CN105979268A - Safe information transmission method based on lossless watermark embedding and safe video hiding - Google Patents

Abstract

The invention provides a safe information transmission method based on lossless watermarking embedding and safe video hiding. The method includes the steps of employing a lossless watermark embedding algorithm to embed a lossless watermark in a first sensitive image to obtain a sensitive image with embedded watermark, employing a safe video hiding algorithm to embed the sensitive image with the embed watermark as hidden information in an original video to obtain a video X with secrete information, a receiving end extracting the hidden information from the received video X with secrete information, further extracting and analyzing the extracted hidden information, determining whether the image is tempered during a transmission process, if yes, positioning the tempered position and restoring the image. The lossless watermark embedding technology and high capacity safe video hiding technology are effectively combined, the watermark embedding capacity is substantial increased, the tempering detection is highly accurate, the image is restored in a high quality way, and safe image transmission is ensured.

Description

Information security transmission method based on lossless watermark embedding and security video hiding

Technical Field

The invention belongs to the technical field of information transmission, and particularly relates to an information security transmission method based on lossless watermark embedding and security video hiding.

Background

With the development of networks and multimedia, digital medical information systems, such as hospital information systems, electronic medical record systems, etc., play an increasingly important role. When the medical images are transmitted on the public network, the medical images are easy to be maliciously tampered, so that the medical image tamper detection and recovery are very important.

At present, the integrity of an image is mainly detected by adopting a tamper detection algorithm based on lossless digital watermarking, and the main detection principle is as follows: and (3) partitioning the image into blocks with fixed size, and then selecting the characteristic values of the blocks as watermarks to be embedded in the image blocks. If the image is tampered, the recalculated block characteristics are not matched with the extracted watermark information, and therefore positioning of tampering detection is achieved. For example, Chiang et al divides an image into 4 × 4 blocks, selects the mean value of each block as a recovery feature value, and then embeds the feature value by using a lossless digital watermarking method based on difference expansion. Guo et al first divides the image into 2 × 2 blocks, and then calculates the authentication information of each block using a hash function as watermark embedding.

However, the current partitioning method based on fixed blocks has the following disadvantages: the accuracy of tamper detection is in the size of the block, the smaller the size the more accurate, but the more watermark information needs to be embedded. In addition, the algorithm of the existing watermark embedding technology is poor in embedding capacity, tampering detection accuracy and image quality recovery method, and cannot meet the safety requirement of actual information encryption.

Disclosure of Invention

Aiming at the defects in the prior art, the invention provides an information security transmission method based on lossless watermark embedding and security video hiding, which can effectively solve the problems.

The technical scheme adopted by the invention is as follows:

the invention provides an information security transmission method based on lossless watermark embedding and security video hiding, which comprises the following steps:

step 1, watermark embedding process:

embedding a lossless watermark into the 1 st sensitive image by adopting a lossless watermark embedding algorithm to obtain a watermark-embedded sensitive image;

step 2, embedding the sensitive image embedded with the watermark into an original video by adopting a safe video hiding algorithm to obtain a video carrying hidden information, namely a secret-carrying video X;

step 3, transmitting the secret-carrying video X from the transmitting end to the receiving end through the network; in addition, the sending end also sends the 1 st authentication information to the receiving end in a key mode;

step 4, watermark extraction, tamper detection and recovery:

the receiving end extracts hidden information from the received secret-carrying video X; then further extracting and analyzing the extracted hidden information, and extracting a 2 nd sensitive image and watermark information;

recalculating according to the watermark information extracted in the step 4 to obtain the 2 nd authentication information, and further obtaining the 1 st authentication information provided by the sending end; comparing whether the 1 st authentication information and the 2 nd authentication information are the same, if so, indicating that the 2 nd sensitive image is the same as the 1 st sensitive image, and indicating that the 1 st sensitive image is not tampered in the transmission process; on the contrary, if the 1 st authentication information is different from the 2 nd authentication information, the 1 st sensitive image is tampered in the transmission process, the sensitive image is positioned to the tampered position, and the image is recovered.

Preferably, the watermark embedding process of step 1 specifically includes:

step 1.1, setting the value of a parameter gamma, wherein the gamma is a decimal with a value range of [0,1 ]; the 1 st sensitive image is a square image, the 1 st sensitive image is subjected to quadtree decomposition, and the 1 st sensitive image is decomposed into n image blocks; the n image blocks are denoted image block B1, image block B2 … image block Bn, respectively; recording the size information of each image block after decomposition; a set formed by arranging the size information of all the image blocks according to a set rule is the quadtree decomposition information q;

step 1.2, for any image block B_iAccording to the size information, calculating by adopting a linear weighted interpolation technology to obtain a characteristic value F of the image block_i；

Eigenvalues F of image Block B1, image Block B2 … image Block Bn₁、F₂…F_nAre arranged in reverse direction and spliced together to obtain a total characteristic value F represented asWherein,representing a splicing operation;

step 1.3, mixingConverting the total characteristic value F into binary form as the l-level watermark information w₁；

Step 1.4, according to the I-level watermark information w₁Setting an embedding threshold value T;

step 1.5, setting an initial chaotic parameter x of the chaotic system_kAnd mu, respectively converting the characteristic values F by a reversible integer transform method based on chaos_n、F_n-1…F₁Embedded as watermarks in the image block B1 and the image block B2 … image block Bn in sequence, thereby obtaining a level 1 watermarked image

Step 1.6, calculating by adopting MD5 algorithm to obtain the 1 st level watermark imageThe 1 st-level watermark image authentication code of (1);

step 1.7, performing Huffman coding on the quadtree decomposition information q obtained in the step 1.1, and converting the quadtree decomposition information q into a binary form to obtain quadtree coding huf (q);

the quad-tree coding huf (q), the embedding threshold T and the level 1 watermark image authentication code are combined to form level 2 watermark information w₂；

Step 1.8, adopting LSB replacement algorithm based on lossless compression to carry out watermarking on the 2 nd level information w₂Embedding into level 1 watermark imagesIn the method, a 2 nd-level watermark image is obtained

Preferably, in step 3, the 1 st authentication information sent by the sending end to the receiving end in a key manner includes: initial chaotic parameter x_kAnd mu, and coding length and coding table of the quadtree coding huf (q)。

