-06-21 | Faraday Technology Corp. | Watermark generating circuit and related method thereof |
| CN102223540A (en) * | 2011-07-01 | 2011-10-19 | 宁波大学 | Information hiding method facing to H.264/AVC (automatic volume control) video |
| CN102905134A (en) * | 2012-10-22 | 2013-01-30 | 山东省计算中心 | Adaptive video digital steganography method |
| CN103338376A (en) * | 2013-07-02 | 2013-10-02 | 中国科学技术大学 | Video steganography method based on motion vector |
| CN103826128A (en) * | 2014-02-25 | 2014-05-28 | 宁波工程学院 | Encryption domain H.264/AVC video reversible data hiding method |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW200832276A (en) * | 2007-01-31 | 2008-08-01 | Benq Corp | Method and related apparatus for hiding data inside video signals and transmitting the video signals to a display device |
| CN102547297B (en) * | 2012-02-28 | 2014-12-03 | 中国传媒大学 | MPEG2 (Moving Picture Experts Group 2) video watermarking realization method based on DC (Discrete Cosine) coefficient |
| CN104581176B (en) * | 2014-12-26 | 2017-07-07 | 宁波大学 | The insertion of H.264/AVC compression domain robust video watermark and extracting method without frame in error drift |
-
2019
- 2019-02-01 CN CN201910102368.1A patent/CN109819260B/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7965838B2 (en) * | 2007-11-30 | 2011-06-21 | Faraday Technology Corp. | Watermark generating circuit and related method thereof |
| CN102223540A (en) * | 2011-07-01 | 2011-10-19 | 宁波大学 | Information hiding method facing to H.264/AVC (automatic volume control) video |
| CN102905134A (en) * | 2012-10-22 | 2013-01-30 | 山东省计算中心 | Adaptive video digital steganography method |
| CN103338376A (en) * | 2013-07-02 | 2013-10-02 | 中国科学技术大学 | Video steganography method based on motion vector |
| CN103826128A (en) * | 2014-02-25 | 2014-05-28 | 宁波工程学院 | Encryption domain H.264/AVC video reversible data hiding method |
Non-Patent Citations (2)
| Title |
|---|
| Prediction mode modulated data-hiding algorithm for H.264/AVC;Dawen Xu; Rangding Wang; Jicheng Wang;《Journal of Real-Time Image Processing》;20100819;第205-214页 * |
| 视频隐写分析技术研究综述;张弘，尤玮珂，赵险峰;《信息安全学报》;20181130;第3卷(第6期);第13-27页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109819260A (en) | 2019-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109819260B (en) | Video steganography method and device based on multi-embedded domain fusion | |
| Yang et al. | An information hiding algorithm based on intra-prediction modes and matrix coding for H. 264/AVC video stream | |
| Yang et al. | An efficient information hiding method based on motion vector space encoding for HEVC | |
| CN101715135B (en) | Template matching based method for encoding self-adaption intra-frame prediction filtering | |
| CN108965887B (en) | Video information hiding method and device based on inter-block decoupling | |
| CN109391846B (en) | Video scrambling method and device for self-adaptive mode selection | |
| WO2010137324A1 (en) | Image encoding device, image decoding device, image encoding method, and image decoding method | |
| CN106101714B (en) | A kind of and tightly coupled H.264 video information hiding method of compression encoding process | |
| JP2013511874A (en) | Method for coding a block of an image and method for reconstructing a block of an image | |
| US9210435B2 (en) | Video encoding method and apparatus for estimating a code amount based on bit string length and symbol occurrence frequency | |
| CN1695381A (en) | Sharpness enhancement in post-processing of digital video signals using coding information and local spatial features | |
| CN104581176B (en) | The insertion of H.264/AVC compression domain robust video watermark and extracting method without frame in error drift | |
| CN104160703A (en) | Object detection informed encoding | |
| Wang et al. | An Information Hiding Algorithm for HEVC Based on Angle Differences of Intra Prediction Mode. | |
| CN114786019A (en) | Image prediction method, encoder, decoder, and storage medium | |
| Wang et al. | A large-capacity information hiding method for HEVC video | |
| CN104902281A (en) | Hamming code plus one-based information hiding method of HEVC video | |
| CN113329229A (en) | High-capacity hiding method for H.265 video information with high-efficiency fidelity | |
| JP5938424B2 (en) | Method for reconstructing and encoding image blocks | |
| CN109361926B (en) | Lossless reversible information hiding method for H.264/AVC video visual quality | |
| Sakazawa et al. | H. 264 native video watermarking method | |
| CN112714230A (en) | Robust video steganography method and device based on audio side channel | |
| Liu et al. | Adaptive video data hiding with low bit-rate growth based on texture selection and ternary syndrome-trellis coding | |
| Jiaji et al. | An information hiding algorithm for HEVC based on intra prediction mode and block code | |
| Li et al. | An information hiding approach based on integer transform coefficient and virtual space encoding for H. 264/AVC |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20201106 Termination date: 20210201 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |