CN103152608B - Realize the method for the Video Multicast transmission of digital finger-print encryption - Google Patents

Abstract

The invention discloses the method for the Video Multicast transmission realizing digital finger-print encryption, comprise digital finger-print encrypted transmission and digital finger-print and encrypt and receive, the former comprises the following steps: that the carrier information source data of image and video are compressed through Moving Picture Experts Group-2; To the digital fingerprint data based on image or video process, finally form digital fingerprint data the probability that collusion attack occurs mutually according to the different branches of tree structure determines the energy size that digital finger-print embeds and relation thereof; Combine quadrature modulation process through CDMA/TDMA and form blended data and form encrypting traffic, finally by transmission; The latter comprises the following steps: according to key K ^m, K ⁽ⁱ⁾deciphering, compression, Recovery image and video source data, meanwhile, carry out demodulation, and through process, relevance function detector tree generation collusion attack at different levels, finally obtains digital fingerprint data.Fail safe of the present invention is high, reduces communications cost, strengthens right and the validity thereof of the illegal person of system tracks simultaneously.

Description

Realize the method for the Video Multicast transmission of digital finger-print encryption

Technical field

The present invention relates to the technical field of information security, is the technology of the method for the Video Multicast transmission realizing digital finger-print encryption specifically.

Background technology

Digital finger-print video is the process Video service signal that user is given being embedded to the fingerprint Digital Signal Processing of different identification, utilizes the different meaning bits of appearance to identify between user.This mark, also referred to as fingerprint load pay-load, is normally realized by robust digital watermark telescopiny.In digital copyright management DRM text, the main purpose of fingerprint embedding and encryption is that protection video content does not disturb by single or collusion attack.Wherein, the collusion attack of digital video is that it applies the sixty-four dollar question faced, and therefore, the resistance against colluders ability of digital video is the important technology index that it can be applied.

Current Digital Fingerprinting Technology reaches the practical stage far away, this is because people ignore the research of watermarked security.Digital finger-print belongs to digital watermarking research category.For a long time, the robustness of watermark is more paid close attention in the research of digital watermarking, namely studies the attack of how to resist some non-malicious, as: additive noise, filtration, lossy compression method and resist geometric attacks etc., and ignore the research of watermarked security.Wait application for fingerprint recognition and copy control, the fail safe of watermark is a very important problem, and wherein, collusion attack is the most effectively attacked digital finger-print.In collusion attack, the new copy of generations one conspired by multiple copies of the same works that multiple conspirator has with them, in new copy, finger print information is removed or weakens, thus, digital finger-print not only requires that can resist general signal transacting attacks and single copy-attack, and can also resist collusion attack.

Actualizing technology mainly orthogonal fingerprint, the correlation fingerprint of current digital finger-print, also can be divided into continuous fingerprint and discrete fingerprint and collusion resistant code ACC digital finger-print.

(1) shortcoming of orthogonal fingerprint is: the fingerprint application few for user is more effective, but along with increasing of user, the complexity of fingerprint detection and tracing algorithm significantly increases, and the energy of fingerprint significantly reduces in collusion, and this makes detector fingerprint to be detected.

(2) correlation fingerprint shortcoming is: introduce correlation in the design, even if the duplicate of digital content is subjected to collusion attack, in fingerprint, the energy of relevant portion also can not significantly reduce.Correlation fingerprint forms primarily of encoded fingerprint, and the C-safety code wherein proposed by Boneh and Shaw the earliest, proposes double-deck code construction method and resist to the attack of the conspirator of determined number.This finger-print codes scheme is improved further by Yacobi, and the coding layer of the embeding layer of a direct spreading sequence and C-safety code combines by he, improves number of users.But the major defect of this method is, a) when synchronous distortion, be difficult to extract effective digital finger-print.B) resistance against colluders person limited amount.

(3) people such as WadeTrappe proposes a kind of collusion resistant code ACC in conjunction with the theory (as BIBD etc.) of Combination Design and district's group coding, its major defect is: finger-print codes and embedding are combined, although improve the ability of the resistance against colluders of fingerprint, considerably increase the instructions for use of transmission bandwidth.Therefore, to realize cost higher for system.

Summary of the invention

The object of this invention is to provide a kind of fail safe high, effectively can reduce the method realizing the Video Multicast transmission of digital finger-print encryption of communications cost, solve the impact of transmission bandwidth and system attack collusion attack factor, reduce the computing cost of transmitting terminal, strengthen right and the validity thereof of the illegal person of system tracks simultaneously.

The present invention is achieved through the following technical solutions.

Realize a method for the Video Multicast transmission of digital finger-print encryption, it is characterized in that: the described method realizing the Video Multicast transmission of digital finger-print encryption comprises digital finger-print encrypted transmission and digital finger-print and encrypts and receive.

Described digital finger-print encrypted transmission comprises the following step:

1) the carrier information source data of image and video are compressed through Moving Picture Experts Group-2, form MPEG-2 master data S _l;

2) digital fingerprint data of image or video utilize DGHM multi-wavelet transformation to set to process, form a kind of m ultiwavelet tree;

3) according to the importance ranking of digital finger-print, the different pieces of information of digital fingerprint image characteristic point is become the fingerprint vector base of different-energy according to the different decomposition of m ultiwavelet tree " father-child node " relation and form digital fingerprint data

4) the energy size that the probability decision digital finger-print according to the different branches of tree structure, collusion attack occurring mutually embeds and relation thereof;

5) in compression domain, according to key K ^m, K ⁽ⁱ⁾by CDMA/TDMA quadrature modulation matrix, digital finger-print modulation is embedded into S _lin, form the blended data comprising carrier master data and digital finger-print wherein, K ^mfor PKI, K ⁽ⁱ⁾for user u ⁽ⁱ⁾private key;

