CN101872398B - Anti-collusion digital fingerprinting method based on code division multiple access and diversity technology - Google Patents

Abstract

The invention relates to an anti-collusion digital fingerprinting method based on a code division multiple access and diversity technology, comprising the following steps of: distributing a user identification code consisting of multiple stages of numbers for each purchaser; dividing an original carrier image into sub-images when fingerprints are embedded, then embedding all stages of numbers of a user into local DCT (Discrete Cosine Transformation) coefficients of the sub-images in all the sub-images by adopting a CDMA technology; and when finding a dubious copy, by combining with the original carrier image, firstly, detecting all stages of numbers and fingerprint strength values corresponding to the numbers of the dubious user in each sub-image, then synthesizing detection results of all the sub-images by adopting a diversity technology, and accurately judging one collusion member. The invention has less influence on the quality of the carrier image, low complexity of collusion detection and high detection precision rate.

Description

Anti-collusion digital fingerprinting method based on CDMA and diversity technique

Technical field

What the present invention relates to is a kind of method of technical field of information processing, specifically is a kind of based on CDMA (code divisionmultiple access, CDMA) anti-collusion digital fingerprinting method of He Fenji (diversity) technology.

Background technology

The scope that the fast development of infotech spreads through the internet works such as all kinds of literal, picture, video display unprecedentedly enlarges.How digital product is carried out copyright protection and become one of key problem of information age copyright protection.Some digital copyright managements (digital rights management has been proposed at present; DRM) scheme, this type scheme are based on cryptography means such as encryption, signatures, to guarantee the secure distribution of data; The illegal abuse of restricting data, and copyright proved purpose.But in case the user has obtained digital media information, then DRM disappears to the protection of information immediately, can't stop the user that the digital product that is obtained by legal means is carried out bootlegging and issue again.

Digital finger-print is a kind of technology of new Digital copyright protection, can remedy the deficiency of traditional DRM.Its core concept is: the product supplier adopts digital watermark technology in original copy, to embed and subscriber-related information before each part of issue copy, and this information has uniqueness, can not be by user's perception, and can distinguish different users effectively; In case the discovery illegal copies can be followed the trail of the user who scatters illegal copies through the finger print information that extracts in the copy.Yet some buyers that have identical multimedia messages and different fingerprints may join together fingerprint is carried out conspiracy attack, the copy that generates a finger print information decay even be removed.Therefore, design can be resisted the conspiracy behavior and differentiated that collaborator's fingerprint is a urgent demand.

Through existing literature search is found that correlation technique is following:

1, people such as Cox is at paper " Secure spread spectrum watermarking for multimedia (the multimedia spread-spectrum watermark of safety) " (IEEE transactions on image processing (IEEE Flame Image Process journal); Vol.6; No.12; Pp.1673-1687 has provided a kind of safe spread spectrum data waterprint embedded method in Dec.1997).This method uses the pseudo-random sequence of obeying standardized normal distribution as watermark information; Through picture being carried out overall discrete cosine transform (discrete cosine transform; DCT), choose the maximum a part of coefficient embed watermark of amplitude except that DC coefficient then.When carrying out watermark detection, extract watermark information by means of original image.Calculate the watermark information and the related coefficient of original watermark information extract then, thereby whether the judgement watermark information exists.Above-mentioned Cox watermark embedding method can be used in the digital fingerprint system.At first, distribute mutually orthogonal pseudo-random sequence as their fingerprint separately to different user; Adopt the Cox watermark embedding method then, fingerprint sequence is embedded in the picture respectively.In case the discovery illegal copies can be from conspiring to extract finger print information the copy, the related coefficient of calculated fingerprint information and each user's finger print information is confirmed the collaborator according to operation result then.This digital finger-print scheme can be resisted processing malice or non-malice such as lossy compression method, filtering, shearing, convergent-divergent well, and has the ability of stronger opposing multi-user conspiracy attack.But the fidelity of the picture that employing Cox embedding grammar obtains is lower, and the finger print information quantity that needs is directly proportional with number of users, and the computational complexity that detects the collaborator also is directly proportional with number of users.

2, people such as Trappe is at paper " Anti-collusion fingerprinting for multimedia (the anti-multimedia fingerprint of conspiring) " (IEEE transactions on signal processing (IEEE signal Processing journal); Vol.51; No.4; Pp.1069-1087 has proposed a kind of fingerprint schemes based on composite design in Apr.2003).This scheme by means of BIB DESIGN (Balancedincomplete block design, BIBD), with each binary element negate in the BIBD incidence matrix, each row of the matrix that obtains can be used as a user's fingerprint code; By preset rule fingerprint code is mapped to fingerprint to be embedded then; Fingerprint is embedded in the DCT coefficient of carrier.After finding illegal copies, from this copy, extract finger print information, adopt hard Threshold detection method, soft door limit detection method or sequential detection method to detect a plurality of collaborators.The random number sequence quantity of using in this fingerprint schemes is directly proportional with the square root of number of users, but when number of users is very big, asking for of corresponding BIBD incidence matrix will become very difficult.

3, people such as Wang is at paper " Group-oriented fingerprinting for multimedia forensics (the packet-based fingerprint that is used for the multimedia messages evidence obtaining) " (EURASIP journal on applied signal processing (the EURASIP application signal is handled periodical); Vol.2004; No.14; Pp.2153-2173 has proposed packet-based fingerprint schemes in Oct.2004).This scheme uses the pseudo-random sequence of obeying standardized normal distribution as finger print information.At first, distribute mutually orthogonal pseudo-random sequence as user profile to different user; Then, the user is divided into groups, the user who most possibly initiates conspiracy attack is each other divided in same group according to prior imformation; User in giving same group distributes the group information of same pseudo-random sequence as this group user, and on the same group group information is not mutually orthogonal; Group information and user profile addition are just obtained user's fingerprint; At last, fingerprint is embedded in the DCT coefficient of carrier.When detecting the collaborator, judge the group at collaborator place earlier through related operation, in group, judge the collaborator through related operation then.Compare the quadrature fingerprint, this algorithm can improve correct detection collaborator's probability, and the related operation number of times reduces; But employed pseudo-random sequence quantity still is directly proportional with total number of users.

4, people such as Naoki is at paper " Collusion-resistant fingerprinting scheme based on the CDMAtechnique (based on the anti-collusion digital fingerprinting scheme of CDMA) " (International Workshop on Security; Nara; Japan (information security international symposium in 2007), Oct.2007, LNCS; Vol.4752; Pp.28-43) proposed a kind of fingerprint schemes based on CDMA technology in, its method is: the user is divided into groups, and each user distributes a group # and Customs Assigned Number as its identification code; Picture is carried out overall dct transform, choose the DCT coefficient sequence that a part of medium and low frequency coefficient is formed two equal in length; Adopt CDMA technology that Customs Assigned Number and group # are embedded into respectively in two DCT coefficient sequence; All DCT coefficients are advanced the picture that the DCT inverse transformation obtains containing fingerprint.After finding suspicious copy, at first picture is carried out dct transform, select the DCT coefficient sequence of having carried user identification code; Adopt the CDMA technology information that takes the fingerprint in conjunction with original image, detect a plurality of collaborators.This scheme does not need related operation when detecting the collaborator, and detection complexity is low, but when the conspiracy number of users is big, can a lot of innocent persons be judged to be the collaborator.

In sum, present digital fingerprinting method still can not be in the finger print information amount, conspire detection complexity and collaborator and detect and reach gratifying effect simultaneously aspect three of the accuracy.

Summary of the invention

The objective of the invention is to overcome the above-mentioned deficiency of prior art, a kind of anti-collusion digital fingerprinting method based on CDMA and diversity technique is provided.The present invention embeds digital finger-print in the medium and low frequency coefficient of dct transform domain, guarantee that finger print information has very strong opposing normal signal ability of processing and well disguised; The process that embeds fingerprint adopts CDMA technology, can distinguish different user effectively, opposing multi-user conspiracy attack; Diversity technique is dissolved in the total system, in each piece, embedding identical fingerprint behind the carrier picture piecemeal, after obtaining suspicious copy; Testing result through comprehensive each piece; Only a people is judged to be the collaborator, has reduced the probability that the innocent person is judged to be the collaborator dramatically, the finger print information amount that this method needs seldom; The copy visual effect that generates is good, and detection collaborator's complexity is very low.

The present invention realizes through following technical scheme, the present invention includes following steps:

The first step generates buyer's fingerprint: distribute a unique user identification code for the buyer of each digital product.

Described user identification code is a H level numbering (s ₁, s ₂..., s _H), H>=2,1≤s _i(1≤i≤H), L is the maximum occurrences of every grade of numbering in the user identification code to≤L.

In second step, embed buyer's fingerprint: before the issue digital product, adopt fingerprint embedding method, generate the copy that contains user identification code with in the user identification code embedded product original copy.

Described fingerprint embedding method may further comprise the steps:

2.1) image division that needs are carried out copyright protection is that D opens equal-sized sub-pictures, respectively each sub-pictures carried out overall dct transform, obtains the overall dct transform coefficient of each sub-pictures;

2.2) from each sub-pictures, select H original DCT coefficient sequence that length is L of HL medium and low frequency overall situation dct transform coefficient composition respectively, H the original DCT coefficient sequence note of selecting in the individual sub-picture of k made v _{K, i}={ v _{K, i}(1), v _{K, i}(2) ..., v _{K, i}(L) }, wherein: 1≤k≤D, 1≤i≤H, H are the numbering progression of user identification code, and L is the maximum occurrences of every grade of numbering in the user identification code;

2.3) i length adopting CDMA technology in k sub-picture, to select respectively is that the i level that embeds in the user identification code in the original DCT coefficient sequence of L is numbered s _i, obtain containing the DCT coefficient sequence v ' of finger print information _{K, i}=v ' _{K, i}(1), v ' _{K, i}(2) ..., v ' _{K, i}(L) };

I the length that described employing CDMA technology is selected in k sub-picture respectively is that the i level that embeds in the user identification code in the original DCT coefficient sequence of L is numbered s _i, may further comprise the steps:

2.3.1) the use key K _iGenerating length equals L and element and is ± 1 binary pseudo-random sequence

$\mathrm{PN}\left(K_i\right)=\{r_{K_i}\left(1\right),r_{K_i}\left(2\right),...,r_{K_i}\left(L\right)\};$

Described key K _iBe meant:

$K_i=\left\{\begin{array}{c}{s}_C,i=1\\ s_{i-1},i\≥2\end{array}\right.,$

Wherein: 1≤s _C≤L, s _CBe the definite value of presetting, and each user's s _CAll equate s _CTo user cipher device.

Described binary pseudo-random sequence is through being that 0 of 0 and 1 pseudo-random sequence is mapped as-1 and obtains with element.

Described element is that 0 and 1 pseudo-random sequence is the m sequence, or the M sequence, or the Gold sequence.

2.3.2) with sequence PN (K _i) and sequence v _{K, i}The element of correspondence position multiplies each other, and the sequence that obtains multiplying each other is then carried out the one dimension dct transform, obtains w _{K, i}={ w _{K, i}(1), w _{K, i}(2) ..., w _{K, i}(L) };

2.3.3) to w _{K, i}S _iIndividual element adds fingerprint intensity level α _i, all the other elements are constant, obtain w ' _{K, i}=w ' _{K, i}(1), w ' _{K, i}(2) ... w ' _{K, i}(L) };

2.3.4) to w ' _{K, i}Carry out one dimension DCT inverse transformation, sequence and sequence PN (K that inverse transformation is obtained _i) element of correspondence position multiplies each other, and obtains containing the DCT coefficient sequence v ' of fingerprint _{K, i}=v ' _{K, i}(1), v ' _{K, i}(2) ..., v ' _{K, i}(L) }.

2.4) use each sub-pictures to contain the DCT coefficient sequence v ' of finger print information respectively _{K, i}Replace original DCT coefficient sequence v _{K, i}, and to the replacement after each sub-pictures carry out overall DCT inverse transformation, obtain containing the sub-pictures of fingerprint;

2.5) all sub-pictures that contain fingerprint are stitched together according to its order corresponding to the atom picture, obtain containing the image of fingerprint.

In the 3rd step, the collaborator detects: after the pirate copies of finding digital product, adopt collaborator's detection method, accurately obtain one of them collaborator.

Described collaborator's detection method may further comprise the steps:

3.1) be D equal-sized sub-pictures with the image division of pirate digital product, respectively each sub-pictures is carried out overall dct transform, obtain the overall dct transform coefficient of each sub-pictures;

3.2) from the overall dct transform coefficient of each sub-pictures, extract the overall dct transform coefficient sequence of having carried all grades of collaborator number information respectively, the overall dct transform coefficient sequence note of selecting in the individual sub-picture of k that carries i level numbering is made v ^* _{K, i}={ v ^* _{K, i}(1), v ^* _{K, i}(2) ..., v ^* _{K, i}(L) }, v wherein ^* _{K, i}(j) at the position and the v of the overall dct transform coefficient of k pirate sub-pictures _{K, i}(i) position of the overall dct transform coefficient in the original sub picture is identical, according to d _{K, i}(j)=v ^* _{K, i}(j)-v _{K, i}(j), the i level numbering that obtains k sub-picture detects sequence d _{K, i}={ d _{K, i}(1), d _{K, i}(2) ..., d _{K, i}(L) }, 1≤i≤H, 1≤k≤D;

3.3) respectively each sub-pictures is carried out the detection of first order numbering, obtain the 1st grade of numbering of suspicious user and the fingerprint intensity level of the 1st grade of numbering;

Described first order numbering detects, and may further comprise the steps:

3.3.1) use key s _CGenerating length equals L and element and is ± 1 binary pseudo-random sequence

$\mathrm{PN}\left(s_C\right)=\{r_{s_C}\left(1\right),r_{s_C}\left(2\right),...,r_{s_C}\left(L\right)\};$

3.3.2) with PN (s _C) and d _{K, 1}Element on the correspondence position multiplies each other, and then the sequence that obtains is carried out the one dimension dct transform, obtains the fingerprint intensity level f of the 1st grade of numbering _{K, 1}={ f _{K, 1}(1), f _{K, 1}(2) ..., f _{K, 1}(L) };

3.3.3) according to following formula setting threshold T _{K, 1}:

T _k，1＝σ _k，1Q ^-1(q _k，1)，

Wherein: ${\mathrm{\σ}}_{k,1}=\sqrt{\frac{1}{p_{k,1}}\underset{m\∈S_{k,1}}{\mathrm{\Σ}}{(f_{k,1}\left(m\right)-\frac{1}{p_{k,1}}\underset{m\∈S_{k,1}}{\mathrm{\Σ}}{f}_{k,1}\left(m\right))}^2},$

$Q\left(x\right)=\frac{1}{\sqrt{2\mathrm{\π}}}{\∫}_x^{+\∞}\exp (-\frac{t^2}{2})\mathrm{dt},$

S _k，1＝{m|f _min，k，1≤f _k，1(m)≤-f _min，k，1}，

f _{Min, k, 1}Be f _{K, 1}In the minimum value of each element, Q ^-1(x) be the inverse function of Q (x), q _{K, 1}Be the parameter of control threshold size, p _{K, 1}The expression S set _{K, 1}The number of element;

3.3.4) satisfy f _{K, 1}(m)＞T _{K, 1}N _{K, 1}Individual m forms the 1st grade of numbered sequence c of suspicious user _{K, 1}={ c _{K, 1}(1), c _{K, 1}(2) ..., c _{K, 1}(n _{K, 1}), the 1st grade of corresponding numbering fingerprint intensity level is f successively _{K, 1}(c _{K, 1}(1)), f _{K, 1}(c _{K, 1}(2)) ..., f _{K, 1}(c _{K, 1}(n _{K, 1})).

3.4) respectively each sub-pictures is carried out second level numbering and detects, obtain can the user the 2nd grade of numbering and the 2nd grade of numbering fingerprint intensity level.

Described second level numbering detects, and may further comprise the steps:

3.4.1) use key c _{K, 1}(j ₁) generate length and equal L and element and be ± 1 binary pseudo-random sequence

$\mathrm{PN}\left(c_{k,1}\left(j_1\right)\right)=\{r_{c_{k,1}\left(j_1\right)}\left(1\right),r_{c_{k,1}\left(j_1\right)}\left(2\right),...,r_{c_{k,1}\left(j_1\right)}\left(L\right)\};$

3.4.2) with PN (c _{K, 1}(j ₁)) and d _{K, 2}Element on the correspondence position multiplies each other, and then the sequence that obtains is carried out the one dimension dct transform, obtains the 1st grade of numbering and equals c _{K, 1}The 2nd grade of numbering fingerprint intensity level of suspicious user (j1)

$f_{k,2}^{\left(j_1\right)}=\{f_{k,2}^{\left(j_1\right)}\left(1\right),f_{k,2}^{\left(j_1\right)}\left(2\right),...,f_{k,2}^{\left(j_1\right)}\left(L\right)\};$

3.4.3) according to following formula setting threshold

$T_{k,2}^{\left(j_1\right)}={\mathrm{\σ}}_{k,2}^{\left(j_1\right)}{Q}^{-1}\left(q_{k,2}\right),$

Wherein: ${\mathrm{\σ}}_{k,2}^{\left(j_1\right)}=\sqrt{\frac{1}{p_{k,2}^{\left(j_1\right)}}\underset{m\∈S_{k,2}^{\left(j_1\right)}}{\mathrm{\Σ}}{(f_{k,2}^{\left(j_1\right)}\left(m\right)-\frac{1}{p_{k,2}^{\left(j_1\right)}}\underset{m\∈S_{k,2}^{\left(j_1\right)}}{\mathrm{\Σ}}{f}_{k,2}^{\left(j_1\right)}\left(m\right))}^2},$

$S_{k,2}^{\left(j_1\right)}=\left\{m\right|f_{\min ,k,2}^{\left(j_1\right)}\≤f_{k,2}^{\left(j_1\right)}\left(m\right)\≤-f_{\min ,k,2}^{\left(j_1\right)}\},$

Be

In the minimum value of each element,

The expression set

The number of element, q _{K, 2}It is the parameter of control threshold size;

3.4.4) satisfy

Individual m forms the 1st grade of numbering and equals c _{K, 1}(j ₁) the 2nd grade of numbered sequence of suspicious user

Accordingly

Individual the 2nd grade of numbering fingerprint intensity level is successively

3.5) according to 3.4) and method; Each sub-pictures is carried out i level numbering to be detected; Obtain every grade of numbering of each sub-pictures and detect the fingerprint intensity level that the i level is numbered and the i level is numbered that detects suspicious user in the sequence; Till detecting all H level numberings and corresponding fingerprint intensity level, 3≤i≤H.

Described i level numbering detects, and may further comprise the steps:

3.5.1) using the key

generate length equal to L and the element is ± 1 binary pseudo-random sequence

3.5.2) will

And d _{K, i}Element on the correspondence position multiplies each other, and then the sequence that obtains is carried out the one dimension dct transform, obtains the 1st grade of numbering and equals c _{K, 1}(j ₁), the 2nd grade of numbering equals

The 3rd level numbering equals

Equal with i-1 level numbering

The i level numbering fingerprint intensity level of suspicious user

3.5.3) according to following formula setting threshold

$T_{k,i}^{(j_1,j_2,...,j_{i-1})}={\mathrm{\σ}}_{k,i}^{(j_1,j_2,...,j_{i-1})}{Q}^{-1}\left(q_{k,i}\right),$

Wherein:

${\mathrm{\σ}}_{k,i}^{(j_1,j_2,...,j_{i-1})}=\sqrt{\frac{1}{p_{k,i}^{(j_1,j_2,...,j_{i-1})}}\underset{m\∈S_{k,i}^{(j_1,j_2,...j_{i-1})}}{\mathrm{\Σ}}{(f_{k,i}^{(j_1,j_2,...,j_{i-1})}\left(m\right)-\frac{1}{p_{k,i}^{(j_1,j_2,j_{i-1})}}\underset{m\∈S_{k,i}^{(j_1,j_2,j_{i-1})}}{\mathrm{\Σ}}{f}_{k,i}^{(j_1,j_2,...,j_{i-1})}\left(m\right))}^2}$

$S_{k,i}^{(j_1,j_2,...,j_{i-1})}=\left\{m\right|f_{\min ,k,i}^{(j_1,j_2,...,j_{i-1})}\≤f_{k,i}^{(j_1,j_2,...,j_{i-1})}\left(m\right)\≤-f_{\min ,k,i}^{(j_1,j_2,...,j_{i-1})}\}$

Be

In the minimum value of each element,

The expression set

The number of element, q _{K, i}It is the parameter of control threshold size.

3.5.4) satisfy

Individual m forms the 1st grade of numbering and equals c _{K, 1}(j ₁), the 2nd grade of numbering equals

... and i-1 level numbering equals

The i level numbered sequence of suspicious user

Accordingly

The fingerprint intensity level of individual i level numbering equals successively

3.6) respectively every grade of fingerprint intensity level in each sub-pictures is carried out addition, obtain the fingerprint intensity sums at different levels of detected suspicious user in each sub-pictures, concrete formula is:

${\mathrm{\β}}_k(j_1,j_2,...,j_{H-1},j_H)=f_{k,1}\left(j_1\right)+f_{k,2}^{\left(j_1\right)}\left(j_2\right)+f_{k,3}^{(j_1,j_2)}\left(j_3\right)+...+f_{k,H}^{(j_1,j_2,...,j_{H-1})}\left(j_H\right);$

3.7) when there being and only existing k ₁Individual, k ₂Individual ... with k _NIndividual sub-picture makes

Then with the fingerprint intensity at different levels of these sub-pictures with carry out addition, obtain the fingerprint intensity level summation of each suspicious user, the maximum user of fingerprint intensity level summation is confirmed as the collaborator.

Compare with existing anti-collusion digital fingerprinting method, the present invention has following advantage:

First; For distinguishing the Customs Assigned Number of different user effectively; Only one group of mutual quasi-orthogonal element of needs is ± 1 pseudo-random sequence; And these can be that 0 in 0 and 1 the pseudo-random sequence (like m sequence, M sequence, Gold sequence etc.) is mapped as-1 and obtains through some elements for random series, and actual data volume is minimum, has avoided the huge problem of bringing as fingerprint with the white Gaussian noise sequence of finger print information data volume.

The second, adopt diversity technique to improve the robustness of fingerprint.Because the carrier picture is divided, be embedded in fingerprint at each sub-pictures, so when detecting the collaborator, can from a plurality of sub-pictures, detect suspicious user at built-in end.The testing result of last comprehensive each sub-pictures is only selected a most possible disabled user who participates in conspiring, and has improved the accuracy that the collaborator detects, and has reduced the probability of faults.

The 3rd, the process that the collaborator detects comprise picture the piecemeal dct transform, contain the extraction of fingerprint DCT coefficient sequence, pseudo-random sequence and contain wise multiplication, one dimension dct transform and the last judgement of fingerprint DCT coefficient sequence.Can find that whole process need not carried out related operation, detection complexity is very low.If the power that the length of the pseudo-random sequence of selecting for use equals 2, the one dimension dct transform can also use fast algorithm to carry out so, with further quickening testing process.

The 4th, this method has good extendability.ID is the combination of multiple stages of numbers, in each sub-pictures, chooses a plurality of overall DCT coefficient sequence and embeds these numberings.Adopt public key to generate the pseudo-random sequence except embedding first order numbering, adopt the upper level numbering to generate and embed the required pseudo-random sequence of next stage numbering as key.The progression of the numbering that can confirm according to actual needs to be taked.

Description of drawings

The former figure of Lena that Fig. 1 adopts for embodiment.

Fig. 2 is an examples of implementation picture dividing mode synoptic diagram.

Fig. 3 contains fingerprint Lena picture for 4 users' among the embodiment;

Wherein: (a) be first user contain fingerprint Lena picture; (b) be second user contain fingerprint Lena picture; (c) be third party contain fingerprint Lena picture; (d) be the 4th user contain fingerprint Lena picture.

Fig. 4 is the illegal Lena picture that the collaborator generates among the embodiment.

Embodiment

Below in conjunction with accompanying drawing embodiments of the invention are elaborated: present embodiment provided detailed embodiment and process, but protection scope of the present invention is not limited to following embodiment being to implement under the prerequisite with technical scheme of the present invention.

Embodiment

Present embodiment adopt size be 512 * 512 Lena gray scale picture as initial carrier, as shown in Figure 1, specifically may further comprise the steps:

1) generates buyer's fingerprint: distribute a unique user identification code for the buyer of each digital product.

User identification code is a two-stage numbering (s in the present embodiment ₁, s ₂), and 1≤s ₁, s ₂≤63, so the number of users that this system of fingerprints can hold equals 63 ²=3969.

2) embed buyer's fingerprint: before the issue digital product, adopt fingerprint embedding method, generate the copy that contains user identification code with in the user identification code embedded product original copy.

2.1) image division that needs are carried out copyright protection is 4 equal-sized sub-pictures; The size of each sub-pictures is 256 * 256; The sequence number of 4 sub-pictures is as shown in Figure 2, respectively each sub-pictures is carried out overall dct transform, obtains the overall dct transform coefficient of each sub-pictures;

2.2) from each sub-pictures, select H original DCT coefficient sequence that length is L of HL medium and low frequency overall situation dct transform coefficient composition respectively, H the original DCT coefficient sequence note of selecting in the individual sub-picture of k made v _{K, i}={ v _{K, i}(1), v _{K, i}(2) ..., v _{K, i}(L) }, wherein: 1≤k≤D, 1≤i≤H, H are the numbering progression of user identification code, and L is the maximum occurrences of every grade of numbering in the user identification code;

From the overall dct transform coefficient of the individual sub-picture of k (1≤k≤4), selecting 2 * 63 medium and low frequency coefficients, to form two length be 63 sequence v _{K, 1}={ v _{K, 1}(1), v _{K, 1}(2) ..., v _{K, 1}And v (63) } _{K, 2}={ v _{K, 2}(1), v _{K, 2}(2) ..., v _{K, 2}(63) }, be respectively applied for the 1st grade of numbering of embedding s ₁With the 2nd grade of numbering numbering s ₂V wherein _{K, 1}(j) the overall dct transform coefficient that is positioned at the capable 192-j row of 65+j of k sub-picture, v are taken from (1≤j≤63) _{K, 2}(j) the overall dct transform coefficient that is positioned at the capable 129-j row of 128+j of k sub-picture is taken from (1≤j≤63).

2.3) i length adopting CDMA technology in k sub-picture, to select respectively is that the i level that embeds in the user identification code in the original DCT coefficient sequence of L is numbered s _i, obtain containing the DCT coefficient sequence v ' of finger print information _{K, i}=v ' _{K, i}(1), v ' _{K, i}(2) ..., v ' _{K, i}(L) }, wherein: 1≤i≤2,1≤k≤4:

2.3.1) the use key K _iGenerate length equal 63 and element be ± 1 binary pseudo-random sequence

Described key K _iSpecifically: K ₁=1, K ₂=s ₁

For obtain length equal 63 and element be ± 1 binary pseudo-random sequence, at first adopt primitive polynomial x ⁶+ x+1 generate length be 63 and element be 0 and 1 m sequence, be mapped to-1 with 0 again, obtain an element and be ± 1 binary pseudo-random sequence a={a (1), a (2) ..., a (63).A given key K (1≤K≤63), the pseudo-random sequence of generation are PN (K)={ r _K(1), r _K(2) ..., r _K(63) }, r wherein _K(i)=a (mod (i+K-1,63)).

2.3.2) with PN (K _i) and v _{K, i}The element of correspondence position multiplies each other, and then the sequence that obtains is carried out the one dimension dct transform, obtains w _{K, i}={ w _{K, i}(1), w _{K, i}(2) ..., w _{K, i}(63) };

2.3.3) to w _{K, i}S _iIndividual element adds fingerprint intensity level α _i=500, all the other elements are constant, obtain w ' _{K, i}=w ' _{K, i}(1), w ' _{K, i}(2) ... w ' _{K, i}(63) };

2.3.4) to w ' _{K, i}Carry out one dimension DCT inverse transformation, with sequence that obtains and PN (K _i) element of correspondence position multiplies each other, and obtains containing the DCT coefficient sequence v ' of fingerprint _{K, i}=v ' _{K, i}(1), v ' _{K, i}(2) ..., v ' _{K, i}(63) }.

2.4) use each sub-pictures to contain the DCT coefficient sequence v ' of finger print information respectively _{K, i}Replace original DCT coefficient sequence v _{K, i}, and to the replacement after each sub-pictures carry out overall DCT inverse transformation, obtain containing the sub-pictures of fingerprint;

2.5) all sub-pictures that contain fingerprint are stitched together according to its order corresponding to the atom picture, obtain containing the image of fingerprint.

Use user identification code (15,44) in the present embodiment, (17; 27); What (35,5) and (44,30) generated contains the fingerprint copy respectively shown in Fig. 3 (a), Fig. 3 (b), Fig. 3 (c) and Fig. 3 (d); The PSNR of 4 parts of copies (Y-PSNR) is followed successively by 39.2571dB, 39.2574dB, 39.2575dB and 39.2539dB, and the fidelity of visible picture is fine.

The computing method of PSNR are in the present embodiment:

$\mathrm{PSNR}={10\log }_{10}\frac{{255}^2}{\frac{1}{{512}^2}\underset{m=1}{\overset{512}{\mathrm{\Σ}}}\underset{n=1}{\overset{512}{\mathrm{\Σ}}}{[g(m,n)-h(m,n)]}^2},$

Wherein g (m, n) and h (m n) representes original image respectively and contains the gray-scale value of fingerprint picture.

3) collaborator detects: after the pirate copies of finding digital product, adopt collaborator's detection method, accurately obtain one of them collaborator.

