TWI375190B - - Google Patents

Description

[Technical Field] The present invention relates to an information hiding method, and more particularly to a method for directly hiding and hiding a large number of repeatedly hidden bit blocks to cover the last two significant bits of a pixel. . [Prior Art] Image steganography is one of the most widely recognized technical fields in information hiding applications; it uses cover function (covering functi〇n) to hide secret information to cover images. Therefore, the third party cannot find the covered secret information during the overlay image transmission process. The general information hiding technology must have the following characteristics: (1) Impercity: The camouflage image that covers the hidden information of the image, which has little change in the human sensory vision, and is not easy to detect and cover the image. Differences exist, in short, after hiding secret information, there should be good transparency for overlay images; (2) hidden information capacity (Capacity): information hiding technology helps to capture large amounts of information and can The higher the hidden secret information capacity, the better, the higher the better the hiddenness; that is, the ability to hide more secret information while maintaining the same image quality is the preferred hidden 1375190 technology; (3) Efficiency: Information hiding technology is simple and efficient. 'Hidden data capture and capture is fast and time-saving; (4) Recoverability: Recoverability hidden in camouflage image removal After the secret information, 'will disguise the image as a reply action' to make the camouflage image return to the overlay image without distortion; and (5) low distortion: a hidden secret Camouflage image information, in the human sense of Lu must be no undue distortion 'and shall generally increases as the hidden secret information capacity leads to higher distortion. In the hiding process, the cover function is an important key to reducing the degree of steganographic modifications and maintaining the quality of good stego images. Previous approaches by experts and scholars have attempted to produce high embedding rates and high efficiency hidden images; however, increasing the high hiding rate also reduces the hidden performance. In the study of image hiding techniques, I?. Crandall, “Some notes on steganography,, 5 Posted on steganography mailing list and A. Westfield, “F5: asteganography algorithm”, Proc. of 4th Int. Workshop Information Hiding 2001, Lecture Notes in Computer Science, vol. 2137, pp. 289-302, 2001, proposes a matrix coding method that uses block-wise as a processing unit for hidden information to increase hidden performance. (embeddingefficiency) and maintain low coverage image distortion. In addition, D. Sctoifeld and

A. Winkler, “Embedding with syndrome coding based on BCH codes Proc. 8th ACM Workshop on Multimedia and Security, PP. 214-223, 2006 ’ proposes a BCH code (Bose, Chaudhuri,

HocquenghemCode) is the main coverage method, which can achieve the purpose of balanceoff (converting overhead) and hidden effects. In order to improve the hidden performance of binary covering codes, F. Willems and M. Dijk, "Capacity and codes for embedding information in gray-scale signals", IEEE Trans. Inf. Theory, vol· 51, no 3, pp. 1209-1214, Mar. 2005, X. Zhang . and S. Wang, “Efficient steganographic embedding by exploiting modification direction,” IEEE Commun. Letter, vol. 10, no. 11, pp. 78 783 , Nov. 2006, and W. Zhang, S. Wang, and X. Zhang, "Improving Embedding efficiency of covering codes for applications in steganography," IEEE Commun. Letter, vol. 11, no. 8, pp. 680-682 , Aug. 2007, is trying to use the plus/minus 1 strategy (or the ±i method for short) to develop hidden information capacity for overlay images; .\^1^1115 311 (1 肘 · Dijk, ^Capacity and codes for embedding information in gray-scale signals", IEEE Trans. Inf. Theory, vol. 51, no. 3, pp. 1209-1214, Mar. 2005, and X. Zhang and S. Wang, "Efficient steganographic embedding by exploiting modification direction," IEEE Commun. Letter, vol. 10, no. 11, pp. 781-783, Nov. 2006, using Hamming coding ( Hamming Code) method to hide ternary information into the coverage pixel 'however, this method needs to convert binary secret information into ternary information first' has the burden of conversion; and W. Zhang, S. Wang, And X. Zhang, 44Improving Embedding efficiency of covering codes for applications in steganography,,, IEEE Commun. Letter, vol. 11, no. 8, pp. 680-682, Aug·2007, designed to convert methods can be directly hidden Binary secret information without first converting to ternary information, however, if the processed binary secret information is uncompressed and may contain high frequency repeated hidden blocks, then the same hidden processing operation It is repeated many times, so it is necessary to change a plurality of overlay pixels (c〇vering pixels). Based on the above problem, the inventors have proposed a direct information on how to reduce the number and value of the covered pixels to be changed and to reduce the number of hidden processing operations. The method of hiding is to overcome the drawbacks of the prior art. [Invention] The purpose of the present invention is to provide a reduction in the number of coverage pixels that are changed, 13/5190

Cj-i ^Cj^oj Cj +1, ^Cj ^25^y, 3·10 V, under other conditions

[Embodiment] Although the present invention has been explained using several preferred embodiments, the following, formula and specific embodiments are merely preferred embodiments of the present invention;

It is to be understood that the present invention is limited to the following drawings and specific embodiments. Hereinafter, embodiments of the method of the present invention will be specifically described. The matrix coding is used to hide a number of secret bits by a linear check code (linear parity code), which can hide a secret data block of an M-ary element. To the last significant bit (LSB) of the covered pixels, where Α

The secret data block of the bit is represented as ··, ~), and the "covering pixels can be C = (C|, C2,...'c„) indicating 'and the relationship with 々 and the radius of coverage ( Covering radius) is set to /? 'The coverage radius is the number of maximum possible changes used to limit the conversion code. The parity check matrix (indicated by ") can be used to implement a covering function to represent and achieve the purpose of matrix coding: K=(//xC;T)T@·?; I', "and ten" in the equation represent the symmetry vector (Syndr〇me vector) and the logical operation of the transition 9 i (transpose) and the mutual exclusion or (exciusive-〇R). Let a good representation of a parity check matrix with g columns and a rows, and the parity check matrix elements //(,, y) are represented by 〜, and the 7th row vector of the parity check matrix " can be defined as ~." Aj7. If W is equal to / /;, where 丨 “, then the last significant bit LSB of the overlay pixel will be modified to c@v where ^ represents ^ unit vector; 'will not change the last valid of any covered pixel Located in the LSB. For example, let 々:3,5 = (i〇i), C| = _, u, 〇, 〇, and the parity check matrix be set as follows: [10 0 11 Η ~ 〇1〇〇1 , 00111 1 0\ (1);

The characteristic to the banknote can be calculated as follows: Since Φ5: (4), 1) ten _=_); shows) the hidden last significant bit LSBs of the ^7-type pixel (with c, table C| ~ C Θ (丨, G, 〇, 0, 0, 0, 0) = (1,0,0,1,:,0,1). When the hidden information is to be retrieved, the hidden secret bit ^ can be counted as follows Different from C, the lieutenant is taken out:

S

Η X C is a continuous, repeated block of secret bits representing the last two significant bits of the w-covered pixels. The last two significant bits (four) of the resident pixels are. The hidden method of this method is developed to (4) directly hidden into c, and therefore if "equal to G, it means that the secret data will be s; otherwise, the secret data is ^ and can be repeated by the second. 1 and get. For example, 'make W, ^_, c+=(1_import), and the parity check matrix // as shown in formula (1); first set the flag bit to 1, then you can use formula (2) to calculate The eigenvector v = _, because ν' = (1 scale will be equal to the matrix row vector % ', so the secret data f can be directly hidden by the person covering the 9 in the pixel c+, and the hidden result can be obtained by the calculation of the compact function. That is, ci= £/(cP〇〇〇〇0011); where the bits ~ and ~ in the vector v are m(c+ 6,) results and 1. respectively. Please refer to the diagram to show that the smart direct information of the present invention is closely hidden. The preferred method of the law is the block diagram. In the general-type direct Zisaki secret concealment method (CCS), the value of the secret data y can only be a fixed positive integer; however, the appropriate value is high. Depending on the frequency of occurrence of the repeated block, please refer to Figure 2 again. The system shows the average ratio of the repeated blocks of all blocks or all i, and the total ratio of all blocks is 0 or all 1s. The two most frequently covered blocks appear in the job image. Therefore, the present invention will be presented below by the average ratio of total G and full 丨. - A smart close-in method (las) to tightly compress the full-length or all-one repeating block to cover the image, and increase the number of hidden blocks that can be compressed as much as possible. The hiding method uses a number of overlapping functions to repeat the number of blocks W value is + l, and then determines whether the secret data belongs to a hidden block S of all ones; ', and the number of repeated blocks w is 2^^2 (^ + 1 ; Step SA6: If the secret data belongs to the hidden block π of all 1, and the number of repeated blocks W is + 1, the flag bit = 1 is set, and the value of the pixel c, Set to / = (w-1) + 2 (four); Step SA7: If the secret data does not belong to the hidden block y of all 1, and the number of repeated blocks "value is 2^^2 (Μ + 1), then set the flag The locator 6/=0, and the eigenvector is calculated. (/^(:/ ten brothers; step SA8: judging whether the value of the overlay pixel C, / = 2λ; step SA9: if the value of the overlay pixel is Then, the feature step SA10 is calculated: if the value of the covered pixel c is not /= 2*, the feature vector F = (//xC/ten///' is calculated; step SA11: determining the turn of the feature vector r ./ matrix π is equal to the first row vector of the parity check matrix // // ,, i.e. determines whether

Vr = Η.; Step SA12: If κ7 = / /, the founder sets the last significant bit of the octet vector v, = l; Step SA13: If it is not true, then set the octet vector of 1375190 Bit a = 9 ' and covers the pixel. The value is determined by a random value from 1 to „; and step SA14: the last second significant bit of the 8-bit vector is calculated by a labeling function ¥(〇:+~) and a compact function £, (c , v) Calculate the pixel cy to be hidden from the modified pixel c,., that is, and c; = £:U, where ❿ from (C+,6,)=the mod 2, and, is| &gt;-;, 5 Bu 0) and Bu = 0) and (v2V, = 1 〇, 1 right ^ 255) and (~=1) and (¥1=11) ~®v, under other conditions} C 4* EAcj ^) - Referring to Figure 3' is a flow block diagram showing the secret information extracted from the overlay image of the present invention. When the secret 7 is to be retrieved from the overlay, the <received overlay pixel 4 is hidden and modified. The result is that the flag bit is - calculated by using the above formula (3). If the flag bit (4) ^ can be ": the secret data will be a single hidden area... and it can be obtained by = (call) simple; side,秘_料. By the single-block #I of #, and the basin 疋 is connected, and the repeat- owed: ,, the medium I value is obtained from the execution of the lower touch (1) if (, equal to ~, where Call, then w value will be r value; if salt is not greater than 2 (' The secret information is, the secret information is "&quot; and the value is (Chuan μ2Μ. (2) If (&quot;Yu Yi, the professional face is depreciated 17 and 2 Μ 'The secret of the poor can be repeated by Obtained by a single block deduction or s/. _ ? = 3,5〇=(0,0,0)5 ^ c+=(10^10^^01^^01) f 5 k-matrix // The above formula (1); first set the flag bit ~ for the feature vector can be calculated according to the formula (4), where: ΜΜ __ 'μ, = (10, 0); Ζ曰 equal to &quot;,, then the secret The data is "can be hidden directly into the C+ overlay i pixel £|, hiding the pixel pixel result as: C; = JMC丨, 0000 0011); where the bit vector V of the ~ bit can be calculated by calculation The function is obtained from (c+,~) and V, and the bit is set to 1. Therefore, after the aforementioned intelligent direct information hiding method of L, the method of extracting secret data from the cover shirt image can be summarized as the following steps: SB1: Define a single A-bit hidden data block, &amp;), the parity check matrix is /^, where „ = 2*_1, C+^cApC^Cw.,^”,)' indicates “overlay pixels” The last two have The bit 'and c; 2 means the value of the modified C + median C z' and the modified cover pixel c, and the result C is obtained; and cn means the cover pixel. The last valid bit; Step SB2: Enter a: value; 1375190 Step SB3: Calculate a flag bit %, where mod 2 ; \/=l ) Step SB4: Determine if the flag bit &amp; is equal to the value 1; Step SB5: If the flag bit % If the value is not equal to 1, the single female bit hides the block meaning 2 (Z/xC; 7 to capture the secret data; Step SB6: If the flag bit is equal to the value 1, then judge (// X c;7) = . Is it true, where / / indicates the j-th row vector of the parity check matrix //; Step SB7: If (//&gt;&lt;〇//, does not hold, the secret data belongs to All 1 hidden block β, and its number of repeated blocks "value = w = 2 (*_|) +1, in order to extract secret poor material, step SB8: If it is the founder, then determine whether the value of y is j&lt;2{k-]); Step SB9: If (4) is not a hill, the secret data belongs to the hidden block V of all 1, and the number of repeated blocks w is w = (y + i) - 2 (*-1) ) 'to take secret education , And step SB10: If the founders, to hide the confidential information belonging to a whole block β 0, and the number of repeating block "value w = y +1, to take the secret-lean operation. Therefore, by the above method, the hidden blocks are tightly compressed to all 0s or all 1s to cover the image, and the number of hidden blocks that can be compressed at one time is increased as much as possible to reduce the coverage pixels being changed. The number of operations is reduced and the number of operations that are hidden by 19 丄J/B) is handled, as well as the effect of direct tight concealment. Although the present invention has been explained in connection with the preferred embodiments, the present invention is not limited to the invention, and the field is not invented. Similar embodiments 'all of these are within the scope of the invention.

20 1375190 - [Simplified description of the drawings] Fig. 1 is a flow block diagram showing a preferred embodiment of the smart direct information close concealment method of the present invention; Fig. 2 is a diagram showing repeated blocks showing all 0s or all 1s. An average ratio comparison graph for all blocks; and Fig. 3 is a block diagram showing the flow of hidden secret information from the overlay image of the present invention. [Main component symbol description]

Claims (1)

1375190 If it is not established, the secret data belongs to the all-hidden block V, and the number of the repeated blocks is "= 〇· + 1)-2 (*-&quot; to capture the secret data. And if β2(4) is established, the secret data belongs to a hidden block β of all 0s, and the number of the repeated blocks is “= 7 + 1, to capture the secret data. 25