6) data flow forming transmission is encrypted respectively, finally by transmission to the data of multicast and clean culture;

Described digital finger-print encryption reception comprises the following steps:

1) according to key K ^m, K ⁽ⁱ⁾the compression bit flow data V of deciphering multicast and clean culture, wherein V may be subject to collusion attack, also may not have;

2) carry out MPEG-2 decompression, Recovery image and video source data, meanwhile, combine orthogonal matrix according to CDMA/TDMA and carry out demodulation;

3) DGHM decomposition is carried out to image and video data, and the different frequency sub-bands coefficient on same spatial location at different levels and direction is divided into same group and forms one and have the symmetrical junction paper mulberry of " father-child node " relation;

4) odd even is calculated to the sub-band coefficients of every grade of tree structure, extract the digital finger-print watermark data Y embedded ⁽ⁱ⁾;

5) detect with normalizated correlation coefficient, coefficient correlation is defined as follows:

$\mathrm{\η}=\frac{\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{W_l}^{(i_1,...,i_l)}\·Y^{(i_1,...,i_l)}}{\sqrt{\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{W_l}^{{(i_1,...,i_l)}^2}\·\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{Y}^{{(i_1,...,i_l)}^2}}}--------\left(1\right)$

If η > is T (detection threshold of setting), then digital fingerprint data detects, and wherein T is the detection threshold arranged, and η, more close to 1, illustrates Y ⁽ⁱ⁾more be approximately equal to original fingerprint data ;

6) according to statistics, relevance function detector tree generation collusion attack at different levels, finally obtains digital fingerprint data.

Described step 4) with step 6) in collusion attack comprise the following steps:

If SC detects and points out the index of collusion attack probability of happening, the suspicious copied content that definition collusion attack function produces is k is conspirator, and g (.) is collusion function, then for the 1≤l≤L level of tree, detect collusion process as follows from V:

A) the fingerprint tree first order detects: detector sets up the digital finger-print Y and each D1 fingerprint that extract in the first order statistic correlation:

$T^{i_1}=\frac{<Y,a^{i_1}>}{\left|\right|a^{i_1}\left|\right|},i_1=\mathrm{1,2},...,D_1-----\left(2\right)$

Wherein, || a|| is the Euclid norm of a, and first order detection is conspired the region of distorting and is h1 is first order predicted estimate thresholding;

B) fingerprint tree 2nd≤l≤L level detects: the estimation region of known (l-1) level collusion attack is GR (l-1), for any [i ₁..., i _l-1] result of ∈ GR (l-1), 2nd≤l≤L level detection statistics is:

$T^{i_1,...,i_{l-1},i_l}=\frac{<Y,a^{i_1,i_2,...,i_{l-1},i_l}>}{\left|\right|a^{i_1,i_2,...,i_{l-1},i_l}\left|\right|},i=\mathrm{1,2},...,D_l-------\left(3\right)$

Detect the region pointed to collusion and occur $\mathrm{GR}\left(l\right)=\left\{\right[i_1,...,i_l]:[i_1,...,i_{l-1}]\&Element;\mathrm{GR}(l-1).T^{i_1,...,i_l}\&GreaterEqual;h_l\},$ H _lfor the prediction thresholding of l level fingerprint detection;

C) detector estimates that the index exporting collusion generation is set to

Described digital finger-print CDMA/TDMA combines orthogonal matrix to carry out being provided with fingerprint drift compensation method in demodulation, be exactly embed digital fingerprint data in the motion vector of MPEG-2 video sequence reference frame, described fingerprint drift compensation method, contain the digital finger-print secret information that in reference frame, motion vector embeds, and validated user can only be transferred to by unicast tunnel, specifically comprise the following steps:

A) for the DCT coefficient that each MPEG-2 encodes, if it does not modulate embedding digital finger-print, so carry out Variable Length Code VLC operation, and pass through K together with motion vector, quantizing factor and side information ^mencryption, all users are given in multicast;

B) for the DCT coefficient that each MPEG-2 encodes, if its modulation embedded in digital finger-print, so first inverse quantization Q is carried out ^-1, then to user u ⁽ⁱ⁾modulation embeds corresponding digital fingerprint data and adds corresponding drift compensation signal data, is formed like this and comprises drift-compensated digital finger-print coefficient, then carry out quantification Q and Variable Length Code VLC.

C) last, pass through key K ⁽ⁱ⁾also clean culture is encrypted to user i to blended data.

Described digital finger-print CDMA/TDMA hybrid modulation matrix P=P ^jointbe a upper triangular matrix:

Known { p _{l, l}} _{l=1 ..., L}and P ^jointtime, we can find N ₁, N ₂..., N _lmeet relation:

$P^{\mathrm{Joint}}{\left[\begin{array}{ccc}{N}_1& N_2...& N_L\end{array}\right]}^T=N{\left[\begin{array}{ccc}{\mathrm{\&rho;}}_1& {\mathrm{\&rho;}}_2...& {\mathrm{\&rho;}}_L\end{array}\right]}^{T}s.t.\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{N}_l=N,0\&le;N_l\&le;N-----\left(12\right)$

Wherein p _{l, L}> ρ _l, p _{1, l}, p _{2, l}..., p _{l, l}be respectively tree structure Data Modulation coefficient at different levels; ρ ₁, ρ ₂..., ρ _lfor the probability of collusion attack appears in tree structure node at different levels.