Above-mentioned four users conspire in the present embodiment; They average the respective pixel value of the gray scale picture that obtains, and obtain a new picture, and then picture being carried out quality factor is that 75 JPEG compression obtains conspiring copy; As shown in Figure 4, carry out illegal distribution at last.The copyright owner detects the collaborator through following steps behind the copy that has obtained illegal distribution:

3.1) be 4 equal-sized sub-pictures with the image division of pirate digital product, the dividing mode of dividing mode when embedding fingerprint is consistent, respectively each sub-pictures is carried out overall dct transform, obtains the overall dct transform coefficient of each sub-pictures;

3.2) from the overall dct transform coefficient of each sub-pictures, select and carried all overall dct transform coefficient sequences of number information at the same level not of collaborator, the overall dct transform coefficient sequence note of selecting in the individual sub-picture of k (1≤k≤4) that carries i (1≤i≤2) level numbering is made v ^* _{K, i}={ v ^* _{K, i}(1), v ^* _{K, i}(2) ..., v ^* _{K, i}(63) }, v wherein ^* _{K, i}(j) at the position and the v of the overall dct transform coefficient of k pirate sub-pictures _{K, i}(i) position of the overall dct transform coefficient in the original sub picture is identical, according to d _{K, i}(j)=v ^* _{K, i}(j)-v _{K, i}(j) (1≤j≤63), the i level numbering that obtains the individual sub-picture of k (1≤k≤4) detects sequence d _{K, i}={ d _{K, i}(1), d _{K, i}(2) ..., d _{K, i}(63) }.

3.3) detect sequence d from the 1st grade of numbering of the individual sub-picture of k (1≤k≤4) _{K, 1}Middle the 1st grade of numbering that detects suspicious user obtained n _{1, k}The 1st grade of numbered sequence c that individual element is formed _{K, 1}={ c _{K, 1}(1), c _{K, 1}(2) ..., c _{K, 1}(n _{K, 1}) and the 1st grade of corresponding numbering fingerprint intensity level f _{K, 1}(c _{K, 1}(1)), f _{K, 1}(c _{K, 1}(2)) ..., f _{K, 1}(c _{K, 1}(n _{K, 1})):

3.3.1) use key s _C=1 generate length equal 63 and element be ± 1 binary pseudo-random sequence

3.3.2) with PN (s _C) and d _{K, 1}Element on the correspondence position multiplies each other, and then the sequence that obtains is carried out the one dimension dct transform, obtains the fingerprint intensity level f of the 1st grade of numbering _{K, 1}={ f _{K, 1}(1), f _{K, 1}(2) ..., f _{K, 1}(63) };

3.3.3) establish f _{K, 1}In the minimum value of each element be f _{Min, k, 1}, S set _{K, 1}={ m|f _{Min, k, 1}≤f _{K, 1}(m)≤-f _{Min, k, 1}, calculate P wherein _{K, 1}The expression S set _{K, 1}The number of element;

3.3.4) setting threshold T _{K, 1}=σ _{K, 1}Q ^-1(q _{K, 1}), q wherein _{K, 1}=0.005.Satisfy f _{K, 1}(m)＞T _{K, 1}N _{K, 1}Individual m forms the 1st grade of numbered sequence c of suspicious user _{K, 1}={ c _{K, 1}(1), c _{K, 1}(2) ..., c _{K, 1}(n _{K, 1}), the 1st grade of corresponding numbering fingerprint intensity level is f successively _{K, 1}(c _{K, 1}(1)), f _{K, 1}(c _{K, 1}(2)) ..., f _{K, 1}(c _{K, 1}(n _{K, 1})).

After all 4 sub-pictures all traveled through completion, the 1st grade of numbering of suspicious user detected and finishes, and obtained corresponding the 1st grade of numbering fingerprint intensity level.

3.4) according to the 1st grade of suspicious user numbering c _{K, 1}(j ₁) (1≤j ₁≤n _{K, 1}), the 2nd grade of numbering from the individual sub-picture of k (1≤k≤4) detects sequence d _{K, 2}Middle the 2nd grade of numbering that detects suspicious user, what obtain has

The sequence of individual element

And obtain corresponding The fingerprint intensity level of individual i level numbering

3.4.1) use key c _{K, 1}(j ₁) generate length equal 63 and element be ± 1 binary pseudo-random sequence

3.4.2) with PN (c _{K, 1}(j ₁)) and d _{K, 2}Element on the correspondence position multiplies each other, and then the sequence that obtains is carried out the one dimension dct transform, obtains the 1st grade of numbering and equals c _{K, 1}(j ₁) the 2nd grade of potential suspicious user numbering fingerprint intensity level

3.4.3) Let

the minimum value for each element

collection

calculate where

represents a collection

number of elements;

3.4.4) setting threshold

Q wherein _{K, 2}=0.005.Satisfy

Individual m forms the 1st grade of numbering and equals c _{K, 1}(j ₁) the 2nd grade of numbered sequence of suspicious user Accordingly

Individual the 2nd grade of numbering fingerprint intensity level is successively

As all c _{K, 1}(j ₁) (1≤k≤4) (1≤j ₁≤n _{K, 1}) all travel through accomplish after, the 2nd grade of numbering of suspicious user detects and finishes, and obtains corresponding the 2nd grade of numbering fingerprint intensity level.