Preferably, in step 4, the specific processes of watermark extraction, tamper detection and recovery are as follows:

assuming that the receiving end has acquired the initial chaotic parameter x_kAnd mu, the coding length and coding table of the quadtree coding huf (q) are known, then:

step 4.1, the receiving end extracts the 2 nd level watermark image from the videoThen, using LSB extraction algorithm to extract 2 nd level watermark imageAnd recovering the 1 st level watermark image from the secret dataExtracting to obtain a level 1 watermark image authentication code, a quadtree code huf (q) and an embedding threshold value T;

step 4.2, calculating the level 1 watermark image restored in the step 4.1 by reusing the MD5 algorithmThe watermark image authentication code of (1); comparing the authentication code with the level 1 watermark image extracted in the step 4.1, if the authentication code is the same as the authentication code, indicating that the original image is not tampered in the transmission process, and performing lossless recovery on the original image, namely executing the step 4.3; otherwise, executing step 4.3-step 4.7, and carrying out tampering detection and recovery;

step 4.3, according to the obtained initial chaotic parameter x_kMu and an embedding threshold value T, extracting an embedded total characteristic value F by utilizing chaos-based reversible integer transform, and obtaining a recovered image I_R；

Step 4.4, decoding the quadtree coding huf (q) extracted in the step 4.1 to obtain quadtree decomposition information q;

step 4.5, restoring the image I according to the quadtree decomposition information q obtained in the step 4.4_RPerforming quadtree decomposition, decomposing the quadtree into a plurality of image blocks, and obtaining the size information of each image block;

step 4.6, calculating by adopting a linear weighted interpolation technology to obtain the characteristic value of each image block obtained in the step 4.5 to obtain a new characteristic value F', and then obtaining a new total characteristic value

Step 4.7, comparing the total characteristic value F obtained in the step 4.3 with the new total characteristic value F' obtained in the step 4.6 one by one, so as to position the tampered area; assuming that any j-th image block is tampered, the characteristic value F extracted by the image block is used_jThe tampered area is restored.

Preferably, in step 2, the hidden information is embedded into the original video by the following method to obtain the secret-loaded video X:

step 2.1, for three factors, including: factor 1 f₁Is ARE, namely: the amount of variation in inter-frame correlation; factor 2 f₂Is IDC, namely: the variable quantity of the statistic characteristics of the spatial domain DCT coefficients; factor 3 f₃Is the inverse of the available DCT coefficient utilization, i.e.: the inverse of the utilization of the DCT coefficients that can be used to embed information;

according to the practical application background, the factor f is the 1 st factor₁2 nd factor f₂And factor 3 f₃Setting the 1 st weight omega₁2 nd weight ω₂And the 3 rd weight ω₃；

Step 2.2, respectively for the 1 st factor f₁2 nd factor f₂And factor 3 f₃Are normalized, i.e.

${\hat{f}}_i=(f_i-f_{i,\min })/(f_{i,\max }-f_{i,mim})$

Wherein,is a factor f_i1,2, 3;

step 2.3, construct the objective function, which can be expressed asWherein,ω_iaccording to the practical application background setting, when the requirement for safety is higher, omega₁、ω₂The larger the value, the higher the demand for embedding capacity, ω₃The larger the value. The embedding operation can be viewed as an optimization problem as follows:

min(cost(C,C′))＝min(cost(C,M,ρ))

wherein: rho is an adjusting vector formed by all adjusting variables when secret information M is embedded in the current scene, the optimization goal is to find the adjusting vector which enables the objective function to obtain the minimum value, and when the objective function reaches the convergence condition of the optimization algorithm, the secret-carrying video X can be output.

The information security transmission method based on lossless watermark embedding and security video hiding provided by the invention has the following advantages:

the method has the advantages that the lossless watermark embedding technology and the high-capacity safe video hiding technology are effectively improved and fused, the watermark embedding capacity, the tampering detection accuracy and the image quality are remarkably improved, and the safe transmission of the image is ensured.

Drawings

Fig. 1 is a schematic overall flow chart of an information security transmission method based on lossless watermark embedding and security video hiding according to the present invention;

FIG. 2 is a schematic diagram of the level 1 watermark embedding process of the present invention;

FIG. 3 is a schematic diagram of the level 2 watermark embedding process of the present invention;

fig. 4 is a schematic diagram of the watermark extraction process of the present invention;

FIG. 5 is a schematic diagram of a tamper detection and tamper area recovery process according to the present invention;

fig. 6 is a medical diagnostic header image in DICOM format;

FIG. 7 is a DICOM format medical diagnostic knee image;

fig. 8 is a medical diagnostic brain image in DICOM format;

FIG. 9 is a medical diagnostic lung image in DICOM format;

fig. 10 is a medical diagnostic chest image in DICOM format;

FIG. 11 is a DICOM formatted medical diagnostic liver image;

fig. 12 is a raw MRI _ brain image;

FIG. 13 is an exploded view of a quad-tree;

fig. 14 shows an image in which tampering has occurred in the region (288,112,32,32), with a quality PSNR of 30.069 dB;

fig. 15 shows an image in which tampering has occurred in the region (192,352,24,24), with a quality PSNR of 33.092 dB;

fig. 16 is an image after recovering the tampered region in fig. 14 by Kim, with a PSNR 51.719 dB;

fig. 17 is an image after the tampered region of fig. 14 has been restored using the method of the present invention, with a PSNR of 51.746 dB;

fig. 18 is an image after recovering the tampered region in fig. 15 by Kim, with a PSNR 52.123 dB;

fig. 19 is an image after the tampered region of fig. 15 is restored using the method of the present invention, with a PSNR of 70.658 dB;

FIG. 20 is a graph of the mean of the positive detection rate TPR;

FIG. 21 is a graph of the mean of negative detection rates FNR;

FIG. 22 is a schematic diagram of DCT domain information embedding based on optimization algorithm;

FIG. 23 is a graph of average relative entropy of a dense frame after information embedding is performed using an optimization algorithm;

FIG. 24 is a time-domain inter-frame correlation coefficient value-taking diagram of a secret-carrying frame after information embedding is performed by applying an optimization algorithm.

Detailed Description

In order to make the technical problems, technical solutions and advantageous effects solved by the present invention more clearly apparent, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

In regional medical information sharing, such as remote diagnosis, it is important to authenticate the integrity of images during transmission, in addition to proving the true source of the original medical images. The integrity authentication of the medical image is also called tamper detection, and the tamper detection algorithm of the invention can not only judge whether the image is tampered or not, but also accurately position and approximately recover a tampered area.

The invention provides an information security transmission method based on lossless watermark embedding and security video hiding, which mainly comprises the following steps:

step 1, watermark embedding process:

embedding a lossless watermark into the 1 st sensitive image by adopting a lossless watermark embedding algorithm to obtain a watermark-embedded sensitive image;

step 2, embedding the sensitive image embedded with the watermark into an original video by adopting a safe video hiding algorithm to obtain a video carrying hidden information, namely a secret-carrying video X;

step 3, transmitting the secret-carrying video X from the transmitting end to the receiving end through the network; in addition, the sending end also sends the 1 st authentication information to the receiving end in a key mode;

step 4, watermark extraction, tamper detection and recovery:

the receiving end extracts hidden information from the received secret-carrying video X; then further extracting and analyzing the extracted hidden information, and extracting a 2 nd sensitive image and watermark information;

recalculating according to the watermark information extracted in the step 4 to obtain the 2 nd authentication information, and further obtaining the 1 st authentication information provided by the sending end; comparing whether the 1 st authentication information and the 2 nd authentication information are the same, if so, indicating that the 2 nd sensitive image is the same as the 1 st sensitive image, and indicating that the 1 st sensitive image is not tampered in the transmission process; on the contrary, if the 1 st authentication information is different from the 2 nd authentication information, the 1 st sensitive image is tampered in the transmission process, the sensitive image is positioned to the tampered position, and the image is recovered.

The invention provides an information security transmission method based on lossless watermark embedding and security video hiding, which mainly comprises three parts, namely: a watermark embedding process; watermark extraction, tamper detection and recovery processes; a secure video hiding process. These three sections are described in detail below:

watermark embedding process

Firstly, the following main technical content points involved in the watermark embedding process are introduced:

(1) image quadtree decomposition

Definition 1: quadtree representation of images

The image quadtree representation method adopts a pyramid data structure to store the image. The root of the quad-tree corresponds to the whole image, the leaf nodes correspond to single pixels or a square matrix composed of pixels with the same characteristics, and each non-leaf node has 4 sub-nodes.

The quadtree storage of the image is very suitable for square images and images with the integral power of 2 of the number of pixel points. The quadtree is composed of multiple levels, the root of the quadtree is at level 0, and the quadtree is branched one more level per minute. For a quadtree with N levels, the maximum value of the total number of nodes N is

$N_{max}=\underset{k=0}{\overset{n}{\Σ}}{4}^k=(4^{n+1}-1)/3\≈\frac{4}{3}\×4^n$

Definition 2: quadtree decomposition of images.

The quadtree decomposition of an image divides a square image into 4 equal-sized square blocks, and then judges whether the 4 square blocks meet a given homogeneity criterion. If the current block meets the criterion, it remains unchanged, otherwise it continues to be decomposed into 4 square blocks, and it is determined whether the criterion is met until all square blocks meet the given criterion. The default quadtree decomposition criterion may be expressed as

p_max-p_min＝(g_l-1)×γ

In the formula, p_maxAnd p_minRespectively representing the maximum pixel and the minimum pixel of the grey value in the square block, g_lRepresenting the number of gray levels of the pixel, gamma being a range of [0,1]]The decimal fraction of (c). The default criterion is when the maximum pixel to minimum pixel difference in the square block is greater than (g)_l1) × gamma, the square blocks need to be divided further, from the decomposition method of the quadtree, it belongs to image division of indefinite size, and from the result of the division, the pixels in the image block obtained by the decomposition have high homogeneity, suitable for lossless information hiding, the minimum block size is specified to be 2 × 2 in the algorithm of the invention.

The medical image is subjected to quadtree decomposition, the correlation among pixel blocks can be fully utilized, and the length of a characteristic value can be reduced. The medical image is characterized by more continuous background areas, if a fixed block division method is adopted and the mean value of each block is selected as a characteristic value, in order to realize the accuracy of tamper detection, the size of each block is set to be 4 x 4, 16384 blocks (512 x 512/16) are shared, each block needs 8bit storage space, and the characteristic value is 131072 bits shared. By adopting the quadtree decomposition, the decomposition results are 12097 blocks in total, the characteristic values are 96776 bits in total, and the method has obvious advantages compared with a fixed block division method.

(2) Linear weighted interpolation

In the conventional Kim method, the mean value of the image blocks is used as a characteristic value, and two problems exist: 1) the average value of the image block does not consider the correlation between pixels and the texture structure of the block, and when the texture of a tampered area is complex or the gray difference between pixels is large, the average value is not suitable as the gray value of all pixels of the block; 2) choosing the block mean as the feature value is vulnerable to tampering with the block mean, and an attacker is most likely to make the mean constant by simply modifying the pixels in the block. The invention adopts a linear weighted interpolation technology to obtain the characteristic value of the image block. Since the image block after the quadtree decomposition has high homogeneity, it is desirable that the present invention calculates the feature values using only a part of the pixel values in the block, and the amount of calculation is advantageous for the region where the block size is large.

Definition 3: linear weighted eigenvalues of the image blocks. The linear weighted feature values of an image block are represented by a weighted average of diagonal pixels in the block, as described in detail below:

let B be a square image block with size m. Let d⁽¹⁾And d⁽²⁾Pixel values on the positive and negative diagonal of B, d⁽¹⁾＝[B(1,1),B(2,2),…,B(m,m)]，d⁽¹⁾＝[B(1,m),B(2,m-1),…,B(m,1)]And x is the pixel mean value on the positive and negative diagonals and is expressed as x ═ d⁽¹⁾+d⁽²⁾) The table below shows a 4 × 4 pixel block and its positive and negative diagonal pixels.

Let w_j(1. ltoreq. j. ltoreq.2) represents d⁽¹⁾And d⁽²⁾Calculating the characteristic value of B

$F_i=\underset{j=1}{\overset{2}{\Σ}}{w}_j{d}_{i,k}^{\left(j\right)}/(2\×m),1\≤i\≤n,1\≤k\≤m$

In the formula,representing the diagonal elements in the ith image block, n being the total number of image blocks, and m being the size of the current square image block. The set of eigenvalues F is denoted as Representing the stitching operation, the feature values are arranged in reverse order mainly to embed the feature values of the current block in other blocks. In order to reduce the error generated by interpolation, the pixel weight values on the positive and negative diagonals are important. Let D ═ D⁽¹⁾；d⁽²⁾]The weight w is

w＝(DD^T)^-1Dx^T

The derivation process is as follows:

let e be the error generated by the predictive interpolation, which can be expressed as

$e_i=x_i-F_i=x_i-\underset{j=1}{\overset{2}{\Σ}}{w}_j{d}_k^{\left(j\right)}/(2\×m)$

In the formula, i is more than or equal to 1 and less than or equal to m, k is more than or equal to 1 and less than or equal to m, and m is a constant and is omitted when calculating the optimal weight. The set of weights may be obtained by minimizing the sum of squares of the predictive interpolation, which may be expressed as E

$E=\underset{i=1}{\overset{n}{\Σ}}{e}_i^2=\underset{i=1}{\overset{n}{\Σ}}{(x_i-\underset{j=1}{\overset{2}{\Σ}}{w}_j{d}_k^{\left(j\right)})}^2$

Let X be ═ X₁,x₂,…x_m]Then E can be represented as

$E=\left|\right|X-\underset{j=1}{\overset{2}{\Σ}}{w}_j{d}_k^{\left(j\right)}|{|}^2$

Where | l | · | | | is the norm of the vector. Minimizing E, i.e. finding a vectorSuch that its sum is closest to X, which can be determined by finding the vector X at the vectorObtained by spatially orthogonal projection, equivalent to obtaining an orthogonal toVector of (2)The dot product of these two orthogonal vectors is zero and can be expressed as

$d^{\left(h\right)}\·(X-\underset{j=1}{\overset{2}{\Σ}}{w}_j{d}_k^{\left(j\right)})=0,1\≤h\≤2$

The above formula can be transformed into

$d^{\left(h\right)}\underset{j=1}{\overset{2}{\Σ}}{w}_j{d}_k^{\left(j\right)}=d^{\left(h\right)}\·X$

The above formula can be expressed in a matrix form as

DD^Tw＝DX^T

Wherein D ═ D⁽¹⁾；d⁽²⁾]，w＝[w₁,w₂]^TThen a formula for calculating the weight w, w ═ DD (DD), can be derived^T)^{- 1}Dx^T。

For the smooth region of the image texture, the characteristic value obtained by adopting the linear weighted interpolation method is basically consistent with the average value, however, for the complex region, the obtained characteristic value has higher image quality when the tampered region is recovered, and the method can effectively resist the attack of keeping the average value and improve the accuracy of tampering detection. This conclusion is demonstrated in the experimental section below.

(3) Chaos based reversible integer transform

The basic idea of the reversible integer transform based on chaos is to randomly select a pixel as a reference pixel in each image block by using a chaos system, and then embed a watermark according to the relation between other pixels and the reference pixel in the block.

1) Based on chaotic reference pixel selection. The algorithm selects a reference pixel ref in a block by using a highly efficient logistic chaotic map, wherein the value of the ref does not change in the data embedding process. The selection method comprises

Wherein, mu is more than 0 and less than or equal to 4, x_k+1∈[0,1]M × m is the size of the block index (ref) is the index value of ref in the block (default scan order is from top to bottom, left to right). when 3.5699456 < μ ≦ 4, x_k+1The value of (a) is in a chaotic state, and the initial value x_kAnd mu is used as a key to well ensure the security of ref, thereby improving the security performance of the watermarking algorithm.

2) A reversible integer transform. The embedding steps are as follows:

step 1, for any pixel a of the image block except ref, if | a-ref | < T, embedding a bit of watermark, namely

$\left\{\begin{array}{cc}{a}_b=a+T& b\left(i\right)=1\\ a_b=a-T& b\left(i\right)=0\end{array}\right.$

Where T is an embedding threshold, the larger the value of T, the larger the embedding capacity, the larger the distortion introduced by the embedding. b represents the binary data stream to be embedded and l (b) represents its length. a is_bRepresents the pixel after embedding the watermark, and 1 is less than or equal to i less than or equal to l (b). Otherwise, turning to the step 2.

Step 2, adjusting the pixels which satisfy | a-ref | ≧ T, namely:

$\left\{\begin{array}{cc}{a}_b=a+T& a\&GreaterEqual;ref\\ a_b=a-T& a<ref\end{array}\right.$

and 3, repeatedly executing the step 1 and the step 2 until all the blocks are processed or the data to be embedded is embedded.

$\left\{\begin{array}{c}{a}_b<2^d-1\\ a_b>0\end{array}\right.$

Because the pixels with similar gray values are divided into the same block by the quadtree decomposition, for the image block with the pixel value at the upper boundary or the lower boundary, the image block at the lower boundary (the pixel value is close to 0) can be shifted to the right, and the flag is set to be 1; moving the image block at the upper boundary to the left, wherein the image block is equal to-1; the values are embedded as watermark payload for lossless restoration of pixels in the image block.

Initial parameter x of quadtree decomposition result and logistic chaotic mapping of assumed image_kμ known, the data extraction and pixel recovery steps are as follows:

step 1 according to x_kAnd μ, the reference pixel ref in each block of pixels, the relationship of the other pixels in the block to ref,if a is_bGreater than or equal to ref, then a_r＝a_b-T, whereas a_r＝a_b+T，a_rIs a recovered pixel;

step 2 judgment of a_rRelation to ref if a_bIs greater than or equal to ref and | a_r-ref | < T, 1 is extracted, if a_b< ref and | a_r-ref | < T then extract 0;

and 3, repeatedly executing the step 1 and the step 2 until all the blocks are processed or the watermark is extracted.

(4) LSB replacement based on lossless compression

Least Significant Bit (LSB) replacement is one of the simplest methods in the field of multimedia digital watermarking, and in order to correctly extract data and losslessly recover image pixels, the quad-tree decomposition information of the original image (denoted as q, with length l (q)) must be provided to the extractor. The algorithm embeds the quadtree decomposition information by adopting an LSB replacement algorithm based on lossless compression, the lossless compression algorithm is used for generating extra embedding space in an LSB data stream of an original image to store q, and n is set_AIs a set of image pixels (I)₁,I₂,…,I_n) The number of pixels in the pixel. The space for embedding q is generated using the following equation.

l(q)+l(R_RLC(P_LSB(I₁,I₂,…,I_n))＝n_A

In the formula, P_LSB(I₁,I₂,…,I_n) Is I₁,I₂,…,I_nOf the least significant bit plane, R_RLC(P_LSB(I₁,I₂,…,I_n) For using RLC method to P_LSB(I₁,I₂,…,I_n) The result after compression. Once the embedding space is generated, q and P are added_LSB(I₁,I₂,…,I_n) And (4) embedding.

Since each image block after the quadtree decomposition can be used with 2ⁿ×2ⁿFormally expressed, each block can be divided intoIs converted into a binary form. The tile sizes and corresponding binaries are shown in the table below.

Size coding of image blocks after quadtree decomposition

Considering that the number of blocks decomposed into large sizes is small, the length of q, denoted as huf (q), can be further reduced using Huffman coding according to the frequency of occurrence of blocks of different sizes.

Based on the related contents described above, the watermark embedding process provided by the present invention is performed in two stages, where the l-th stage is used for embedding the feature value, and the 2-nd stage is used for embedding the image authentication code and the quadtree decomposition information in the 1-th stage watermark image. Wherein the level 1 watermark embedding process is shown in fig. 2 and the level 2 watermark embedding process is shown in fig. 3.