In the embedding of described CDMA/TDMA hybrid modulation fingerprint, known P ^joint{ N _l} _{l=1 ..., L}, in the 1≤l≤L level of fingerprint tree structure, for fingerprint base there is relation:

$a^{i_1,...,i_l}={a_l}^{i_1,...,i_l}\mathrm{\&Pi;}{a_{l+1}}^{i_1,...,i_l}...\mathrm{\&Pi;}{a}_L^{i_1,...,i_l}----\left(13\right)$

Here, gaussian distributed and separate, length is Nk, and is embedded in Sk, and symbol " ∏ " represents the series connection factor.

For user u ⁽ⁱ⁾the l part number fingerprint content received is here, digital finger-print concrete form be:

$W_i^{(i_1,...,i_l)}=\sqrt{p_{1,l}}{a}_l^{i_1}+\sqrt{p_{2,l}}{a}_l^{i_1,i_2}+...+\sqrt{p_{l,l}}{a}_l^{i_1,...,i_l}----\left(14\right)$

During collusion attack, suppose total K conspirator, SC is the detection index pointing to conspirator, and collusion attack person divides L for each grouping for 1≤l < L, known digital content assailant in SCl by produce l part and conspire copied content, wherein, nl is the additive noise introduced by collusion attack person, supposes that conspiring the digital copies content obtained is V=V ₁Ω V ₂... Ω V _l.

In the detection of described CDMA/TDMA hybrid modulation fingerprint, in reception test side, the known digital content by secret key decryption is V, and for 1≤l < L, detecting step is as follows:

1) first, detector is from V ₁middle extraction Y ₁;

2) first order of fingerprint tree detects: at the 1st grade, detector makes the fingerprint { Y of extraction _l} _{l=1 ..., L}with each D1 fingerprint base be correlated with and calculate ASSOCIATE STATISTICS:

$T^{i_1}=\frac{\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}<Y_k,{a_k}^{i_1}>}{\sqrt{\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}}{\left|\right|{a_k}^{i_1}\left|\right|}^2},i_1=\mathrm{1,2},...,D_1----\left(15\right)$

Detector estimates that the first order collusion crime region exported is:

$\mathrm{GR}\left(1\right)=\left\{\right[i_1]:T^{i_1}>h_1-----\left(16\right)$

Wherein, h ₁for the preset thresholding of tree structure first order fingerprint detection;

3) 2nd≤l < L level of fingerprint tree detects: the collusion region that known detector exported in the past is GR (l-1), for each [i ₁, i ₂..., i _l-1] ∈ GR (l-1), detector calculates ASSOCIATE STATISTICS:

$T^{i_1,...,i_{l-1},i_l}=\frac{\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}<Y_k,{a_k}^{i_1,...,i_{l-1},i_l}>}{\sqrt{\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}\left|\right|{a_k}^{i_1,...,i_{l-1},i_l}{\left|\right|}^2}},i_1=\mathrm{1,2},...,D_l--------\left(17\right)$

And point out that collusion attack region is: $\mathrm{GR}\left(l\right)=\left\{\right[i_1,...,i_l]:[i_1,...,i_{l-1}]\&Element;\mathrm{GR}(l-1),T^{i_1,...,i_l}>h_l,$ Wherein, h _lfor the preset thresholding of tree structure l level fingerprint detection;

4) last, collusion attack person's index that detector exports is:

In described digital finger-print encrypted transmission, the process that digital finger-print embeds and encryption distributes is as follows:

A) according to the tree structure of digital finger-print, by formula $W_l^{(i_1,...,i_l)}=\sqrt{p_{1,l}}{a}_l^{i_1}+\sqrt{p_{2,l}}{a}_l^{i_1,i_2}+...+\sqrt{p_{l,l}}{a}_l^{i_1,...,i_l}$ Generate user u ⁽ⁱ⁾digital fingerprint data W ⁽ⁱ⁾;

B) MPEG-2 compression bit stream is divided into two parts: part 1 comprises motion vector, quantizing factor and other side information, and part 2 comprises DCT coefficient and VLD, the wherein variable normal decoding of VLD;

C) for the DCT coefficient not embedding fingerprint, VLC coding is adopted.For the DCT coefficient embedding fingerprint, first carry out inverse quantization, if it belongs to Sl, grouping in the corresponding finger print data digital finger-print that adopts SS (spread spectrum) mode to embed be then, carry out VLC to all DCT coefficient comprising finger print data to form the hybrid digital content comprising digital finger-print and be

D) DCT coefficient of non-embedded finger print data uses K ^mencryption and with motion vector, side information etc. together with multicast to all users.For 1≤l < L, use encrypt also multicast to grouping user digital content use private key K ⁽ⁱ⁾encrypt also clean culture to user u ⁽ⁱ⁾;

During described digital finger-print encryption receives, the process that digital finger-print embeds and encryption distributes is as follows:

A) according to key K ^m, K ⁽ⁱ⁾the digital content data of deciphering multicast and clean culture wherein, V may be subject to collusion attack, also may not have;

B) MPEG-2 decompression is carried out: first inverse variable-length decoding (i.e. VLC is carried out to the data of deciphering ^-1decoding); Then inverse quantization Q is carried out ^-1, obtain non-overlapped 8 × 8DCT block of digital picture and video source data, and according to key obtain the position of the DCT coefficient embedding digital finger-print, finally, in conjunction with motion compensation and inverse discrete cosine transformation operation recovery MPG-2 sequence of video images;

D) combine orthogonal matrix according to the CDMA/TDMA of formula (11) and demodulation is carried out to digital fingerprint data;

E) the symmetrical structure fingerprint based on Fig. 6 " father-child node " relation is set, and utilizes mould 2 to operate, and calculates odd even, extract the digital finger-print watermark data Y embedded to the sub-band coefficients of every grade of tree structure ⁽ⁱ⁾;

F) similarity measurement of digital finger-print is carried out with normalizated correlation coefficient defined formula (1);

G) according to statistic correlation function formula (15)-(17), the contingent collusion attack situation of detector tree, and export the last index detecting collusion attack.

