CN105915916B - Video steganalysis method based on the estimation of motion vector distortion performance - Google Patents
Video steganalysis method based on the estimation of motion vector distortion performance Download PDFInfo
- Publication number
- CN105915916B CN105915916B CN201610313236.XA CN201610313236A CN105915916B CN 105915916 B CN105915916 B CN 105915916B CN 201610313236 A CN201610313236 A CN 201610313236A CN 105915916 B CN105915916 B CN 105915916B
- Authority
- CN
- China
- Prior art keywords
- video
- motion vector
- steganalysis
- sub
- distortion performance
- 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
- 239000013598 vector Substances 0.000 title claims abstract description 142
- 238000000034 method Methods 0.000 title claims abstract description 76
- 238000000605 extraction Methods 0.000 claims abstract description 21
- 238000004458 analytical method Methods 0.000 claims description 21
- 238000004364 calculation method Methods 0.000 claims description 7
- 238000005259 measurement Methods 0.000 claims description 3
- 238000007906 compression Methods 0.000 description 12
- 238000013461 design Methods 0.000 description 10
- 230000006835 compression Effects 0.000 description 9
- 230000006870 function Effects 0.000 description 9
- 238000005457 optimization Methods 0.000 description 9
- 238000001514 detection method Methods 0.000 description 6
- 238000011161 development Methods 0.000 description 5
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005192 partition Methods 0.000 description 2
- 238000013139 quantization Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
- 208000011580 syndromic disease Diseases 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
- H04N19/51—Motion estimation or motion compensation
- H04N19/567—Motion estimation based on rate distortion criteria
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/46—Embedding additional information in the video signal during the compression process
- H04N19/467—Embedding additional information in the video signal during the compression process characterised by the embedded information being invisible, e.g. watermarking
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The present invention relates to a kind of video steganalysis methods based on the estimation of motion vector distortion performance.Several frame groups will be divided into measured compressed video first, each frame group is made of continuous video frame, and any video frame belongs to and only belongs to some frame group.Then steganalysis feature extraction is carried out for the frame group that some includes several motion vectors, it include: for each motion vector in some frame group, obtain the set being made of the motion vector and its nearby motion vectors, the distortion performance of each motion vector in the set is calculated, preset steganalysis feature then is extracted to the frame group；Above-mentioned steps are repeated, steganalysis feature extraction successively is carried out to all frame groups of video to be measured.Then the classifier based on steganalysis feature is used, steganography classification judgement is carried out to each frame group in video to be measured.The present invention can effectively detect current existing motion vector field video steganography method.
Description
Technical Field
The invention relates to a Video steganography analysis (Video steganography) method, in particular to a Video steganography analysis method based on motion vector Rate-Distortion (Rate-Distortion) performance estimation and application thereof in the aspect of digital media security protection, and belongs to the information hiding sub-field in the technical field of information security.
Background
Modern information hiding techniques mainly include Steganography, and Digital Watermarking. Steganography mainly researches how to embed secret information into digital multimedia files such as digital images, videos, audios and the like so as to achieve the aim of covert communication; the steganography analysis mainly adopts methods of machine learning, pattern recognition and the like to carry out steganography classification judgment on the file to be tested.
With the widespread popularity of video on demand, internet streaming media, and handheld portable camera devices, digital video has become the most influential distribution medium in today's entertainment industry. In addition, with the rapid development of technologies such as video compression, computer networking, and high-performance computing, the original video material can be rapidly compressed and transmitted over the internet in real time while ensuring higher encoding efficiency and visual fidelity. Therefore, digital video is an ideal carrier for steganographic communication.