With reference to fig. 2 and 3, the watermark embedding process specifically includes:

step 1.1, setting the value of a parameter gamma, wherein the gamma is a decimal with a value range of [0,1 ]; the 1 st sensitive image is a square image, the 1 st sensitive image is subjected to quadtree decomposition, and the 1 st sensitive image is decomposed into n image blocks; the n image blocks are denoted image block B1, image block B2 … image block Bn, respectively; recording the size information of each image block after decomposition; a set formed by arranging the size information of all the image blocks according to a set rule is the quadtree decomposition information q;

step 1.2, for any image block B_iAccording to the size information, calculating by adopting a linear weighted interpolation technology to obtain a characteristic value F of the image block_i；

Eigenvalues F of image Block B1, image Block B2 … image Block Bn₁、F₂…F_nAre arranged in reverse direction and spliced together to obtain a total characteristic value F represented asWherein,representing a splicing operation;

step 1.3, converting the total characteristic value F into a binary form as the I-level watermark information w₁；

Step 1.4, according to the I-level watermark information w₁Setting an embedding threshold value T;

step 1.5, setting an initial chaotic parameter x of the chaotic system_kAnd mu, respectively converting the characteristic values F by a reversible integer transform method based on chaos_n、F_n-1…F₁Embedded as watermarks in the image block B1 and the image block B2 … image block Bn in sequence, thereby obtaining a level 1 watermarked image

Step 1.6, calculating by adopting MD5 algorithm to obtain the 1 st level watermark imageThe 1 st-level watermark image authentication code of (1);

step 1.7, performing Huffman coding on the quadtree decomposition information q obtained in the step 1.1, and converting the quadtree decomposition information q into a binary form to obtain quadtree coding huf (q);

the quad-tree coding huf (q), the embedding threshold T and the level 1 watermark image authentication code are combined to form level 2 watermark information w₂；

Step 1.8, adopting LSB replacement algorithm based on lossless compression to carry out watermarking on the 2 nd level information w₂Embedding into level 1 watermark imagesIn the method, a 2 nd-level watermark image is obtained

The 1 st authentication information sent by the sending end to the receiving end in a key mode comprises: initial chaotic parameter x_kAnd μ, and a coding length and coding table of the quadtree coding huf (q).

The specific processes of watermark extraction, tamper detection and recovery are as follows:

referring to fig. 4, a schematic diagram of a watermark extraction process is shown; referring to fig. 5, a process of tamper detection and recovery of a tampered area is schematically illustrated.

Assuming that the receiving end has acquired the initial chaotic parameter x_kAnd mu, the coding length and coding table of the quadtree coding huf (q) are known, then:

step 4.1, the receiving end extracts the 2 nd level watermark image from the videoThen, using LSB extraction algorithm to extract 2 nd level watermark imageAnd recovering the 1 st level watermark image from the secret dataExtracting to obtain a level 1 watermark image authentication code, a quadtree code huf (q) and an embedding threshold value T;

step 4.2, calculating the level 1 watermark image restored in the step 4.1 by reusing the MD5 algorithmThe watermark image authentication code of (1); comparing the authentication code with the level 1 watermark image extracted in the step 4.1, if the authentication code is the same as the authentication code, indicating that the original image is not tampered in the transmission process, and performing lossless recovery on the original image, namely executing the step 4.3; otherwise, executing step 4.3-step 4.7, and carrying out tampering detection and recovery;

in the step 4.3, the step of the method,according to the acquired initial chaotic parameter x_kMu and an embedding threshold value T, extracting an embedded total characteristic value F by utilizing chaos-based reversible integer transform, and obtaining a recovered image I_R；

Step 4.4, decoding the quadtree coding huf (q) extracted in the step 4.1 to obtain quadtree decomposition information q;

step 4.5, restoring the image I according to the quadtree decomposition information q obtained in the step 4.4_RPerforming quadtree decomposition, decomposing the quadtree into a plurality of image blocks, and obtaining the size information of each image block;

step 4.6, calculating by adopting a linear weighted interpolation technology to obtain the characteristic value of each image block obtained in the step 4.5 to obtain a new characteristic value F', and then obtaining a new total characteristic value

Step 4.7, comparing the total characteristic value F obtained in the step 4.3 with the new total characteristic value F' obtained in the step 4.6 one by one, so as to position the tampered area; assuming that any j-th image block is tampered, the characteristic value F extracted by the image block is used_jThe tampered area is restored.

Performance analysis:

the performance of image tamper detection is mainly measured by four indexes: tamper detection accuracy, embedding capacity, covert image quality, and quality of tamper recovery images.

1) The tamper detection accuracy is calculated by the following two indicators:

positive detection rate (TPR): the rate of the tampered blocks in the total number of all the tampered blocks is correctly detected;

negative detection ratio (FNR): i.e. the ratio of the blocks that were falsely detected as having been tampered to the total number of all tampered blocks.

2) Embedding capacity. To better compare with other tamper detection algorithms, the embedding capacity only calculates the amount of data of the level 1 embeddable characteristic value.

The embedding capacity of the invention is related to an embedding threshold T and a quadtree decomposition parameter gamma, when the gamma is smaller, the block size is more, the correlation among image block pixels is larger, and the freedom of selecting the threshold T is large. In general, the smaller γ, the smaller the threshold T. For medical images, by setting reasonable parameters γ and threshold T, the embedding capacity of the algorithm ranges between (0.75 bit/pixel, 1 bit/pixel), the lower limit represents that all blocks are 2 × 2 blocks, and the upper limit represents that all blocks are close to 256 × 256.

3) Quality of the stego image. The quality of the stego image is directly related to the threshold T, which is ≦ 6 in order to ensure that no large distortions are caused. In an extreme case, all pixels are adjusted by 6 gray levels, and the minimum PSNR is obtained according to a calculation formula of PSNR peak signal-to-noise ratio_min10 × lg (255 × 255/36) is 32.75 dB.

The accuracy of tamper detection of the algorithm and the image quality after restoration of the tampered area are given by later experimental results.

Experimental results and discussion:

in the experiment, 100 medical images in DICOM format (512 × 512 images) were randomly selected as experimental objects (all from the data center of the Xiangya medical college of the university of Central and south), which are limited to sections, and 6 of the medical images are taken as examples, and as shown in FIGS. 6 to 11, medical diagnosis head images, knee images, brain images, lung images, chest images and liver images in DICOM format are respectively selected. The key point of the experiment is to compare the reversibility of watermark embedding, embedding capacity, secret image quality, tampering detection precision and image quality after the tampered area is recovered. The experimental results are all from simulation data of matlab7.0 tool, and the experimental design scheme is as follows: 1) the reversibility of the algorithm is that the consistency of the recovered image and the original image is tested when the watermark image is not tampered; 2) testing the quadtree decomposition condition, the embedding threshold, the capacity and the secret image quality of the carrier image; 3) comparing the performance of the reversible integer transform method adopted by the invention with that of a similar method; 4) comparing the restored image quality of the tampered region with the Kim method; 5) the accuracy of tamper detection is compared to Kim's method. When the watermark image is not tampered, the extracted authentication code is consistent with the recalculated authentication code, and the finally recovered image is consistent with the original image, so that the reversibility of the algorithm is proved.

The quadtree decomposition, embedding threshold T, embedding capacity and quality of the stego image of 6 carrier images in DICOM format are shown in the following table.