The present invention compared with prior art has the following advantages.

The present invention considers in Video Multicast transmission application, the capability problems of digital finger-print opposing collusion attack; Follow the tracks of the ability of distorting disabled user based on digital finger-print to judge; Effective transmission bandwidth of digital finger-print and encryption method; In decoding end, reconstructing video technical problem.In the present invention 1) spread spectrum is utilized to embed digital finger-print in video streaming; In order to reduce communications cost, to the video coefficients that all user's multicast transmission do not embed; Authorized user clean culture is embedded to the video coefficients of digital finger-print.

The present invention proposes and design and allocative decision based on tree structure fingerprint multicasting technology and combination fingerprint.With existing orthogonal spectrum expansion fingerprint and encoded fingerprint Technical comparing, advantage of the present invention is:

1) consider the communications cost of system transfers and collusion resistant performance, according to the difference of number of users and sequence characteristic, save the communications cost of 48%-84% based on tree structure fingerprint multicasting technology than orthogonal spectrum expansion unicast technique.

2) in order to reduce transmission bandwidth further, the CDMA/TDMA proposed based on tree structure combines fingerprint design and distribution technique.It, compared with tree structure fingerprint multicasting technology, has saved 3%-9% communications cost.Orthogonal fingerprint technique, tree structure fingerprint multicasting technology and CDMA/TDMA combine three kinds of Technical comparing such as fingerprint design and distribution technique, and when number of users is less (an about 10-20 user), orthogonal fingerprint technique unicast technique performance is best; When number of users is more (as several thousand users), tree structure combination fingerprint is designed with best bandwidth usage effect.Therefore, for the video sequence (as personage's scene) embedding less fingerprint, tree structure fingerprint multicasting technology has better bandwidth efficiency.And for embedding the video sequence (as complex environment) of more fingerprint, under certain network and design conditions, tree structure CDMA/TDMA combines fingerprint design and has minimum communications cost.

3) in order to improve the video quality of receiving terminal reconstruct, under the prerequisite not increasing communications cost, compare tree structure fingerprint multicasting technology, the fingerprint drift compensation method that this patent proposes, the picture quality PSNR making CDMA/TDMA combine the reconstruct of fingerprint designing technique on average improves 1-1.5dB.

In order to improve the video quality of receiving terminal reconstruct, under the prerequisite not increasing communications cost, compare tree structure fingerprint multicasting technology, this patent proposes a kind of digital finger-print drift compensation method, the picture quality PSNR making CDMA/TDMA combine the reconstruct of fingerprint modulation embedded technology on average improves 1-1.5dB, namely improves the picture quality of recovery.

4) according to the not at the same level of tree structure and different branch, the ability of the mathematical method that patent adopts the digital finger-print modulation of different-energy size to embed and opposing collusion attack.

Accompanying drawing explanation

Fig. 1 is the digital finger-print encryption transmission system block diagram that the present invention realizes the method for the Video Multicast transmission of digital finger-print encryption;

Fig. 2 is the digital finger-print encryption receiving system block diagram that the present invention realizes the method for the Video Multicast transmission of digital finger-print encryption;

Fig. 3 is the sub-band coefficients graph of a relation that the present invention realizes 4 grades of m ultiwavelet decomposition of the method for the Video Multicast transmission of digital finger-print encryption;

Fig. 4 be the present invention realize the Video Multicast transmission of digital finger-print encryption method more than the Triangle Tree division figure of wavelet coefficient, the wherein a) grouping of m ultiwavelet tree in the one-dimensional space, Triangle Tree of b) dividing into groups;

Fig. 5 be the present invention realize digital finger-print encryption Video Multicast transmission method more than wavelet tree subband the each subtree of coefficient distributes fingerprint method schematic configuration diagram;

Fig. 6 is fingerprint allocative decision figure (L=3, the D based on tree structure that the present invention realizes the method for the Video Multicast transmission of digital finger-print encryption ₁=D ₂=2, D ₃=3);

Fig. 7 is the schematic diagram ([ρ that carrier signal divides as fingerprint tree L=3 that the present invention realizes the method for the Video Multicast transmission of digital finger-print encryption ₁, ρ ₂, ρ ₃]=[1/4,1/4,1/2]);

Fig. 8 is the schematic diagram of mutual collusion attack in fingerprint tree system that the present invention realizes the method for the Video Multicast transmission of digital finger-print encryption;

Fig. 9 is the MPEG-2 transmission system schematic diagram based on the modulation of CDMA/TDMA combined digital fingerprint that the present invention realizes the method for the Video Multicast transmission of digital finger-print encryption, wherein a) server transmitting terminal digital finger-print embeds and assigning process, b) user's receiving terminal decoding process implementation procedure;

Figure 10 is the block diagram that the digital finger-print of method that the present invention realizes the Video Multicast transmission of digital finger-print encryption is embedded into the drift compensation method of motion vector in reference frame.

Embodiment

The method below in conjunction with accompanying drawing, the present invention being realized to the Video Multicast transmission of digital finger-print encryption is described in further detail.

The present invention realizes the method for the Video Multicast transmission of digital finger-print encryption, comprises digital finger-print encrypted transmission and digital finger-print and encrypts and receive.

Described digital finger-print encrypted transmission comprises the following step:

1) the carrier information source data of image and video are compressed through Moving Picture Experts Group-2, form MPEG-2 master data S _l;

2) digital fingerprint data of image or video utilize DGHM multi-wavelet transformation to set to process, form a kind of m ultiwavelet tree;

3) according to the importance ranking of digital finger-print, the different pieces of information of digital fingerprint image characteristic point is become the fingerprint vector base of different-energy according to the different decomposition of m ultiwavelet tree " father-child node " relation and form digital fingerprint data

4) the energy size that the probability decision digital finger-print according to the different branches of tree structure, collusion attack occurring mutually embeds and relation thereof;

5) in compression domain, according to key K ^m, K ⁽ⁱ⁾by CDMA/TDMA quadrature modulation matrix, digital finger-print modulation is embedded into S _lin, form the blended data comprising carrier master data and digital finger-print $X_l^{\left(i\right)}=S_l+W_l^{\left(i\right)};$

6) data flow forming transmission is encrypted respectively, finally by transmission to the data of multicast and clean culture;

Described digital finger-print encryption reception comprises the following steps:

1) according to key K ^m, K ⁽ⁱ⁾the compression bit flow data V of deciphering multicast and clean culture, wherein V may be subject to collusion attack, also may not have;

2) carry out MPEG-2 decompression, Recovery image and video source data, meanwhile, combine orthogonal matrix according to CDMA/TDMA and carry out demodulation;

3) DGHM decomposition is carried out to image and video data, and the different frequency sub-bands coefficient on same spatial location at different levels and direction is divided into same group and forms one and have the symmetrical junction paper mulberry of " father-child node " relation;

4) odd even is calculated to the sub-band coefficients of every grade of tree structure, extract the digital finger-print watermark data Y embedded ⁽ⁱ⁾;

5) be defined as follows with normalizated correlation coefficient:

$\mathrm{\&eta;}=\frac{\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{W_l}^{(i_1,...,i_l)}\&CenterDot;Y^{(i_1,...,i_l)}}{\sqrt{\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{W_l}^{{(i_1,...,i_l)}^2}\&CenterDot;\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{Y}^{{(i_1,...,i_l)}^2}}}--------\left(1\right)$

If η > is T (detection threshold of setting), then digital fingerprint data detects, and wherein T is the detection threshold arranged, and η, more close to 1, illustrates Y ⁽ⁱ⁾more be approximately equal to original fingerprint data ;

6) according to statistics, relevance function detector tree generation collusion attack at different levels, finally obtains digital fingerprint data.

Be described below in detail:

One. system survey with realize principle

Realize as digital watermarking the powerful guarantee that copyrighted digital content protection is information security application with digital finger-print; wherein the transmission technology of Information hiding is the core technology of information security application, and collusion resistant and incompressible attack are the Key Performance Indicators of Information hiding core technology.Especially under different transmission channel conditions, realizing digital of digital video data and be reliably transferred to multiple users simultaneously, and meet the requirement of cbr (constant bit rate) and low distortion, is the focal spot techniques of current multi-casting communication system research.According to the traditional theory of Shannon ' s, when given channel capacity, source and channel coding branch realizes.But under multi-user environment, this technology is difficult to the optimization encoding and decoding reaching information source and channel.Therefore, there is resources configuration optimization problem in traditional multi-casting communication system.

This patent, for MPEG-2 transmission and the safety issue of application technology, proposes a kind of multicast transmission system utilizing digital finger-print to be encrypted as digital watermarking.The structure of whole system as shown in Figure 1-2, in figure, symbol the fingerprint vector data that presentation video or video data obtain after multi-wavelet transformation; be respectively tree structure Data Modulation coefficient at different levels; K ^mfor PKI, K ⁽ⁱ⁾for user u ⁽ⁱ⁾private key; The data content that V receives after expressing possibility and suffering collusion attack; Y is the digital fingerprint data extracted after decompressing.

(1) system principle and implementation procedure illustrate:

Transmission ends realizes principle and basic step:

A) the carrier information source data of image and video are compressed through Moving Picture Experts Group-2, form MPEG-2 master data S _l.B) digital fingerprint data of image or video the DGHM multi-wavelet transformation tree utilizing the people such as Kumsawat to propose processes, and forms a kind of m ultiwavelet tree.

C) according to the importance ranking of digital finger-print, the different pieces of information of digital fingerprint image characteristic point is become the fingerprint vector base of different-energy according to the different decomposition of m ultiwavelet tree " father-child node " relation and form digital fingerprint data

D) the energy size that the probability decision digital finger-print according to the different branches of tree structure, collusion attack occurring mutually embeds and relation thereof.

E) in compression domain, according to key K ^m, K ⁽ⁱ⁾by CDMA/TDMA quadrature modulation matrix, digital finger-print modulation is embedded into S _lin, form the blended data comprising carrier master data and digital finger-print $X_l^{\left(i\right)}=S_l+W_l^{\left(i\right)}.$

F) data flow forming transmission is encrypted respectively, finally by transmission to the data of multicast and clean culture.

Receiving terminal realizes principle and basic step:

A) according to key K ^m, K ⁽ⁱ⁾the compression bit flow data V of deciphering multicast and clean culture.Wherein, V may be subject to collusion attack, also may not have.

B) MPEG-2 decompression is carried out, Recovery image and video source data.Meanwhile, combine orthogonal matrix according to CDMA/TDMA and carry out demodulation.

C) DGHM decomposition is carried out to image and video data, and the different frequency sub-bands coefficient on same spatial location at different levels and direction is divided into same group and forms one and have the symmetrical junction paper mulberry of " father-child node " relation.

D) operate based on mould 2, odd even is calculated to the sub-band coefficients of every grade of tree structure, extracts the digital finger-print watermark data Y embedded ⁽ⁱ⁾.

E) similarity measurement: be defined as follows with normalizated correlation coefficient:

$\mathrm{\&eta;}=\frac{\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{W_l}^{(i_1,...,i_l)}\&CenterDot;Y^{(i_1,...,i_l)}}{\sqrt{\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{W_l}^{{(i_1,...,i_l)}^2}\&CenterDot;\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{Y}^{{(i_1,...,i_l)}^2}}}-------\left(1\right)$

If η > is T (detection threshold of setting), then digital fingerprint data detects.Obviously, η more illustrates Y close to 1 ⁽ⁱ⁾more be approximately equal to original fingerprint data .

