CN111127290B - An Image Authentication Method Based on AMBTC - Google Patents

Abstract

The present invention provides an AMBTC-based image authentication method, by acquiring an image to be authenticated, and extracting a corresponding first verification watermark; image authentication is performed on the to-be-authenticated image according to the first verification watermark and preset rules , which improves the verification accuracy and resists collage attacks.

Description

An Image Authentication Method Based on AMBTC

technical field

The invention relates to the technical field of image processing, in particular to an AMBTC-based image authentication method.

Background technique

In 2018, Hong Weien et al. proposed Detectability Improved TamperDetection Scheme for Absolute Moment Block Truncation (an effective AMBTC compressed image authentication scheme) in the journal Symmetry, which includes two strategies: LSBR and MSBP. In the LSBR strategy, the traditional AMBTC technology is used to compress the image to obtain the AMBTC compression code. Here, it is assumed that a certain set of compressed codes is (h, l, B). Wherein, h and l become the high quantization value and the low quantization value of 8 bits respectively, and B is the bitmap corresponding to the image block. In the LSBR scheme, the lower a bits of h and the lower b bits of l are vacated and used to carry the verification watermark ac of (a+b) bits. The calculation of the verification watermark ac is shown in the following announcement:

ac=f _a+b (h ^M , l ^M , B),

where f _a+b (·) is a hash operation and the output length is (a+b) bits. ^hM refers to the additional (8-a) bits of h, and ^lM refers to the additional (8-b) bits of l. Finally, the first a bits of the generated (a+b) bit verification watermark are replaced with the lower a bits of h, and the other b bits are replaced with the lower b bits of l, so as to realize the embedding of the verification watermark.

At the same time, in order to further improve the quality of watermarked images, MSB transform technology is proposed to find a set of verification watermarks that can minimize the distortion of image blocks. By sacrificing a certain amount of computing power, a higher quality watermark image quality is achieved. Specifically do the following:

The main strategy of MSBP is: On the basis of LSBR, allow h ^M and l ^M to have a certain range of variation (±1), and then find a group from 9 groups of combinations that can minimize the variance of the local block through operations, so that Each local distortion becomes smaller, which improves the quality of the watermarked image as a whole.

However, this method still has the following disadvantages:

Disadvantage 1: The watermark image quality and verification accuracy can be further improved. Generally, the more verification watermarks are embedded, the better the verification accuracy can be achieved, but at the same time, the quality of the watermarked image will drop sharply. Therefore, there is a trade-off between the watermark image quality and the embedded watermark length. This method is the best solution so far to balance and realize the compatibility between the two, that is, it can achieve up to 8-bit verification watermark embedding and at the same time achieve the quality of most watermark images close to 30dB. But there are still some image watermark quality below 30dB. On the other hand, it is also hoped that the watermark quality can be further improved.

Disadvantage 2: Can't resist collage attack (collage attack: an attack method that copies a part from a watermark-embedded image and pastes it into another watermark-embedded image). In this method, each image block calculates the hash result (verifies the watermark) through its own compression code (partially), and then implements self-embedding in the form of LSB replacement. Therefore, this method cannot withstand collage attacks.

SUMMARY OF THE INVENTION

(1) Technical problems to be solved

In order to solve the above problems of the prior art, the present invention provides an AMBTC-based image authentication method with high verification accuracy and resistance to collage attacks.

(2) Technical solutions

In order to achieve the above-mentioned purpose, the main technical scheme adopted in the present invention includes:

An AMBTC-based image authentication method, comprising steps:

S1, obtain the image to be authenticated, and extract the corresponding first verification watermark;

S2. Perform image authentication on the image to be authenticated according to the first authentication watermark and a preset rule.

(3) Beneficial effects

The beneficial effects of the present invention are: by acquiring the image to be authenticated, and extracting the corresponding first verification watermark; image authentication is performed on the image to be authenticated according to the first verification watermark and preset rules, thereby improving the verification accuracy and high And can resist collage attack.

Description of drawings

1 is a flowchart of an AMBTC-based image authentication method according to an embodiment of the present invention;

2 is a schematic diagram of a reference table RT according to an embodiment of the present invention;

3 is a partial schematic diagram of a misjudgment solution according to an embodiment of the present invention;

4 is a schematic diagram of an adaptive weight reconstructed image according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of an image verification result according to an embodiment of the present invention.

Detailed ways

In order to better explain the present invention and facilitate understanding, the present invention will be described in detail below with reference to the accompanying drawings and through specific embodiments.

Example 1

Please refer to Fig. 1, a kind of image authentication method based on AMBTC, including steps:

Before step S1, it also includes:

S0. Generate a watermark image including the second verification watermark by AMBTC compression according to the original image.

Step S0 specifically includes:

S01. According to the original image, the corresponding AMBTC compression code is obtained after the new variant AMBTC compression processing;

Step S01 is specifically:

S011, each 4*4 image block in the original image is formed into a one-dimensional pixel sequence by raster scanning, and is represented as I _b ={b ₁ ,b ₂ ,...,b ₁₆ };

并使得I_b中每个像素值都包含在U中且U中两两元素之间均不相同，若U的元素个数|U|≥2,则从其中选择两个元素构建一个初始质心集合C＝{c₁,c₂}，并执行步骤S013；若|U|＝1，c₁＝c₂，则执行步骤S016；S012, the centroid is initialized, and U is derived from the I _b , wherein,

And make each pixel value in I _b included in U and the two elements in U are different, if the number of elements in U |U| ≥ 2, select two elements from them to construct an initial centroid set C={c ₁ , c ₂ }, and go to step S013; if |U|=1, c ₁ =c ₂ , go to step S016;

S013. Group calculation, according to Euclid's theorem, divide all elements in I _b into two groups S ₁ and S ₂ , and the basis for the grouping is

|S_r|表示S_r中元素个数；S014, centroid iteration, calculate the centroid of each group

|S _r | represents the number of elements in S _r ;

决定迭代是否终止，ε是一个小于0的常量，若不满足，则c₁＝c′₁,c₂＝c′₂，执行步骤S013；否则执行步骤S016；S015. According to the formula

Determine whether the iteration is terminated, ε is a constant less than 0, if not satisfied, then c ₁ =c' ₁ , c ₂ =c' ₂ , go to step S013; otherwise, go to step S016;

S016. Export AMBTC compression code:

h _k =max(c ₁ ,c ₂ ),

l _k =min(c ₁ ,c ₂ ),

对每个图像区块重复上述步骤，即可得到整个原始图像的AMBTC压缩码

其中,h_k表示高阶量化水平值，l_k表示低阶量化水平值，

表示对应图像区块的位图。

Repeat the above steps for each image block to get the AMBTC compression code of the entire original image

Among them, h _k represents the high-order quantization level value, l _k represents the low-order quantization level value,

A bitmap representing the corresponding image block.

S02, obtain a new compression code after N variant AMBTC compression processing according to the AMBTC compression code, and generate a corresponding second verification watermark by hash function according to the new compression code and position information;

Step S02 is specifically:

S021, generate a 4*4 first reference table RT;

的B_k中的前N比特位置腾出，得到新的压缩码

并根据所述新的压缩码

和位置信息通过hash函数生成相应的N比特的第二验证水印，其中B′_k表示压缩处理后的位图。S022, according to the first parameter N, compress the AMBTC code

The first N bits in B _k are vacated to obtain a new compressed code

and according to the new compression code

and location information to generate a corresponding N-bit second verification watermark through a hash function, where B′ _k represents the compressed bitmap.

S03. Generate a watermark image according to the new compression code and the second verification watermark.

Step S03 is specifically:

A watermark image is generated by bit replacement according to the new compression code and the first verification watermark.

S1, obtain the image to be authenticated, and extract the corresponding first verification watermark;

Step S1 also includes:

According to the second parameter N and the second reference table RT, the third verification watermark is calculated from the image to be authenticated.

S2. Perform image authentication on the image to be authenticated according to the first authentication watermark and a preset rule.

Step S2 is specifically:

Image authentication is performed on the image to be authenticated according to the first verification watermark, the third verification watermark and a preset rule.

The image recognition of the to-be-authenticated image according to the first verification watermark, the third verification watermark and the preset rule is as follows:

It is judged whether the first verification watermark is equal to the third verification watermark, and if so, it is judged that the image to be authenticated has not been tampered with; otherwise, it is judged that the image to be authenticated has been tampered with.

Embodiment 2

The difference between this embodiment and Embodiment 1 is that this embodiment will further illustrate how the above-mentioned AMBTC-based image authentication method of the present invention is implemented in combination with specific application scenarios:

The present invention mainly includes three aspects: watermark generation, watermark embedding and image verification and reconstruction process;

1. Generation of watermark

根据所述AMBTC压缩码通过N变体AMBTC压缩处理后得到新的压缩码，并根据所述新的压缩码和位置信息通过hash函数生成相应的第二验证水印；具体地，对于每一个区块将其对应(h_k,l_k,B′_k)和当前区块位置信息k喂给hash函数以获取第二验证水印ac_k，其长度为N，且ac_k＝f_N(h_k,l_k,B′_k,k).When the original image I is compressed by the new variant AMBTC, the corresponding AMBTC compression code

According to the AMBTC compression code, a new compression code is obtained after N-variant AMBTC compression processing, and a corresponding second verification watermark is generated through a hash function according to the new compression code and location information; Specifically, for each block Feed its corresponding (h _k , l _k , B′ _k ) and the current block position information k to the hash function to obtain the second verification watermark ac _k , whose length is N, and ac _k =f _N (h _k ,l _k ,B′ _k ,k).

2. Embedding of watermarks

(1) New variant AMBTC compression processing

For each 4*4 image block, its 16 pixels are formed into a one-dimensional pixel sequence in a raster scanning manner, and expressed as I _b ={b ₁ ,b ₂ ,...,b ₁₆ }. Next, the new AMBTC compression process can be described as follows:

并使得I_b中出现的每个像素值都含在U中且U中两两元素之间均不相同，若U的元素个数_|U|≥2,则从其中选择两个元素构建一个初始质心集合C＝{c₁,c₂}，并跳转至步骤2)；若|U|＝1，c₁＝c₂，并跳转至步骤5)；1) Centroid initialization, derive U from I _b , where,

And make each pixel value appearing in I _b included in U and the two elements in U are different, if the number of elements in _{U |U|≥2} , select two elements from them to construct an initial The centroid set C={c ₁ , c ₂ }, and jump to step 2); if |U|=1, c ₁ =c ₂ , and jump to step 5);

其中，d(b_v,c_r)＝(b_v-c_r)².，1≤v≤16,1≤r≤2。2) Grouping calculation, according to Euclid's theorem, all elements in I _b are divided into two groups S ₁ and S ₂ , and the basis for the grouping is

Wherein, d(b _v , cr ₎ =(b _v -c _r ) ² . 1≤v≤16, 1≤r≤2.

|S_r|表示S_r中元素个数；3) Centroid iteration, calculating the centroid of each group

|S _r | represents the number of elements in S _r ;

决定迭代是否终止，ε是一个小于0的常量，若不满足，则c₁＝c′₁,c₂＝c′₂，跳转回步骤2)；否则执行步骤5)：4) Determine whether to terminate, according to the formula

Determine whether the iteration is terminated, ε is a constant less than 0, if not satisfied, then c ₁ =c' ₁ , c ₂ =c' ₂ , jump back to step 2); otherwise, go to step 5):

5) Export AMBTC compression code:

h _k =max(c ₁ ,c ₂ ),

l _k =min(c ₁ ,c ₂ ),

对每个图像区块重复上述步骤，即可得到整个原始图像的AMBTC压缩码

其中,h_k表示高阶量化水平值，l_k表示低阶量化水平值，

表示对应图像区块的位图。

Repeat the above steps for each image block to get the AMBTC compression code of the entire original image

Among them, h _k represents the high-order quantization level value, l _k represents the low-order quantization level value,

A bitmap representing the corresponding image block.

(2) N variant AMBTC compression processing

通过实施N变体AMBTC技术，从B_k腾出N比特的空间，用于隐藏第二验证水印，具体步骤如下：when obtained

By implementing the N variant AMBTC technology, N bits of space are freed from B _k to hide the second verification watermark. The specific steps are as follows:

1) Generate a 4*4 first reference table RT, as shown in FIG. 2(a), wherein (a) B _k ; (b) reference table RT; (c) B′ _k (N=2); ( d) _B'k (N=8);

的B_k中的前N比特位置腾出；2) According to the first parameter N, compress the AMBTC code

The first N bit positions in B _k are vacated;

Among them, the determination rule for the position of the first N bits: confirm according to the size of the numbers in the hexagonal box in RT Figure 2(b) from small to large. Fig. 2(c) is B' _k when N is 2; Fig. 2(d) is B' _k when N is 8, 'NA' indicates the vacated position.

并根据所述新的压缩码和位置信息通过hash函数生成相应的N比特的第第二验证水印，其中B′_k表示压缩处理后的位图。3) Implement the above process for each block compressed code to obtain a new compressed code

And according to the new compression code and position information, a corresponding N-bit second verification watermark is generated through a hash function, where B′ _k represents the compressed bitmap.

和相应的N比特的第二验证水印后，通过位元替换的形式，将N比特验证水印依次替换中B′_k腾出的N个位置(顺序根据RT中蓝色六边形框内数字的大小从小到大确认)，从而实现第二验证水印的嵌入，得到水印图像

(3) When a new compressed code is obtained

After and the corresponding N-bit second verification watermark, the N-bit verification watermark is replaced in turn in the N positions vacated by _B'k in the form of bit substitution (the order is based on the number in the blue hexagonal frame in RT). The size is confirmed from small to large), so as to realize the embedding of the second verification watermark, and obtain the watermark image

3. Image verification and reconstruction process

1) Image verification

The first stage:

对于每一对

实施图像验证的步骤如下：Assume that the AMBTC compression code received by the receiver is

for each pair

The steps to implement image validation are as follows:

1) The corresponding RT is generated in the same way, (or shared by sender and receiver).

中提取前N个比特位元(N个比特位元的提取顺序根据RT中六边形框内数字的大小从小到大确认)，并串接成提取的验证水印，记为第一验证水印Eac_k。2) From

Extract the first N bits (the extraction order of N bits is confirmed according to the size of the numbers in the hexagonal frame in RT from small to large), and concatenate them into the extracted verification watermark, which is marked as the first verification watermark Eac _k .

压缩成

并根据水印产生的方法，并从待认证图像中计算得到第三验证水印，即Rac_k。3) According to the second parameter N and the second reference table RT, the

compressed into

And according to the method of watermark generation, the third verification watermark, namely _Rack , is calculated from the image to be authenticated.

Specifically, the generation process of the second verification watermark and the third verification watermark are the same, the difference is that if the watermark image is tampered with, the second verification watermark and the third verification watermark are not the same, otherwise, the calculated third verification watermark is the same as that of the third verification watermark. The second verification watermark obtained during the watermark generation process is the same, the difference between (the second parameter N and the second reference table RT) and (the first parameter N and the first reference table RT) is that the owner is different, but Only when their values are the same can accurate verification be achieved, that is, N and RT are shared by both the sender and the receiver;

4) If the image has not been tampered with, _Eack should be equivalent to _Rack . Therefore, by comparing _Eack (first watermark) and _Rack (third watermark), if they are equal, the block is considered to have not been tampered with; otherwise, it is considered that the block has been tampered with.