Quadtree decomposition result, threshold T, embedding capacity and PSNR of 6 carrier images

Note that the embedding capacities listed in the above table are merely the number of eigenvalues that can be embedded when a reversible integer transform is employed. As can be seen from the above table, with the difference between the decomposition parameter y and the texture structure of the carrier image itself, the number of blocks after the quadtree decomposition is different, which results in different numbers of the feature values to be embedded. If the eigenvalue of each block is converted into an 8-bit binary, it can be seen from the total number of blocks and capacity in the table above that sufficient embedding capacity can be obtained by selecting an appropriate threshold T to embed all binary streams of eigenvalues. Since the reversible integer transform used in the present invention is the same as the method of Deng, the difference in embedding capacity and stego image quality is given in the table below for the two methods, with the same two images (MRI _ brain and US _ liver images) being selected for the test images. Among these, the methods of Deng are described in Deng X H, Chen Z G, Deng X H, et a1. inorganic dual-layer reversible authentication for a mechanical image authentication [ J ]. Advanced science Letters,2011,4 (11): 3678-3684.

Comparison of embedding Capacity and PsNR for the inventive Algorithm and the document Deng

As can be seen from the above table, the algorithm of the present invention is significantly better than the method of Deng in embedding capacity, because the method of Deng adopts fixed block division (size is 2 × 2), when the texture of the current block is complex, the difference between other pixels and the reference pixel is greater than the threshold value, and cannot be used for embedding the watermark, whereas the image block obtained by the algorithm through the quadtree decomposition has high homogeneity, the difference between other pixels and the reference pixel is smaller, and more pixels are used for embedding data. From the aspect of embedding capacity, compared with the method of fixed blocks, the quadtree decomposition can fully consider the texture structure of the image and is a better choice for partitioning the image. On the quality of a secret image, the texture of an MRI _ brain image is complex, more pixels are adjusted by the method of Deng by the amplitude of a threshold value, so that the quality of the secret image is low, and the quality of the secret image is better than that of the method of Deng although more information is embedded. However, for US _ liver images with smoother texture, the algorithm herein adjusts more pixels in embedding more data, and the quality of the stego image is lower than that of Deng.

To demonstrate the advantage of the algorithm herein in feature value selection, the method of Kim is emphasized^[55]The quality of the recovered image in the tampered area is compared. Among these, the method of Kim is described in Kim K S, Lee M J, Lee J W, et a1.region-based sampling detection and recovery using homology analysis in quality-sensitive imaging [ J]Computer Vision and Image Understanding, 201l, 115 (9): 1308.1323. the results of the experiment are given in figure 9. Fig. 12 is an original MRI _ brain image, fig. 13 is a quadtree exploded view (r is 0.0027), fig. 14 is an image in which tampering has occurred in a region (288,112,32,32), quality PSNR is 30.069dB, fig. 15 is an image in which tampering has occurred in a region (192,352,24,24), quality PSNR is 33.092dB, fig. 16 is an image after the tampered region in fig. 14 is restored by Kim, quality PSNR is 51.719dB, fig. 17 is an image after the tampered region in fig. 14 is restored by using the algorithm of this document, quality PSNR is 51.746dB, fig. 18 is an image after the tampered region in fig. 14 is restored by using the algorithm of this document, quality PSNR is 51.746dB, and fig. 18 is an image of fig. 18The image after the tampered region in fig. 15 is restored by Kim, and the quality PSNR is 52.123dB, and fig. 19 is the image after the tampered region in fig. 15 is restored by using the algorithm of this document, and the quality PSNR is 70.658 dB.

In fig. 14, the tamper-evident area is in a smoother zone of the image, (288,112,32,32) represented as a 32 x 32 image block starting from the coordinate point (288, 112). Since the feature values obtained by Kim's method and the algorithm herein in the smoothed region are substantially the same, the restored image quality of both methods is substantially the same in fig. 16 and 17. In fig. 15, however, the tampered region occurs in a region (192,352,24,24) where the texture is more complex, and the restored image quality is more greatly different. As can be seen from the PSNR values in fig. 18 and fig. 19, it is clear that the feature values obtained by linear weighted interpolation proposed herein are significantly better than the mean algorithm because the linear interpolation method takes the correlation between pixels into full account.

In fig. 14, the tamper-evident area is in a smoother zone of the image, (288,112,32,32) represented as a 32 x 32 image block starting at coordinate point (288,112). Since the feature values obtained by Kim's method and the algorithm herein in the smoothed region are substantially the same, the restored image quality of both methods is substantially the same in fig. 16 and 17. In fig. 15, however, the tampered region occurs in the region (192,352,24,24) where the texture is more complex, and the restored image quality is more varied. As can be seen from the PSNR values in fig. 18 and fig. 19, it is clear that the feature values obtained by linear weighted interpolation proposed herein are significantly better than the mean algorithm because the linear interpolation method takes the correlation between pixels into full account.

To test the tamper detection accuracy of the algorithm, the stego-image of the MRI _ brain was tampered with in the experiment. The specific modification method is as follows: a certain number of 32 x 32 blocks are randomly selected and their pixels are modified while keeping the mean value of the block constant, e.g. by adding one grey level to half the pixels and subtracting one grey level from the other half. The average of the tamper detection accuracy of the algorithm and Kim method herein after 50 experiments was performed. Fig. 20 is a graph showing the mean value of the positive detection rate TPR, and fig. 21 is a graph showing the mean value of the negative detection rate FNR. As can be seen from fig. 20 and 21, as the number of tampered image blocks increases, the positive detection rate TPR starts to decrease and FNR starts to increase in the Kim method. The reason is that some tampered blocks happen exactly in the quad-tree decomposed square blocks and the mean does not change. Although the tamper detection accuracy of the proposed method also decreases with the increase of the tampered blocks, the method is obviously superior to the Kim method because the attack of keeping the mean value changes the block eigenvalue obtained by the linear weighted interpolation technique to a great extent. In addition, if the extraction of the layer 2 data is damaged by the tampering operation, i.e. the wrong quadtree decomposition information will result in an increase of the positive detection rate TPR.

The integrity authentication of the medical image under the network environment has very important significance, and the image tampering detection method based on the lossless digital watermark has obvious advantages. Due to the typical regional characteristics and the complex texture characteristics of the key regions of the medical image, the method for dividing the fixed block is not suitable for the medical image. The non-fixed block dividing method based on the quadtree decomposition is provided, the regional and textural features of the medical image are well considered, and in addition, the new feature value is obtained by utilizing a linear weighted interpolation method to serve as watermark information aiming at the defect of the Kim method in selecting the feature value of the image block. Aiming at the high complexity of the watermark embedding algorithm provided by the Kim method, the reversible integer transform method based on chaos is provided to reduce the complexity of the algorithm and improve the safety. Experimental results show that the algorithm has obvious advantages in watermark embedding capacity, tampering detection accuracy and restored image quality. The method is suitable for tamper detection and high-quality recovery of quality sensitive images in a network environment.