F) according to the contingent collusion attack situation of statistic correlation function check tree structure.Wherein, it is a characteristic part of the present invention that collusion attack detects, and is described in detail as follows at this.

The Cleaning Principle of receiving terminal collusion attack and basic step:

If SC detects and points out the index of collusion attack probability of happening, the suspicious copied content that definition collusion attack function produces is k is conspirator, and g (.) is collusion function.Then for the 1≤l≤L level of tree, from V, detect collusion process as follows:

A) the fingerprint tree first order detects: detector sets up the digital finger-print Y and each D1 fingerprint that extract in the first order statistic correlation:

$T^{i_1}=\frac{<Y,a^{i_1}>}{\left|\right|a^{i_1}\left|\right|},i_1=\mathrm{1,2},...,D_1------\left(2\right)$

Wherein, || a|| is the Euclid norm of a.The region of distorting is conspired in first order detection $\mathrm{GR}\left(1\right)=\left\{\right[i_1]:T^{i_1}>h_1\},$ H1 is first order predicted estimate thresholding.

B) fingerprint tree 2nd≤l≤L level detects: the estimation region of known (l-1) level collusion attack is GR (l-1).For any [i ₁..., i _l-1] result of ∈ GR (l-1), 2nd≤l≤L level detection statistics is:

$T^{i_1,...,i_{l-1},i_l}=\frac{<Y,a^{i_1,i_2,...,i_{l-1},i_l}>}{\left|\right|a^{i_1,i_2,...,i_{l-1},i_l}\left|\right|},i=\mathrm{1,2},...,D_l------\left(3\right)$

Detect the region pointed to collusion and occur $\mathrm{GR}\left(l\right)=\left\{\right[i_1,...,i_l]:[i_1,...,i_{l-1}]\&Element;\mathrm{GR}(l-1).T^{i_1,...,i_l}\&GreaterEqual;h_l\},$ H _lfor the prediction thresholding of l level fingerprint detection.

C) detector estimates that the index exporting collusion generation is set to

(2) realization condition of system: compression standard is the video image under MPEG-2 meaning; Transmission code rate is 1Mbps-50Mbps; Multicast users number: 1000-10000.

Two, system components implementation procedure

(1) m ultiwavelet of digital picture and digital video frame decomposes

When DGHM m ultiwavelet decomposes, under different yardsticks, the key property that its correlated wavelets sub-band coefficients has is that direction is identical with position.When yardstick is determined, spatially, except most high-frequency sub-band, each " father node " sub-band coefficients is relevant to all " child node " sub-band coefficients on the same direction of next stage.For the multistage subband of m ultiwavelet, when r=2 ties up, between " father-child node " sub-band coefficients that its 4 grades of m ultiwavelet decompose dependence as Fig. 3 a) shown in,

After pre-filtering, original image or video frame images are divided into the identical subgraph of 4 width sizes.The discrete m ultiwavelet decomposable process of two dimensional image and single wavelet decomposable process similar, but owing to there being multiple yardstick (small echo) function to exist, a subgraph after converting with the single wavelet of image is corresponding, has r subgraph after multi-wavelet transformation.When the 1st layer of m ultiwavelet decomposes, after 2 wavelet low-pass filtering and high-pass filtering, original image becomes 16 subgraphs, and its Relationship of Coefficients is as the b of Fig. 3).When carrying out the 2nd layer and decomposing, similar single wavelet decomposes, and just low frequency part decomposed further, high frequency imaging part remains unchanged.By that analogy.After having decomposed, top 4 low-frequency images are only had to be low-frequency images completely, all containing high-frequency components in all the other subgraphs.Under r ties up yardstick, picture breakdown can be r by L level multi-wavelet transformation ²(3L+1) individual subgraph.After m ultiwavelet decomposes at every turn, define more sub-band coefficients, Fig. 3 " father-child node " relation a) may be destroyed, need its position rearrangement reaction.Detailed process resets each 2 × 2 sub-blocks, and 4 of same spatial location relation coefficients are come together.

Other subband, we are divided into same group the different frequency sub-bands coefficient in same spatial location and direction, and Fig. 4 describes in the one-dimensional space, and 16 coefficients in 4 coefficients of 1 coefficient in HL4 and HL3 and HL2 are divided example in groups.The sum of its grouping equals the sum of LH4, HL4 and HH4 sub-band coefficients, has 3 × 32 × 32=3072.

(2) the m ultiwavelet tree structure of digital picture and digital video frame and the division condition of digital finger-print thereof: according to different distributions and the different importance thereof of digital picture and digital video frame, the present invention constructs the digital finger-print Information Hiding Techniques at a kind of m ultiwavelet tree structure.

M ultiwavelet tree building method is as follows: digital picture and digital video frame are after DGHM decomposes, and we are divided into same group the different frequency sub-bands coefficient in same spatial location and direction and form a subtree.According to importance relation, each sub-band coefficients data of image and frame of video are become the fingerprint vector base of different-energy according to the different demarcation of m ultiwavelet tree " father-child node " relation.The different nodes of tree produce the digital finger-print information of varying strength.As shown in Figure 5, other grouping coefficient is similar for its structure in the one-dimensional space.

The division condition of digital finger-print: in Fig. 5, the degree of depth of tree is L=4, and every first nodes has identical descendant nodes number D _l, l=1,2 ..., L, except each node in tree wherein root node, i=[i ₁..., i _l], i _l=1,2 ..., D _l, the vector basis of unique digital fingerprint data (YZ represents certain sub-band coefficients in m ultiwavelet tree) meets Gaussian Profile and it is separate.