3.5) according to the numberings at different levels and the corresponding fingerprint intensity level of numbering at different levels of detected suspicious user, calculate the fingerprint intensity level sums at different levels of detected suspicious user in each sub-pictures.Be specially: in k the sub-picture, the 1st grade of numbering equals c _{K, 1}(j ₁) and the 2nd grade of numbering equal

2 grades of fingerprint intensity level sums of suspicious user equal

3.6) 2 grades of fingerprint intensity level sums of detected same suspicious user in the different sub that the adds up picture, obtaining the fingerprint intensity level summation of each suspicious user, the maximum user of fingerprint intensity level summation is confirmed as the collaborator.Specific as follows:

Equal c when only in k sub-picture, having detected the 1st grade of numbering _{K, 1}(j ₁) and the 2nd grade of numbering equal

Suspicious user, the fingerprint intensity level summation of this suspicious user equals so:

When there being and only existing k ₁Individual and k ₂Individual sub-picture makes And The time, the fingerprint intensity level summation of this suspicious user equals so:

$\mathrm{\γ}(c_{k,1}\left(j_{\mathrm{1,1}}\right),c_{k_2,2}^{\left(j_{\mathrm{1,1}}\right)}\left(j_{\mathrm{1,2}}\right))={\mathrm{\β}}_{k_1}(j_{\mathrm{1,1}},j_{\mathrm{1,2}})+{\mathrm{\β}}_{k_2}(j_{\mathrm{2,1}},j_{\mathrm{2,2}}).$

When there being and only existing k ₁Individual, k ₂Individual ... with k _NIndividual sub-picture makes

The time, the fingerprint intensity level summation of this suspicious user equals so:

$\mathrm{\γ}(c_{k_1,1}\left(j_{\mathrm{1,1}}\right),c_{k_2,2}^{\left(j_{\mathrm{1,1}}\right)}\left(j_{\mathrm{1,2}}\right))={\mathrm{\β}}_{k_1}(j_{\mathrm{1,1}},j_{\mathrm{1,2}})+{\mathrm{\β}}_{k_2}(j_{\mathrm{2,1}},j_{\mathrm{2,2}})+...+{\mathrm{\β}}_{k_N}(j_{N,1},j_{N,2}).$

In having traveled through all sub-pictures, behind the detected suspicious user, obtain the suspicious user fingerprint intensity level summation table shown in the table 1.Because γ (17,27) maximum, user identification code is that the buyer of (17,27) is confirmed as the collaborator so, and this user is one of collaborator really, thereby the detection of present embodiment method is correct.

Table 1

(s 1，s 2)	(15，44)	(17，27)	(35，5)	(43，46)	(44，30)
						γ(s 1，s 2)	231.0729	238.0529	230.1223	38.7108	231.0438
(s 1，s 2)	(17，44)	(44，40)	(15，47)	(17，36)
						γ(s 1，s 2)	60.3237	63.9258	30.4802	29.0907

Claims (8)

1. the anti-collusion digital fingerprinting method based on CDMA and diversity technique is characterized in that, may further comprise the steps:

The first step generates buyer's fingerprint: distribute a unique user identification code for the buyer of each digital product;

In second step, embed buyer's fingerprint: before the issue digital product, adopt fingerprint embedding method, generate the copy that contains user identification code with in the user identification code embedded product original copy;

In the 3rd step, the collaborator detects: after the pirate copies of finding digital product, adopt collaborator's detection method, accurately obtain one of them collaborator; Described fingerprint embedding method may further comprise the steps:

2.1) picture that needs are carried out copyright protection is divided into equal-sized D sub-picture, and each sub-pictures is carried out overall dct transform, obtains overall dct transform coefficient;

2.2) from each sub-pictures, select H DCT coefficient sequence that length is L of HL medium and low frequency overall situation dct transform coefficient composition respectively, H the original DCT coefficient sequence note of selecting in the individual sub-picture of k made v _{K, i}={ v _{K, i}(1), v _{K, i}(2) ..., v _{K, i}(L) } (1≤i≤H), wherein: 1≤k≤D, 1≤i≤H, H are the numbering progression of user identification code, and L is the maximum occurrences of every grade of numbering in the user identification code;

2.3) i length adopting CDMA technology in k sub-picture, to select respectively is that the i level that embeds in the user identification code in the original DCT coefficient sequence of L is numbered s _i, obtain containing the DCT coefficient sequence v ' of finger print information _{K, i}=v ' _{K, i}(1), v ' _{K, i}(2) ..., v ' _{K, i}(L) };

2.4) use each sub-pictures to contain the DCT coefficient sequence v ' of finger print information respectively _{K, i}Replace original DCT coefficient sequence v _{K, i}, and to the replacement after each sub-pictures carry out overall DCT inverse transformation, obtain containing the sub-pictures of fingerprint;

2.5) all sub-pictures that contain fingerprint are stitched together according to its order corresponding to the atom picture, obtain containing the image of fingerprint;

Described collaborator's detection method may further comprise the steps:

3.1) picture of pirate digital product is divided into D equal-sized sub-pictures, respectively each sub-pictures is carried out overall dct transform, obtain the overall dct transform coefficient of each sub-pictures;

3.2) from the overall dct transform coefficient of each sub-pictures, extract the overall dct transform coefficient sequence of having carried all grades of collaborator number information respectively, the overall dct transform coefficient sequence note of selecting in the individual sub-picture of k that carries i level numbering is made v ^* _{K, i}={ v ^* _{K, i}(1), v ^* _{K, i}(2) ..., v ^* _{K, i}(L) }, v wherein ^* _{K, i}(j) at the position and the v of the overall dct transform coefficient of k pirate sub-pictures _{K, i}(i) position of the overall dct transform coefficient in the original sub picture is identical, according to d _{K, i}(j)=v ^* _{K, i}(j)-v _{K, i}(j), the i level numbering that obtains k sub-picture detects sequence d _{K, i}={ d _{K, i}(1), d _{K, i}(2) ..., d _{K, i}(L) }, 1≤i≤H, 1≤k≤D;

3.3) respectively each sub-pictures is carried out the detection of first order numbering, obtain the 1st grade of numbering and the 1st grade of numbering fingerprint intensity level of suspicious user;

3.4) respectively each sub-pictures is carried out second level numbering detection, obtain the 2nd grade of numbering and the 2nd grade of numbering fingerprint intensity level of suspicious user;

3.5) according to 3.4) and method, each sub-pictures is carried out i level numbering detects, obtain the fingerprint intensity level of the i level numbering and the i level numbering of each sub-pictures suspicious user, till detecting all H levels numberings and fingerprint intensity level accordingly, 3≤i≤H;

3.6) respectively every grade of fingerprint intensity level in each sub-pictures is carried out addition, obtain the fingerprint intensity sums at different levels of detected suspicious user in each sub-pictures, concrete formula is:

3.7) testing result of comprehensive different sub picture, the fingerprint intensity level addition of the suspicious user that numbering is identical obtains the fingerprint intensity level summation of each suspicious user, and the maximum user of fingerprint intensity level summation is confirmed as the collaborator.