The invention also provides a DCT domain information embedding method based on an optimization algorithm, referring to FIG. 22, which is mainly used for embedding hidden information into an original video to obtain a secret-carrying video X, and comprises the following steps:

step 2.1, for three factors, including: factor 1 f₁Is ARE, namely: the amount of variation in inter-frame correlation; factor 2 f₂Is IDC, namely: statistical characterization of spatial domain DCT coefficientsThe amount of change in (c); factor 3 f₃Is the inverse of the available DCT coefficient utilization, i.e.: the inverse of the utilization of the DCT coefficients that can be used to embed information;

according to the practical application background, the factor f is the 1 st factor₁2 nd factor f₂And factor 3 f₃Setting the 1 st weight omega₁2 nd weight ω₂And the 3 rd weight ω₃；

Step 2.2, respectively for the 1 st factor f₁2 nd factor f₂And factor 3 f₃Are normalized, i.e.

${\hat{f}}_i=(f_i-f_{i,\min })/(f_{i,\max }-f_{i,mim})$

Wherein,is a factor f_i1,2, 3;

step 2.3, construct the objective function, which can be expressed asWherein,ω_iaccording to the practical application background setting, when the requirement for safety is higher, omega₁、ω₂The larger the value, the higher the demand for embedding capacity, ω₃The larger the value. The embedding operation can be viewed as an optimization problem as follows:

min(cost(C,C′))＝min(cost(C,M,ρ))

wherein: rho is an adjusting vector formed by all adjusting variables when secret information M is embedded in the current scene, the optimization goal is to find the adjusting vector which enables the objective function to obtain the minimum value, and when the objective function reaches the convergence condition of the optimization algorithm, the secret-carrying video X can be output.

The above steps can be summarized as follows:

when designing a DCT domain information embedding module based on an optimization algorithm, three factors need to be considered: the variation of the inter-frame correlation, the variation of the spatial DCT coefficient statistical characteristics, and the utilization of the DCT coefficients (non-0 AC coefficients) that can be used to embed information. Certain restriction factors exist among the video frames, for example, the improvement of the utilization rate of the DCT coefficients of the embeddable information will cause the transformation of the video time domain inter-frame statistical characteristic and the spatial statistical characteristic to be aggravated. An optimization algorithm will thus be introduced at the embedding module to balance the three types of factors affected by the information embedding operation. The information embedding module takes the corresponding weight omega of the various factor facilities according to the actual application background₁、ω₂And ω₃And by three factors (f)₁:ARE,f₂:IDC,f₃The inverse of the utilization rate of the DCT coefficient can be used) as an objective function of the optimization algorithm, and a closed-loop feedback mechanism is constructed to adjust a specific information embedding mode. And when the target function reaches the convergence condition of the optimization algorithm, the secret-carrying video X can be output.

And (3) performance verification:

the corresponding parameter settings in the test are shown in the following table:

three carrier video file corresponding information

The embedded information is a file with the size of 141.8kB, and corresponding weights of the three factors are respectively set as omega when an objective function is constructed₁＝0.4、ω₂0.4 and ω₃0.2, initial temperature t in simulated annealing algorithm₀1000, upper limit of number of times of continuous non-optimal solution in internal circulationUpper limit of temperature drop time T in external circulation_i100, the temperature drop rate α is 0.9, the parameter β is 0.98 when feasible solution is searched, and fig. 23 and fig. 24 show the average relative entropy value of the encrypted frame and the correlation coefficient value between time domains after the application of the optimization algorithm to implement information embedding.

The security problem of video steganography is one of the key directions of the research of video steganography technology. The invention provides an information steganography frame taking compressed video stream as a carrier, which adjusts the embedding amount of secret information in each frame by applying the statistic invisibility measurement among video frames, thereby improving the security of steganography algorithm to resist the existing steganography analysis method. In the steganography framework, compressed video spatial domain and time domain statistical characteristics related to safety are constructed, and the effectiveness of the safety steganography framework is tested and verified by applying a simulated annealing algorithm.

The information security transmission method based on lossless watermark embedding and security video hiding provided by the invention has the following advantages:

(1) in the aspect of lossless watermark embedding technology research, a tampering recovery authentication watermark technology is combined, comprehensive research is carried out on three technologies of tampering authentication, tampering recovery and lossless watermark embedding, and a medical image lossless watermark scheme capable of tampering positioning and recovering is provided through analyzing, testing and improving the prior related technologies. Due to the typical regional characteristics and the complex texture characteristics of the key regions of the medical image, the method for dividing the fixed block is not suitable for the medical image. The invention utilizes the quadtree decomposition to divide the non-fixed block, well considers the region and texture characteristics of the medical image, and utilizes the linear weighted interpolation method to obtain a new characteristic value as the watermark information, thereby avoiding the problem of low quality of the recovered image when recovering the complicated texture region. In addition, the complexity of the algorithm is reduced and the safety is improved by using the chaos-based reversible integer transform method.

(2) The steganography medical image in the video requires that the steganography scheme not only has large capacity, but also is safe, can resist normal processing such as compression and the like, and can resist malicious damage such as collusion attack and the like. But capacity and security are contradictory, requiring a compromise between the two. In order to increase the practicability of the algorithm, the normal processing such as compression resistance and the like is also the problem to be considered by the algorithm, so that the invention researches the steganography of the compressed domain video and pays attention to the safety problem of the steganography of the compressed video. The method comprises the steps of researching video steganography security in a statistical sense, forming an information steganography frame with compressed video streams as carriers, introducing a compressed video security embedding criterion, giving an optimized compressed domain security information embedding algorithm, adjusting the embedding amount and the embedding position of secret information in each frame based on the security measurement of a video space domain and a time domain so as to improve the security of the steganography algorithm, constructing the compressed video space domain and time domain statistical characteristics related to the security in the steganography frame, and performing test verification on the effectiveness of the security steganography frame by applying a simulated annealing algorithm. And particularly, a compressed video security steganography framework giving consideration to time domain and space domain characteristics based on a global optimization strategy, and when the method is actually applied, an optimization algorithm objective function can be adjusted according to application requirements to determine a proper information embedding scheme.

The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and improvements can be made without departing from the principle of the present invention, and such modifications and improvements should also be considered within the scope of the present invention.

Claims (5)

1.An information security transmission method based on lossless watermark embedding and security video hiding is characterized by comprising the following steps:

step 1, watermark embedding process:

embedding a lossless watermark into the 1 st sensitive image by adopting a lossless watermark embedding algorithm to obtain a watermark-embedded sensitive image;

step 2, embedding the sensitive image embedded with the watermark into an original video by adopting a safe video hiding algorithm to obtain a video carrying hidden information, namely a secret-carrying video X;

step 3, transmitting the secret-carrying video X from the transmitting end to the receiving end through the network; in addition, the sending end also sends the 1 st authentication information to the receiving end in a key mode;

step 4, watermark extraction, tamper detection and recovery:

the receiving end extracts hidden information from the received secret-carrying video X; then further extracting and analyzing the extracted hidden information, and extracting a 2 nd sensitive image and watermark information;

recalculating according to the watermark information extracted in the step 4 to obtain the 2 nd authentication information, and further obtaining the 1 st authentication information provided by the sending end; comparing whether the 1 st authentication information and the 2 nd authentication information are the same, if so, indicating that the 2 nd sensitive image is the same as the 1 st sensitive image, and indicating that the 1 st sensitive image is not tampered in the transmission process; on the contrary, if the 1 st authentication information is different from the 2 nd authentication information, the 1 st sensitive image is tampered in the transmission process, the sensitive image is positioned to the tampered position, and the image is recovered.