After the above process is performed for each block, the tampering detection result of the entire image can be obtained, which is recorded as TM.

second stage:

In order to solve the misjudgment situation in the first stage, the second stage tamper verification is implemented on the TM. The basic principle is: take a block as the center, in a small area of 3*3 around it. The area is divided into 4 parts, which are left, upper, right, and lower, corresponding to the gray-black areas in Fig. 3(a)-(d). Each part is a 2*3 or 3*2 rectangular area. Black indicates a tampered block identified in the first stage, and gray indicates a non-tampered block identified in the first stage. '*' indicates the block that currently needs to be authenticated. If more than 4/6 of the blocks in the above-mentioned 4 parts of the current block (left, top, right, bottom) are identified as tampered blocks, it is further judged that the block is a tampered block, As shown in Figure 3(e)-(f).

By implementing the above two-stage image verification process, the final verification result TM can be obtained.

(2) Image reconstruction

If you need to reconstruct the compressed image, you can reconstruct it according to the following steps:

For a certain pixel to be reconstructed, it uses the reconstructed pixels in the range of 3*3 (part of the pixels can be reconstructed according to the compression code) as a reference, and uses the strategy of adaptive weight to reconstruct its own pixels. For convenience, it is assumed that the pixel to be reconstructed is b' _x , and the reconstructed pixels within the range of 3*3 are X={b' ₁ , b' ₂ , . . . , b' _N }. Then the pixel value of the current b' _x can be estimated as:

为分配到x_t的方差。in,

is the variance assigned to _xt .

A related example is given in Figure 4.

For shortcoming 1, in fact, the image regions have local similarity, so a strategy is proposed to further compress the bitmap bits to make room to hide the verification watermark. Thus, the verification capability of N bits can be realized, N=0-10. First, compared with Hong Weien's method, the verification accuracy is improved by 2 bits, and a novel misjudgment solution is proposed to ensure high-precision and accurate image verification results. At the same time, in order to ensure the quality of the reconstructed image, we propose two optimization methods: the first is to propose a new AMBTC compression strategy; the second is to propose an adaptive weighting strategy to estimate the pixels to be tested, thereby ensuring the best image quality. Table 1 shows the comparison of the experimental results of the image quality between the present invention and the method of Hong Weien et al.

Table 1. Image quality comparison between the present invention and the method of Hong Weien et al. (N=2, 4, 6, 8, 10)

'-': Indicates that the method is not provided or unconditionally implemented.

For disadvantage 2, it cannot resist collage attacks. In the method of Hong Weien et al., each image block calculates the hash result (verifies the watermark) through its own compression code (partially), and then implements self-embedding in the form of LSB replacement. Therefore, this method cannot withstand collage attacks. In the present invention, the location information of the block is further fed into the hash function to ensure that the collage attack can be resisted. At the same time, in the verification stage, a new type of misjudgment solution is proposed, which further improves the image verification effect.

Among them: Figure 5(a) is the tampered compressed image, and 5 areas (5 types of attacks) have been tampered with, and the specific tampered area is shown in Figure 5(a).

A: tamper with the value of the quantization level of the area;

B: tamper with the original bits of the bitmap of the area;

C: tamper with the verification watermark bits of the bitmap of this area;

D: collage attack;

E: copy-paste attack (a flower);

The black part represents the tampered area or the detected area, and the white part represents the untampered area or detected as the untampered area.

Figure 5. Schematic diagram of image verification results (N=10): (a) tampered LeNa image;

Verification results of the present invention: (b) first stage (TDR=99.87%); (f) first stage (TDR=99.97%); (d) blocks not correctly detected; (e) original tampered area;

Verification results of Hong Weien's method: (f) first stage (TDR=87.35%); (g) first stage (TDR=87.61%); (h) blocks that were not detected correctly.