Video steganography can be generally divided into Spatial Domain video steganography and compressed Domain video steganography: the former embeds secret information by directly modifying the original pixel values of the video frame; the latter introduces steganographic disturbance in the video compression coding process, so that the compression coding and steganographic embedding can be carried out simultaneously. The compressed domain video steganography may be classified into several categories, according to the difference of an embedded domain, of video steganography based on a Motion Vector (Motion Vector), video steganography based on a Transform Coefficient (Transform Coefficient), video steganography based on an Intra Prediction Mode (Intra Prediction Mode), video steganography based on an Inter Prediction Mode, video steganography based on a Quantization (Quantization) process, and video steganography based on Entropy Coding (Entropy Coding). The steganographic embedding of the secret information by taking the motion vector as a carrier has the following three advantages: firstly, Motion vectors are generated in a Motion Estimation (Motion Estimation) module of the video, representing the offset between an original Block and a corresponding Prediction Reference Block, and the disturbance introduced by slight modification of the Motion vectors can be automatically processed by subsequent modules such as Motion compensation, transform coding and entropy coding, and only have a tiny influence on the visual quality of the reconstructed video; secondly, the motion vector in the video usually has a larger value range, so that a motion vector domain video steganography method capable of keeping the statistical property of the motion vector can be designed; in addition, a large number of motion vectors are usually contained in the video, which ensures that the steganography of the video based on the motion vectors can have a large steganography embedding capacity.
For the above reasons, motion vector domain video steganography has long been receiving wide attention from scholars in the field and has undergone the following three stages of development: the first generation motion vector domain video steganography method mainly selects a part of motion vectors for steganography embedding of secret information through a preset simple screening rule (reference documents: F. Jordan, M. Kutter, andT. Ebrahimi, "Proposal of a marketing technique for designing/reproducing data in compressed and decompressed video," ISO/IEC Document, JTC1/SC29/WG11, Stockholm, Sweden, Tech. Rep. M2281, Jul.1997.). The second generation motion vector field Video steganography method applies steganography Codes such as STC (Syndrome Trellis Codes) and WPC (Wet Paper Codes) to improve embedding efficiency and steganography security (references: y.cao, x.zhao, d.feng, and r.shell, "Video steganography with structured motion estimation," in proc.1. 20153. high. conf.inf.in.14, LNCS, vol.6958, Prague, Czech reproduction, May2011, pp.193-207.; y.yao, w.zhang, n.yu, and x.zhao, "defying embedding vector removal for motion-based Video decoder," mu.42, 11163, and 11186. 11163). However, since these methods inevitably destroy the local optimality of motion vectors during steganography embedding, they cannot resist against the attack of AoSO (reference: k.wang, h.zhao, and h.wang, "Video perceptual information vector-based analysis by adding or subtracting one motion vector value," IEEE trans.inf.forces Security, vol.9, No.5, pp.741-751, May 2014.) which is the most effective current motion vector field. The third generation Motion vector field Video steganography method in the latest development stage can keep local optimization of modified Motion vectors during embedding of secret information as much as possible, thus greatly improving the anti-steganography performance compared with the previous two generations (references: Y.Cao, H.Zhang, X.ZHao, and H.Yu, "Video steganography based on optimized Motion estimation, in Proc.3rd ACM Workshop Inf.HidingMuting Multimedia Security, Portland, OR, USA, Jun.2015, pp.25-31.; H.Zhang, Y.Cao, and X.ZHao," Motion vector-based Video annotation with compressed localization, "Multimedia," Applications, and Applications).
Through patent inquiry, the related patent application situations existing in the field of the invention are as follows:
(1) chinese patent No. 2015103093163, "an improved video steganalysis method based on motion vector reply," discloses a motion vector domain video steganalysis method. The method designs the steganography analysis characteristic based on the fact that the motion vector in the steganography video after being recompressed has a reply phenomenon, and obtains the key parameters of video compression coding by reconstructing the first compression process, thereby effectively improving the practicability and the steganography analysis accuracy. Since the present invention does not involve the re-compression operation of the video, the method is obviously different from the design idea and the specific implementation manner of the present invention.
(2) Chinese patent No. 2015102222805, "a content adaptive video steganalysis method", discloses a motion vector field video steganalysis method. The method divides a video to be detected into detection intervals with variable lengths by calculating frame dynamics so as to perform content self-adaptive feature extraction and steganalysis. The method provides a content-adaptive steganalysis feature extraction strategy, and does not relate to the specific steganalysis feature design, so the basic purpose, design thought and specific implementation mode of the method and the invention are obviously different.
(3) The chinese patent with the patent application number of 2013100660098, "video steganalysis method based on local cost non-optimal statistics" discloses a motion vector domain video steganalysis method. The method judges whether the video is subjected to steganography by calculating the variable quantity of the local optimal probability of the motion vector before and after the recompression of the video to be detected. Since the present invention does not involve the re-compression operation of the video, the method is obviously different from the design idea and the specific implementation manner of the present invention.