2. The information security transmission method based on lossless watermark embedding and security video hiding according to claim 1, wherein the watermark embedding process of step 1 specifically includes:

step 1.1, setting the value of a parameter gamma, wherein the gamma is a decimal with a value range of [0,1 ]; the 1 st sensitive image is a square image, the 1 st sensitive image is subjected to quadtree decomposition, and the 1 st sensitive image is decomposed into n image blocks; the n image blocks are denoted image block B1, image block B2 … image block Bn, respectively; recording the size information of each image block after decomposition; a set formed by arranging the size information of all the image blocks according to a set rule is the quadtree decomposition information q;

step 1.2, for any image block B_iAccording to the size information, calculating by adopting a linear weighted interpolation technology to obtain a characteristic value F of the image block_i；

Eigenvalues F of image Block B1, image Block B2 … image Block Bn₁、F₂…F_nAre arranged in reverse direction and spliced together to obtain a total characteristic value F represented asWherein,representing a splicing operation;

step 1.3, converting the total characteristic value F into a binary form as the I-level watermark information w₁；

Step 1.4, according to the I-level watermark information w₁Setting an embedding threshold value T;

step 1.5, setting an initial chaotic parameter x of the chaotic system_kAnd μ, wherein μ is a Logistic parameter,

respectively converting the characteristic values F by a chaos-based reversible integer transformation method_n、F_n-1…F₁Embedded as watermarks in the image block B1 and the image block B2 … image block Bn in sequence, thereby obtaining a level 1 watermarked image

Step 1.6, calculating by adopting MD5 algorithm to obtain the 1 st level watermark imageThe 1 st-level watermark image authentication code of (1);

step 1.7, performing Huffman coding on the quadtree decomposition information q obtained in the step 1.1, and converting the quadtree decomposition information q into a binary form to obtain quadtree coding huf (q);

the quad-tree coding huf (q), the embedding threshold T and the level 1 watermark image authentication code are combined to form level 2 watermark information w₂；

Step 1.8, adopting LSB replacement algorithm based on lossless compression to carry out watermarking on the 2 nd level information w₂Embedding into level 1 watermark imagesIn the method, a 2 nd-level watermark image is obtained

3. The information security transmission method based on lossless watermark embedding and security video hiding as claimed in claim 2, wherein in step 3, the 1 st authentication information sent by the sending end to the receiving end in a key manner includes: initial chaotic parameter x_kAnd μ, and a coding length and coding table of the quadtree coding huf (q).

4. The information security transmission method based on lossless watermark embedding and security video hiding according to claim 3, wherein in step 4, the specific processes of watermark extraction, tamper detection and recovery are as follows:

assuming that the receiving end has acquired the initial chaotic parameter x_kAnd mu, the coding length and coding table of the quadtree coding huf (q) are known, then:

step 4.1, the receiving end extracts the 2 nd level watermark image from the videoThen, using LSB extraction algorithm to extract 2 nd level watermark imageAnd recovering the 1 st level watermark image from the secret dataExtracting to obtain a level 1 watermark image authentication code, a quadtree code huf (q) and an embedding threshold value T;

step 4.2, calculating the level 1 watermark image restored in the step 4.1 by reusing the MD5 algorithmThe watermark image authentication code of (1); comparing with the level 1 watermark image authentication code extracted in the step 4.1, if the level 1 watermark image authentication code is the same, indicating that the original image is in the transmission processIf the image is not tampered, performing lossless recovery on the original image, namely executing the step 4.3; otherwise, executing step 4.3-step 4.7, and carrying out tampering detection and recovery;

step 4.3, according to the obtained initial chaotic parameter x_kMu and an embedding threshold value T, extracting an embedded total characteristic value F by utilizing chaos-based reversible integer transform, and obtaining a recovered image I_R；

Step 4.4, decoding the quadtree coding huf (q) extracted in the step 4.1 to obtain quadtree decomposition information q;

step 4.5, restoring the image I according to the quadtree decomposition information q obtained in the step 4.4_RPerforming quadtree decomposition, decomposing the quadtree into a plurality of image blocks, and obtaining the size information of each image block;

step 4.6, calculating by adopting a linear weighted interpolation technology to obtain the characteristic value of each image block obtained in the step 4.5 to obtain a new characteristic value F', and then obtaining a new total characteristic value

Step 4.7, comparing the total characteristic value F obtained in the step 4.3 with the new total characteristic value F' obtained in the step 4.6 one by one, so as to position the tampered area; assuming that any j-th image block is tampered, the characteristic value F extracted by the image block is used_jThe tampered area is restored.

5. The information security transmission method based on lossless watermark embedding and security video hiding as claimed in claim 1, wherein in step 2, the hidden information is embedded into the original video by the following method to obtain the secret video X:

step 2.1, for three factors, including: factor 1 f₁Is ARE, namely: the amount of variation in inter-frame correlation; factor 2 f₂Is IDC, namely: the variable quantity of the statistic characteristics of the spatial domain DCT coefficients; factor 3 f₃Is the inverse of the available DCT coefficient utilization, i.e.: the inverse of the utilization of the DCT coefficients that can be used to embed information;

according to the practical application background, the factor f is the 1 st factor₁2 nd factor f₂And factor 3 f₃Setting the 1 st weight omega₁2 nd weight ω₂And the 3 rd weight ω₃；

Step 2.2, respectively for the 1 st factor f₁2 nd factor f₂And factor 3 f₃Are normalized, i.e.

${\hat{f}}_i=(f_i-f_{i,min})/(f_{i,max}-f_{i,mim})$

Wherein,is a factor f_i1,2, 3;

step 2.3, construct the objective function f, which can be expressed asWherein,ω_iaccording to the setting of practical application background, when the safety requirement is metHigher is ω₁、ω₂The larger the value, the higher the demand for embedding capacity, ω₃The larger the value. The embedding operation can be viewed as an optimization problem as follows:

min(cost(C,C′))＝min(cost(C,M,ρ))

wherein: rho is an adjusting vector formed by all adjusting variables when secret information M is embedded in the current scene, the optimization goal is to find the adjusting vector which enables the objective function to obtain the minimum value, and when the objective function reaches the convergence condition of the optimization algorithm, the secret-carrying video X can be output.