(3) digital finger-print based on tree structure resistance against colluders designs

Based on the digital finger-print structure of Fig. 5, Fig. 6 describes a kind of symmetrical tree structural fingerprint design with resistance against colluders that the present invention proposes.In symmetrical tree structure, the degree of depth of each leaf node is L, (l-1, l=1,2 ..., L) and level has identical child node Dl.

Be associated with collusion attack, suppose here: grouping user U ^1,1in the probability of mutual collusion be all p ₃.U ^1,1each user of grouping and another grouping U ^1,2the probability that user conspires mutually is p ₂, p ₂< p ₃.U ^1,1∪ U ^1,2in grouping, the probability of each user and other groups user collusion is p ₁, and p ₁< p ₂< p ₃.Tree at different levels on, its unique finger print data base tree node [i1 ..., i1] on (except root node) according to Gaussian Profile rule produce.Wherein l is the degree of depth of node, and i is certain user node, and { a} is separate for fingerprint vector base simultaneously.For each user, all digital finger-prints from leaf node to root node distribute to him, namely distribute to user u ⁽ⁱ⁾digital finger-print content X ⁽ⁱ⁾in, L fingerprint of total embedding is such as, in Fig. 6, digital finger-print a ¹, a ^1,1and a ^1,1,1be embedded into digital content X ⁽¹⁾in and distribute to user u ⁽¹⁾.

(4) based on digital finger-print modulation technique and the allocative decision of tree structure

CDMA and the TDMA modulation technique of 4.1 digital finger-prints

Suppose that carrier information source data-signal S has N number of DCT coefficient that may embed fingerprint, so embedding in L fingerprint to S has two kinds of methods usually: based on CDMA with based on TDMA fingerprint modulation embedding grammar.

Based on CDMA digital finger-print modulation embedded technology: in CDMA fingerprint modulation technique, { a} has identical length N and equal energy to fingerprint base.Like this, user u ⁽ⁱ⁾digital finger-print be the cumulative of tree structure fingerprint vector base at different levels, its concrete form is:

$W^{\left(i\right)}=\sqrt{{\mathrm{\&rho;}}_1}{a}^{i_1}+\sqrt{{\mathrm{\&rho;}}_2}{a}^{i_1,i_2}+...+\sqrt{{\mathrm{\&rho;}}_L}{a}^{i_1,i_2,...,i_L}-----\left(4\right)$

Wherein, { ρ _l, l=1,2 ..., L} for setting by difference the sizes values that the digital finger-print index of modulation that collusion attack probability determines occurs mutually in branch, and has relation this illustrates, { ρ _l, l=1,2 ..., L} controls the energy size of the embedding digital finger-print not at the same level of tree and regulates the correlation between different user.Therefore, user u is distributed to ⁽ⁱ⁾and the video content form comprising digital finger-print be: X ⁽ⁱ⁾=S+W ⁽ⁱ⁾, wherein S is carrier signal.

Based on TDMA fingerprint modulation technique: in TDMA fingerprint modulation technique, signal S is divided into S ₁..., S _ldeng L different non-overlapped part.Therefore, S _lin can embed fingerprint carrier signal number of coefficients be N _l=ρ _ln, and in Fig. 7, in every 4s, within first second, all frames belong to Sl, and within second second, all frames belong to S2, and last two seconds in S3.If sequence of video images long enough, S _lin can embed fingerprint carrier signal number of coefficients be approximately N _l=p _ln.Based in TDMA fingerprint modulation technique, set the fingerprint base of l level having length is N _l.Digital finger-print content X ⁽ⁱ⁾distribute to user u ⁽ⁱ⁾, the fingerprint base of l level be embedded into carrier signal S _lin l part.Therefore, the video content comprising digital finger-print is:

4.2 digital finger-print CDMA and TDMA modulation matrix

In order to compare CDMA and the TDMA modulating performance based on tree structure finger prints processing system, we mainly measure the fingerprint energy size embedding different digital fingerprint content.Suppose that carrier signal S is divided into L non-overlapped part wherein N _lfor S _lin can embed the signal number of fingerprint.Meanwhile, we suppose for user u ⁽ⁱ⁾, for S _lthe digital finger-print of middle embedding.For CDMA and TDMA modulation, distribute to user u ⁽ⁱ⁾and the l part number video content concrete form comprising digital finger-print is:

$X_l^{\left(i\right)}=S_l+W_l^{\left(i\right)}-----\left(5\right)$

Definition E _{k, l}for the k level fingerprint base of tree be embedded in in energy, and for gross energy.We define further, and the element that matrix P row k l arranges is represent the k level fingerprint base of tree be embedded in in energy and the ratio of gross energy.Therefore, the concrete form of digital finger-print CDMA and TDMA modulation matrix is respectively formula (6) and formula (7):

And relation is had in based on TDMA modulation technique:

P ^tDMA[N ₁n ₂n _l] ^t=N [ρ ₁ρ ₂ρ _l] ^t-------(8) n is the sum that can embed fingerprint base in carrier signal coefficient.

The effective bandwidth of CDMA and TDMA modulation digital fingerprint compares: first, based in TDMA modulation technique, as k > l, has p _{k, l}=0, therefore, in only tree k≤l level embedded in fingerprint base content.Notice, fingerprint base in grouping in shared by all users, so fingerprint is therefore, in TDMA modulation technique, tree structure fingerprint distribution system is not only to the non-overlapped coefficient of all user's multicasts, and the coefficient of embedding fingerprint that multicast grouping user is shared.And in CDMA modulation technique, as k > l, have p _{k, l}> 0, tree structure fingerprint distribution system is the non-overlapped coefficient of multicast only.Therefore, from bandwidth availability ratio, based on TDMA modulation technique than more effective based on CDMA modulation technique.