Disclosure of Invention
The invention aims to accurately estimate the rate distortion performance of the motion vector in the compressed video so as to accurately judge whether the motion vector in the compressed video is locally optimal in the rate distortion sense, and finally designs the motion vector domain video steganalysis method with high steganalysis performance on the basis.
Compared with other motion vector domain video steganalysis methods, the method adopts the Lagrange cost function to estimate the rate distortion performance of the motion vector, defines two different types of local optimal motion vectors in the rate distortion sense, and finally designs and obtains 36-dimensional motion vector domain steganalysis characteristics. Therefore, the method provided by the invention is different from any previous video steganalysis method, and is particularly suitable for digital multimedia security protection scenes with high security level requirements.
According to research, the current motion vector domain video steganalysis method has the following two limitations: firstly, the design principle of most analysis methods does not fully utilize the intrinsic weakness of motion vector domain video steganography, that is, the embedded modification of motion vectors inevitably destroys the local optima, so that the original local optimal motion vectors are modified into non-local optima. Therefore, when the steganographic load rate is low, it is difficult for these methods to ensure an ideal analysis detection effect. Secondly, although some analysis methods are correct in nature, namely, the steganalysis characteristic design is carried out based on the detection of the local optimization of the motion vector, the analysis methods only judge whether the motion vector in the compressed video is the local optimization according to the Distortion (Distortion), and neglect the key role played by the code rate estimation in the judgment process. These methods are unable to combat the motion vector domain video steganography methods currently in the latest development stage, since they fail to correctly detect the local optimum of the motion vector in the rate-distortion sense.
Specifically, the technical scheme adopted by the invention is as follows:
a video steganalysis method based on motion vector rate distortion performance estimation comprises the following steps:
1) frame group division: dividing a compressed video to be detected into a plurality of frame groups, wherein each frame group consists of continuous video frames, and any video frame belongs to and only belongs to a certain frame group;
2) for a certain group of frames F containing N motion vectorsgExecuting steps 3) to 5) to perform steganalysis feature extraction;
3) pretreatment: for frame group FgEach motion vector V in (1)i＝(xi,yi) Wherein i is 1, 2.. times.n, obtained from ViAnd its neighboring motion vectors form a set omega (V)i)；
4) Estimation of motion vector rate distortion performance: according to a preset motion vector rate distortion performance estimation method, a set omega (V) is calculatedi) The rate distortion performance of each motion vector, where i 1, 2.
5) Calculating and extracting characteristics: according to the calculation result of the step 4), the frame group F is setgExtracting preset steganalysis characteristics;
6) repeating the steps 2) to 5), and sequentially performing steganalysis feature extraction on all frame groups of the video to be detected;
7) steganalysis: and respectively carrying out steganography classification judgment on each frame group in the video to be tested by adopting a classifier based on preset steganography analysis characteristics.
On the basis of the scheme, the invention further improves that when the steganalysis method is adopted for carrying out steganalysis (figure 1), each motion vector V in the compressed video to be testedi＝(xi,yi) Obtained from V as followsiAnd its neighboring motion vectors form a set omega (V)i)：
Due to the simplicity and effectiveness of Lagrangian optimization techniques (Lagrangian optimization techniques), Lagrangian optimization techniques have been widely applied in most video compression coding standards including H.264/AVC and H.265/HEVC to efficiently optimize the rate-distortion performance of video coding.