The above descriptions are only examples of the present invention, and are not intended to limit the scope of the patent of the present invention. Any equivalent transformations made by using the contents of the description and drawings of the present invention, or directly or indirectly applied in related technical fields, are similarly included in the within the scope of patent protection of the present invention.

Claims (6)

1. An AMBTC-based image authentication method is characterized by comprising the following steps:

s1, acquiring an image to be authenticated, and extracting a corresponding first verification watermark;

s2, performing image authentication on the image to be authenticated according to the first verification watermark and a preset rule;

step S1 is preceded by:

s0, generating a watermark image containing a second verification watermark through AMBTC compression according to the original image;

step S0 specifically includes:

s01, obtaining corresponding AMBTC compression codes after new variant AMBTC compression processing according to the original image;

s02, obtaining a new compression code through N-variant AMBTC compression processing according to the AMBTC compression code, and generating a corresponding second verification watermark through a hash function according to the new compression code and the position information;

s03, generating a watermark image according to the new compression code and the second verification watermark;

step S01 specifically includes:

s011, forming a one-dimensional pixel sequence by each 4 x 4 image block in the original image according to a raster scanning mode, and expressing the one-dimensional pixel sequence as I_b＝{b₁,b₂,…,b₁₆}；

S012, centroid initialization, from I_bAnd deriving U, wherein,

and make I_bWherein each pixel value is contained in U, and every two elements in U are different, if the number | U | of the elements of U is ≧ 2, two elements are selected from the U to construct an initial centroid set C ═ C₁,c₂And step S013 is executed; if | U | ═ 1, c₁＝c₂Then, go to step S016;

s013, grouping calculation, according to Euclidean theorem, dividing I_bAll elements in the group S₁And S₂The basis of grouping is

Wherein d (b)_v,c_r)＝(b_v-c_r)²，1≤v≤16,1≤r≤2；

S014, iteration of the centroids, calculating the centroids of each group

b_v∈S_r，|S_rI represents S_rThe number of middle elements;

s015, according to the formula | c₁-c′₂|≤ε&|c₂-c′₂Determining whether iteration is terminated or not, wherein epsilon is a constant larger than 0, if not, c₁＝c′₁,c₂＝c′₂Step S013 is executed; otherwise, executing step S016;

s016, deriving an AMBTC compression code:

h_k＝max(c₁,c₂),

l_k＝min(c₁,c₂),

repeating the above steps for each image block to obtain the AMBTC compression code of the whole original image

Wherein h is_kRepresenting a high order quantization level value,/_kRepresenting the value of the low-order quantization level,

a bitmap representing the corresponding image blocks.

2. The AMBTC-based image authentication method according to claim 1, wherein step S02 is specifically:

s021, generating a first reference table RT of 4 x 4;

s022, based on the first parameter N,compressing the AMBTC

B of (A)_kThe first N bit positions are vacated to obtain new compression codes

And based on the new compression code

Generating a corresponding N-bit second verification watermark by a hash function according to the position information, wherein B_k' denotes a bitmap after compression processing.

3. The AMBTC-based image authentication method according to claim 1, wherein step S03 specifically includes:

and generating a watermark image by a bit replacement mode according to the new compression code and the second verification watermark.

4. The AMBTC-based image authentication method according to claim 1, wherein step S1 further includes:

and calculating to obtain a third verification watermark from the image to be authenticated according to the second parameter N and the second reference table RT.

5. The AMBTC-based image authentication method according to claim 4, wherein the step S2 is specifically as follows:

and performing image authentication on the image to be authenticated according to the first verification watermark, the third verification watermark and a preset rule.

6. The AMBTC-based image authentication method according to claim 5, wherein the image recognition of the image to be authenticated according to the first verification watermark, the third verification watermark and a preset rule is specifically:

and judging whether the first verification watermark and the third verification watermark are identical, if so, judging that the image to be authenticated is not tampered, otherwise, judging that the image to be authenticated is tampered.

Images