2. the anti-collusion digital fingerprinting method based on CDMA and diversity technique according to claim 1 is characterized in that, described user identification code is a H level numbering (s ₁, s ₂..., s _H), H>=2,1≤s _i≤L, 1≤i≤H, L are the maximum occurrences of every grade of numbering of user identification code.

3. the anti-collusion digital fingerprinting method based on CDMA and diversity technique according to claim 1; It is characterized in that i the length that described employing CDMA technology is selected respectively is that the i level that embeds in the user identification code in the original DCT coefficient sequence of L is numbered s in k sub-picture _i, may further comprise the steps:

2.3.1) the use key K _iGenerating length equals L and element and is ± 1 binary pseudo-random sequence

2.3.2) with sequence PN (K _i) and sequence v _{K, i}The element of correspondence position multiplies each other, and the sequence that obtains multiplying each other is then carried out the one dimension dct transform, obtains w _{K, i}={ w _{K, i}(1), w _{K, i}(2) ..., w _{K, i}(L) };

2.3.3) to w _{K, i}S _iIndividual element adds fingerprint intensity level α _i, all the other elements are constant, obtain w ' _{K, i}=w ' _{K, i}(1), w ' _{K, i}(2) ... w ' _{K, i}(L) };

2.3.4) to w ' _{K, i}Carry out one dimension DCT inverse transformation, sequence and sequence PN ((K that inverse transformation is obtained _i) element of correspondence position multiplies each other, and obtains containing the DCT coefficient sequence of fingerprint

v′ _k，i＝{v′ _k，i(1)，v′ _k，i(2)，...，v′ _k，i(L)}。

4. the anti-collusion digital fingerprinting method based on CDMA and diversity technique according to claim 3 is characterized in that, described key K _iBe meant:

Wherein: 1≤s _c≤L, s _cBe the definite value of presetting, and each user's s _cAll equate s _cTo user cipher device.

5. the anti-collusion digital fingerprinting method based on CDMA and diversity technique according to claim 3 is characterized in that, described binary pseudo-random sequence is through being that 0 of 0 and 1 pseudo-random sequence is mapped as-1 and obtains with element; Described element is that 0 and 1 pseudo-random sequence is the m sequence, or the M sequence, or the Gold sequence.

6. the anti-collusion digital fingerprinting method based on CDMA and diversity technique according to claim 1 is characterized in that, described first order numbering detects, and may further comprise the steps:

3.3.1) use key s _CGenerating length equals L and element and is ± 1 binary pseudo-random sequence

3.3.2) with PN (s _C) and d _{K, 1,}Element on the correspondence position multiplies each other, and then the sequence that obtains is carried out the one dimension dct transform, obtains the fingerprint intensity level f of the 1st grade of numbering _{K, 1}={ f _{K, 1}(1), f _{K, 1}(2) ..., f _{K, 1}(L) };

3.3.3) according to following formula setting threshold T _{K, 1}

T _k，1＝σ _k，1Q ^-1(q _k，1)

Wherein:

S _k，1＝{m|f _min，k，1≤f _k，1(m)≤-f _min，k，1}

f _{Min, k, 1}Be f _{K, 1}In the minimum value of each element, Q ^-1(x) be the inverse function of Q (x), q _{K, 1}Be the parameter of control threshold size, p _{K, 1}The expression S set _{K, 1}The number of element;

3.3.4) satisfy f _{K, 1}(m)＞T _{K, 1}N _{K, 1}Individual m forms the 1st grade of numbered sequence c of suspicious user _{K, 1}={ c _{K, 1}(1), c _{K, 1}(2) ..., c _{K, 1}(n _{K, 1}), the 1st grade of corresponding numbering fingerprint intensity level is f successively _{K, 1}(c _{K, 1}(1)), f _{K, 1}(c _{K, 1}(2)) ..., f _{K, 1}(c _{K, 1}(n _{K, 1})).

7. the anti-collusion digital fingerprinting method based on CDMA and diversity technique according to claim 1 is characterized in that, described second level numbering detects, and may further comprise the steps:

3.4.1) use key c _{K, 1}(j ₁) generate length and equal L and element and be ± 1 binary pseudo-random sequence

3.4.2) with PN (c _{K, 1}(j ₁)) and d _{K, 2}Element on the correspondence position multiplies each other, and then the sequence that obtains is carried out the one dimension dct transform, obtains the 1st grade of numbering and equals c _{K, 1}(j ₁) the 2nd grade of suspicious user numbering fingerprint intensity level

3.4.3) according to following formula setting threshold

Wherein:

Be

In the minimum value of each element,

The expression set

The number of element, q _{K, 2}It is the parameter of control threshold size;

3.4.4) satisfy

Individual m forms the 1st grade of numbering and equals c _{K, 1}(j ₁) the 2nd grade of numbered sequence of suspicious user Accordingly Individual the 2nd grade of numbering fingerprint intensity level is successively

8. the anti-collusion digital fingerprinting method based on CDMA and diversity technique according to claim 1 is characterized in that, described i level numbering detects, and may further comprise the steps:

3.5.1) using the key?

generated and the element of length equal to L ± 1 binary pseudo-random sequence

3.5.2) will

And d _{K, i}Element on the correspondence position multiplies each other, and then the sequence that obtains is carried out the one dimension dct transform, obtains the 1st grade of numbering and equals c _{K, 1}(j ₁), the 2nd grade of numbering equals The 3rd level numbering equals Equal with i-1 level numbering

The i level numbering fingerprint intensity level of suspicious user

3.5.3) according to following formula setting threshold

Wherein

Be

In the minimum value of each element,

The expression set

The number of element, q _{K, i}It is the parameter of control threshold size;

3.5.4) satisfy

Individual m forms the 1st grade of numbering and equals c _{K, 1}(j ₁), the 2nd grade of numbering equals

Equal with i-1 level numbering

The i level numbered sequence of suspicious user

Accordingly

The fingerprint intensity level of individual i level numbering equals successively

Images