In the video compression coding process, Motion Estimation (Motion Estimation) for a certain Block (Block) means (fig. 2): defining a search range in an encoded Reference Frame (Reference Frame), comparing an original block to be encoded with all or part of candidate blocks in the search range, and selecting a best matching block as a best prediction Reference block of the original block to be encoded. In particular, given that the Lagrangian multiplier λ is used to control the balance between Distortion (discrimination) and code Rate (Rate), the Rate-Distortion-optimized motion estimation can be performed on the partitions S by minimizing the Lagrangian Cost Function, i.e.:
wherein, S'mRepresenting the prediction reference block, D (S, S '), corresponding to the partition S, to which the motion vector m points'm) Represents S and S'mThe distortion between them is generally measured in Sum of Absolute Difference (SAD) of block pixels, i.e. the Sum of Absolute Differences (SAD)Wherein k represents the position index of the pixel, and A represents the position index set of all pixels in the block; r (m, ref _ ind) represents the number of bits required to encode the motion vector m and the corresponding reference frame index ref _ ind; omega represents a position ofMotion estimation in a reference frame searches for regions. In addition to SAD, SATD (sum of Absolute transformed differences) is also a common block matching metric in motion estimation, i.e.Where T stands for Hadamard Transform (Hadamard Transform), α for the normalization factor.
Based on the lagrangian cost function and the introduction of motion estimation in video compression coding, the definition of local optimal motion vectors and 36-dimensional high-performance steganalysis characteristics in the motion vector rate-distortion performance estimation method provided by the invention in the rate-distortion sense are explained in detail as follows.
[1] The motion vector rate distortion performance estimation method comprises the following steps:
according to the motion estimation based on rate distortion optimization in video coding, the method adopts the Lagrange cost function to estimate the rate distortion performance of the motion vector in the compressed video.
Given a motion vector V, its rate-distortion performance can be estimated according to the following equation:
JD(V)＝D(Srec,SV)+λR(V).
wherein S isrecRepresents a Reconstructed Block (Reconstructed Block), S, having a motion vector VVRepresenting the motion vector V pointed to corresponding to SrecPredicted reference block of, D (S)rec,SV) Denotes SrecAnd SVλ is the lagrange multiplier, r (V) represents the number of bits required to encode the motion vector V.
[2] Definition of local optimal motion vectors in the rate-distortion sense:
based on the rate-distortion performance estimation of the motion vector, the invention defines the local optimal motion vector in the rate-distortion sense in two compressed videos.
Definition 1: i type local optimal motion vector. Given a motion vector V in compressed video, if it satisfies
V is called type I locally optimal motion vector.
Definition 2: type II locally optimal motion vectors. Given a motion vector V in compressed video, if it satisfies
V is called type II locally optimal motion vector.
In definitions 1 and 2, Ω (V) denotes a set consisting of V and its neighboring motion vectors,representing a reconstructed block with V, SmCorresponding to the orientation of the motion vector mλ is the lagrange multiplier, r (m) represents the number of bits required to encode the motion vector m.
[3] The 36-dimensional high-performance motion vector domain steganalysis feature:
given a certain group F of frames containing N motion vectors in a compressed videogAccording to the following stepsgAnd 4 types of sub-features are extracted and finally combined to obtain 36-dimensional steganalysis features.
a) Pretreatment: for FgEach motion vector V in (1)i＝(xi,yi)(i∈[1,N]) First obtaining ViAnd its neighboring motion vector, i.e. omega (V)i)＝{xi-1,xi,xi+1}×{yi-1,yi,yi+1 }; and then to set omega (V)i) Each motion vector ofWherein j ∈ [1,9 ]](FIG. 3), calculating the rate distortion performanceAndfinally determine the set { JSAD(m)|m∈Ω(Vi) Minimum value of element in (1), note asSimilarly, obtain
b) Type 1 sub-feature extraction: each dimension of a type 1 sub-feature corresponds to a given kAndequal probability, is defined as
Wherein the functionThe same applies below.
c) Type 2 sub-feature extraction: type 2 sub-feature and JSAD(Vi) Andrelative error relationship therebetween, is defined as
d) Type 3 sub-feature extraction: each dimension of a type 3 sub-feature corresponds to a given kAndequal probability, is defined as
e) Type 4 sub-feature extraction: type 4 sub-feature and JSATD(Vi) Andrelative error relationship therebetween, is defined as
f) And (4) final feature combination: the final 36-dimensional steganalysis feature is obtained by combining the above 4 types of sub-features and is defined as
Based on the above description, the video steganalysis method for estimating performance based on motion vector rate distortion by adopting the lagrangian cost function provided by the invention comprises the following steps (if no special description exists, the following steps are executed by a computer):
1) frame group division: the method comprises the steps of dividing a compressed video to be detected into a plurality of frame groups, wherein each frame group consists of continuous video frames, and any video frame belongs to and only belongs to a certain frame group.
2) For a certain group of frames F containing N motion vectorsgAnd executing steps 3) to 5) to perform steganalysis feature extraction.
3) Pretreatment: for frame group FgEach motion vector V in (1)i＝(xi,yi) Wherein i is 1, 2.. times.n, obtained from ViAnd its neighboring motion vectors form a set omega (V)i)＝{xi-1,xi,xi+1}×{yi-1,yi,yi+1}；
4) Estimation of motion vector rate distortion performance: for the set Ω (V)i) Each motion vector ofWherein j ∈ [1,9 ]](FIG. 3), using SAD and SATD as block matching measurement standard, respectively, and calculating by Lagrange cost function to obtain rate distortion performanceAnd
5) calculating and extracting characteristics: according to the calculation result of the step 4), the frame group F is setgAnd extracting preset 4 types of sub-features, and finally combining to obtain 36-dimensional steganalysis features.
6) And (5) repeating the steps 2) to 5), and sequentially carrying out steganalysis feature extraction on all frame groups of the video to be detected.
7) Steganalysis: and respectively carrying out steganography classification judgment on each frame group in the video to be tested by adopting a classifier based on preset steganography analysis characteristics.
The video steganalysis method of the invention has the following beneficial effects in the relevant technical field:
1) the existing motion vector field video steganography method can be effectively detected. According to rate-distortion optimization in video compression coding, any motion vector is locally optimal in a rate-distortion sense, and thus any modification to a motion vector will inevitably destroy its local optimization in rate-distortion performance. The invention adopts the Lagrange cost function to accurately estimate the rate distortion performance of the motion vector in the compressed video and designs the steganalysis characteristic on the basis, thereby fully utilizing the intrinsic weakness of the steganalysis of the motion vector domain video and grasping the key core for implementing the steganalysis of the motion vector domain video, therefore, the invention has ideal detection effect on the current steganalysis method of the motion vector domain video.
2) The carrier source mismatch phenomenon in steganalysis can be effectively relieved to a certain extent. The carrier Source Mismatch (Cover Source Mismatch) phenomenon in steganalysis refers to: when a steganalysis detector trained on one carrier source is used to analyze samples from different carrier sources, the difference between the two carrier sources will have a large negative impact on steganalysis accuracy. The carrier source mismatch phenomenon is ubiquitous in real network environments, and is the biggest obstacle to practical application of steganalysis. The invention fully utilizes the essential weakness of motion vector field video steganography, and accurately identifies local optimal motion vectors and effectively improves the steganography analysis performance and the stability thereof by accurately estimating the rate distortion performance of the motion vectors in the compressed video, so that the invention can effectively relieve the phenomenon of carrier source mismatch to a certain extent, such as: the steganography analysis detector obtained by training on the basis of extracting steganography analysis characteristics from samples with the same size, the same code rate and the same embedding rate by adopting the method has better analysis and detection effects on steganography videos with different sizes, different code rates and different embedding rates.
3) The method is widely applicable to different video coding standards. The partial motion vector field video steganalysis method is designed under the framework of a specific video coding standard and depends on the unique characteristics of the standard, so that the application range of the analysis methods is greatly influenced. The method and the device mainly carry out accurate estimation on the rate distortion performance of the motion vector to identify the local optimal motion vector so as to effectively improve the steganalysis performance, and the realization of the method and the device does not depend on a specific video coding standard, so the method and the device have a larger application range and can effectively analyze the steganalysis of the motion vector domain video based on different video coding standards.
4) The method has expandability. When extracting steganalysis characteristics from a frame group, for each motion vector V (x, y), a set omega (V) consisting of V and adjacent motion vectors thereof (x-1, x, x + 1), x (y-1, y, y + 1) is obtained. Therefore, different video steganalysis methods based on motion vector rate distortion performance estimation can be expanded and customized by modifying the calculation mode of omega (V) so as to be applied to different steganalysis scenes.
Drawings
FIG. 1 is a schematic illustration of video steganalysis employing the present invention;
FIG. 2 is a schematic diagram of block-based motion estimation in video compression coding;
FIG. 3 shows a certain motion vector ViCorresponding set omega (V)i) A schematic diagram of the spatial relative positions of all the elements in the list;
fig. 4 is a flow chart of video steganalysis employing the present invention.
Detailed Description
The invention will be further described with reference to the following specific embodiment in conjunction with fig. 4.
The video steganalysis method for estimating performance based on motion vector rate distortion by adopting the Lagrange cost function provided by the invention has the following specific operation details:
1) frame group division: the method comprises the steps of dividing a compressed video to be detected into a plurality of frame groups, wherein each frame group consists of continuous video frames, and any video frame belongs to and only belongs to a certain frame group.
2) For a certain group of frames F containing N motion vectorsgAnd executing steps 3) to 5) to perform steganalysis feature extraction.
3) Pretreatment: for frame group FgEach motion vector V in (1)i＝(xi,yi) Wherein i is 1, 2.. times.n, obtained from ViAnd its neighboring motion vectors form a set omega (V)i)＝{xi-1,xi,xi+1}×{yi-1,yi,yi+1}, i.e.
4) Estimation of motion vector rate distortion performance: for the set Ω (V)i) Each motion vector ofWherein j ∈ [1,9 ]](FIG. 3), using SAD and SATD as block matching measurement standard, respectively, and calculating by Lagrange cost function to obtain rate distortion performanceAndnamely, it is
Wherein，Representing possession of ViThe reconstructed block of (a) is reconstructed,representing motion vectorsDirected to correspond toλ is the lagrange multiplier,representing coded motion vectorsThe number of bits required. Subsequently, a set { J } is determinedSAD(m)|m∈Ω(Vi) Minimum value of element in (1), note asSimilarly, obtain
5) Calculating and extracting characteristics: according to the calculation result of the step 4), adopting the described characteristic extraction process to frame the group FgAnd 4 preset types of sub-features are extracted so as to combine to obtain 36-dimensional steganalysis features.
6) And (5) repeating the steps 2) to 5), and sequentially carrying out steganalysis feature extraction on all frame groups of the video to be detected.
7) Steganalysis: and performing steganography classification judgment on each frame group in the video to be detected by adopting a classifier based on the 36-dimensional steganography analysis characteristics.
As can be seen from the above detailed description: firstly, the method mainly identifies the local optimal motion vector in the compressed video and carries out steganalysis by accurately estimating the rate distortion performance of the motion vector, and the realization of the method does not depend on a specific video coding standard; secondly, according to different practical application scenes, different video steganalysis characteristics based on motion vector rate distortion performance estimation can be generated by changing the calculation mode of omega (·). Therefore, the invention has wider application range and stronger flexibility.
In order to highlight that the invention provides a high-performance motion vector domain video steganalysis method, the steganalysis experiment is carried out by adopting the following experimental configuration:
1) YUV sequence: 250 YUV420 sequences with CIF (352X 288) resolution and length between 90-376 frames are collected through the Internet.
2) Video encoder and its configuration: and preparing compressed video samples by adopting an x264 open source video encoder, and setting the encoding grade as a basic grade Baseline Profile in order to reduce time overhead.
3) Compressing video parameters: setting the bit rate (bitrate) to be 0.5mb/s, 3mb/s or 10mb/s, the frame rate to be 50fps, and the motion estimation fast search algorithm to be DIA, HEX or UMH.
4) Steganographic embedding rate: the hidden embedding rate is expressed by a per-frame Motion Vector modification rate (CMVR), the CMVR of the carrier video is set to be 0, and the CMVR of the hidden video is set to be 0.1 or 0.2.
5) The steganography method comprises the following steps: the high concealment motion vector field Video steganography method proposed by Cao et al was chosen for analysis (references: Y. Cao, H. Zhang, X. ZHao, and H.Yu, "Video stereoscopic based on optimal temporal prediction," in Proc.3rd ACM Workshop inf.
6) The steganalysis method comprises the following steps: as AoSO is the most effective steganalysis method of the current motion vector field, the method is compared with the method of the invention (the reference documents are K.Wang, H.ZHao, and H.Wang, "Video hierarchical analysis acquisition information vector by adding or subtracting one motion vector value," IEEE trans.Inf.Forensics Security, vol.9, No.5, pp.741-751, May 2014.).
7) Training and detecting: in each group of steganalysis experiments, 60% of carrier-steganalysis sample pairs are randomly selected for training a Support Vector Machine (SVM), the remaining 40% of sample pairs are used for steganalysis classification judgment, each group of steganalysis experiments are repeated for 50 times, and the obtained data are averaged.
According to the experimental configuration, the obtained steganalysis results are shown in table 1, and it can be seen that the method can effectively detect the motion vector field video steganalysis which is in the latest development stage and has the highest steganalysis security at present.
TABLE 1 average detection accuracy (%) -for steganalysis using AoSO and the method of the invention proposed by Cao et al
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 (4)
1. A video steganalysis method based on motion vector rate distortion performance estimation is characterized by comprising the following steps:
1) dividing a compressed video to be detected into a plurality of frame groups, wherein each frame group consists of continuous video frames, and any video frame belongs to and only belongs to a certain frame group;
2) for a certain frame group containing a plurality of motion vectors, executing steps 3) to 5) to perform steganalysis feature extraction;
3) for a certain group of frames FgEach motion vector V in (1)i＝(xi,yi) Where i 1, 2.. times.n, is obtained from the motion vector ViAnd its neighboring motion vectors form a set omega (V)i)＝{xi-1,xi,xi+1}×{yi-1,yi,yi+1}；
4) According to a preset motion vector rate distortion performance estimation method, the rate distortion performance of each motion vector in the set is calculated, namely the rate distortion performance of each motion vector in the set is calculated for the set omega (V)i) Each motion vector ofWherein j ∈ [1,9 ]]Respectively taking SAD and SATD as block matching measurement standards, and calculating by Lagrange cost function to obtain the rate distortion performanceAnd
5) extracting preset steganalysis characteristics from the frame group according to the calculation result of the step 4);
6) repeating the steps 2) to 5), and sequentially performing steganalysis feature extraction on all frame groups of the video to be detected;
7) and performing steganography classification judgment on each frame group in the video to be detected by adopting a classifier based on steganography analysis characteristics.
2. The method of claim 1, wherein the rate distortion performance in step 4)Andthe calculation formula of (2) is as follows:
wherein,representing possession of ViThe reconstructed block of (a) is reconstructed,representing motion vectorsDirected to correspond toλ is the lagrange multiplier,representing coded motion vectorsThe number of bits required.
3. The method of claim 2, wherein step 5) is performed for a group of frames FgAnd extracting preset 4 types of sub-features, and finally combining to obtain 36-dimensional steganalysis features.
4. The method of claim 3, wherein step 5) comprises:
a) pretreatment: rate distortion performance obtained according to step 4)Anddetermine the set { JSAD(m)|m∈Ω(Vi) Minimum value of element in (1), note asSimilarly, obtain
b) Type 1 sub-feature extraction: each dimension of a type 1 sub-feature corresponds to a given kAndequal probability, defined as:
wherein the functionThe same applies below;
c) type 2 sub-feature extraction: type 2 sub-feature and JSAD(Vi) Andrelative error relationship between, defined as:
d) type 3 sub-feature extraction: each dimension of a type 3 sub-feature corresponds to a given kAndthe probability of being equal is determined,is defined as:
e) type 4 sub-feature extraction: type 4 sub-feature and JSATD(Vi) Andrelative error relationship between, defined as:
f) and (4) final feature combination: the final 36-dimensional steganalysis feature is obtained by combining the above 4 types of sub-features, defined as:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610313236.XA CN105915916B (en) | 2016-05-12 | 2016-05-12 | Video steganalysis method based on the estimation of motion vector distortion performance |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610313236.XA CN105915916B (en) | 2016-05-12 | 2016-05-12 | Video steganalysis method based on the estimation of motion vector distortion performance |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105915916A CN105915916A (en) | 2016-08-31 |
| CN105915916B true CN105915916B (en) | 2019-02-22 |
Family
ID=56748123
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610313236.XA Expired - Fee Related CN105915916B (en) | 2016-05-12 | 2016-05-12 | Video steganalysis method based on the estimation of motion vector distortion performance |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105915916B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107613303B (en) * | 2017-09-08 | 2019-10-22 | 中国科学院信息工程研究所 | Video steganalysis method, device, equipment and computer readable storage medium |
| CN107682703B (en) * | 2017-10-27 | 2019-11-26 | 中国科学院信息工程研究所 | Video steganalysis method, device, equipment and computer readable storage medium based on the detection of inter-frame forecast mode recovery characteristic |
| CN110324634B (en) * | 2019-07-05 | 2021-10-01 | 中国科学技术大学 | Video steganography method based on motion vector embedding distortion decomposition |
| CN112312141B (en) * | 2020-08-17 | 2021-08-13 | 中国科学技术大学 | HEVC video steganography method based on pixel adaptive compensation |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103108188A (en) * | 2013-03-01 | 2013-05-15 | 武汉大学 | Video steganalysis method based on partial cost non-optimal statistics |
| CN104837011A (en) * | 2015-05-04 | 2015-08-12 | 中国科学院信息工程研究所 | Content self-adaptive video steganalysis method |
| CN104853186A (en) * | 2015-06-08 | 2015-08-19 | 中国科学院信息工程研究所 | Improved video steganalysis method based on motion vector reply |
-
2016
- 2016-05-12 CN CN201610313236.XA patent/CN105915916B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103108188A (en) * | 2013-03-01 | 2013-05-15 | 武汉大学 | Video steganalysis method based on partial cost non-optimal statistics |
| CN104837011A (en) * | 2015-05-04 | 2015-08-12 | 中国科学院信息工程研究所 | Content self-adaptive video steganalysis method |
| CN104853186A (en) * | 2015-06-08 | 2015-08-19 | 中国科学院信息工程研究所 | Improved video steganalysis method based on motion vector reply |
Non-Patent Citations (1)
| Title |
|---|
| 一种基于运动向量局部最优保持的视频隐写方法;张弘等;《第十二届全国信息隐藏暨多媒体信息安全学术大会论文集》;20150331;第1-4章，图1-5 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105915916A (en) | 2016-08-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Motion vector-based video steganography with preserved local optimality | |
| Cao et al. | Video steganalysis exploiting motion vector reversion-based features | |
| Wang et al. | Video steganalysis against motion vector-based steganography by adding or subtracting one motion vector value | |
| Su et al. | A video steganalytic algorithm against motion-vector-based steganography | |
| CN105915916B (en) | Video steganalysis method based on the estimation of motion vector distortion performance | |
| Milani et al. | Multiple compression detection for video sequences | |
| Wang et al. | Real-time compressed-domain video watermarking resistance to geometric distortions | |
| CN105933711B (en) | Neighborhood optimum probability video steganalysis method and system based on segmentation | |
| CN104837011B (en) | Content self-adaptive video steganalysis method | |
| Yang et al. | A clustering-based framework for improving the performance of JPEG quantization step estimation | |
| Tian et al. | A semi-fragile video watermarking algorithm based on chromatic residual DCT | |
| CN109819260B (en) | Video steganography method and device based on multi-embedded domain fusion | |
| Kancherla et al. | Novel blind video forgery detection using markov models on motion residue | |
| CN105979269B (en) | Motion vector field video steganography method based on novel insertion cost | |
| Zhao et al. | A novel video watermarking scheme in compression domain based on fast motion estimation | |
| CN104853186A (en) | Improved video steganalysis method based on motion vector reply | |
| Wang et al. | Motion vector reversion-based steganalysis revisited | |
| Yang et al. | A robust DCT-based video watermarking scheme against recompression and synchronization attacks | |
| Swaraja et al. | Video watermarking based on motion vectors of H. 264 | |
| US9087377B2 (en) | Video watermarking method resistant to temporal desynchronization attacks | |
| Wang et al. | Segmentation based video Steganalysis to detect motion vector modification | |
| CN112954318A (en) | Data coding method and device | |
| Ahuja et al. | Robust Video Watermarking Scheme Based on Intra-Coding Process in MPEG-2 Style. | |
| Su et al. | A source video identification algorithm based on motion vectors | |
| Jing et al. | Motion vector based information hiding algorithm for H. 264/AVC against motion vector steganalysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into 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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190222 Termination date: 20190512 |