The resistance against colluders of CDMA and TDMA modulation digital fingerprint compares: based in TDMA modulation technique, as k ≠ l, have p _{k, l}=0, l level fingerprint base only be embedded in digital finger-print content in l part.So collusion attack person can distinguish the different piece of digital content after the embedding fingerprint not at the same level of tree.They can point out that the structure that fingerprint is set and conspirator are in the position of tree.According to the information of collecting, they can apply special attack pattern-mutual collusion attack, attack TDMA modulation technique.Suppose that SC is that the index pointing to collusion attack is arranged, { X ⁽ⁱ⁾} _{i ∈ SC}for the digital content containing finger print data received.In mutual collusion attack, collusion attack person is divided into L grouping and at least exist 1≤l < L make l divide into groups SC _lto divide into groups SC with (l+1) _l+1be in the different branches of fingerprint tree, namely have relation collusion copied content V comprises L non-overlapped part { V _l} _{l=1 ..., L}, at grouping SC _lconspirator by collusion function produce the l partial content in V.Fig. 8 describes a kind of example that mutual collusion attack may occur and also detect on fingerprint tree structure.The index SC={1=[1,1,1] that collusion attack occurs, 2=[1,1,2], 4=[1,2,1], 7=[2,1,1] is supposed in figure } (namely there is dash area in the drawings in collusion attack).Conspirator selects SC ₁={ 7}, SC ₂={ 4}, SC ₃={ 1,2}.

The collusion digital content produced is:

$V_1=X_1^{\left(7\right)}=S_1+a^2,$

$V_2=X_2^{\left(4\right)}=S_2+a^{\mathrm{1,2}}$

$V_3=(X_3^{\left(1\right)}+X_3^{\left(2\right)})/2=S_3+(a^{\mathrm{1,1,1}}+a^{\mathrm{1,1,2}})/2------\left(8\right)$

During detection, in the first order of tree, although a ¹and a ²all be suspected of committing a crime, but detector only can detect a ², because a ¹directly do not form any part of V.At this moment, detector output detections region is GR (1)=[2].In the second level, detector is attempted to detect whether [2,1] or [2,2] is crime subregion, but because a ^2,1and a ^2,2not in V, therefore, detector points out that they are not crime subregions.Due to special construction and unique multistage process thereof of the design of fingerprint tree, CDMA fingerprint modulation technique collusion resistant ability is better than TDMA fingerprint modulation technique.

4.3 digital finger-print CDMA/TDMA hybrid modulation technology

In conjunction with the advantage of digital finger-print CDMA in 4.2 discussions and TDMA modulation matrix technology, patent proposes a kind of digital finger-print CDMA/TDMA hybrid modulation embedded technology further.In the design of combination fingerprint and distribution system, first digital content owner or copyright owner generate a digital finger-print tree according to Fig. 6.Then, CDMA/TDMA hybrid modulation technology is used to embed finger print data, to improve effective bandwidth.Finally, Content owner or copyright owner are by certain form distribute digital content.

Digital finger-print CDMA/TDMA hybrid modulation matrix

In order to realize resisting the robustness of collusion attack and meet the validity requirement of transmission bandwidth, at this, a kind of CDMA/TDMA hybrid modulation technology is proposed, now modulation matrix P=P ^jointit is a upper triangular matrix.At CDMA/TDMA hybrid modulation matrix P ^jointin, its structural condition is:

P _{k, l}=0,-----(9) as k > l

0 < p _{k, l}≤ 1,------(10) as k≤l

Wherein, (9) formula achieves effective bandwidth, the robustness (robustness means the ability of opposing collusion attack) of (10) formula feasible system.

According to the requirement of formula (9) and (10), for digital finger-print tree, at the 1st grade, p _1,1=1.In 2≤l≤L level, known p _{l, l}, we find { p _{k, l}} _{k < l}meet a kind of fingerprint energy distribution relation: E _{1, l}: E _{2, l}: ...: E _{l-1, l}=ρ ₁: ρ ₂: ...: ρ _l-1.At this moment, as k < l, there is relation p _{k, l}=ρ _k(1-p _{l, l})/(ρ ₁+ ... + ρ _l-1), the final CDMA/TDMA hybrid modulation matrix form formed is:

Known { p _{l, l}} _{l=1 ..., L}and P ^jointtime, we can find N ₁, N ₂..., N _lmeet relation:

$P^{\mathrm{Joint}}{\left[\begin{array}{ccc}{N}_1& N_2...& N_L\end{array}\right]}^T=N{\left[\begin{array}{ccc}{\mathrm{\&rho;}}_1& {\mathrm{\&rho;}}_2...& {\mathrm{\&rho;}}_L\end{array}\right]}^{T}s.t.\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{N}_l=N,0\&le;N_l\&le;N---\left(12\right)$

From (11), work as p _{l, L}=ρ _ltime, P ^jointit is exactly CDMA fingerprint modulation technique.Thus, we only consider p _{l, L}> ρ _lsituation.

The embedded technology of CDMA/TDMA hybrid modulation fingerprint

Known P ^joint{ N _l} _{l=1 ..., L}, in the 1≤l≤L level of fingerprint tree structure, for fingerprint base there is relation: $a^{i_1,...,i_l}={a_l}^{i_1,...,i_l}\mathrm{\&Pi;}{a_{l+1}}^{i_1,...,i_l}...\mathrm{\&Pi;}{a}_L^{i_1,...,i_l}-----\left(13\right)$

Here, gaussian distributed and separate, length is Nk, and is embedded in Sk, and symbol " ∏ " represents the series connection factor.

For user u ⁽ⁱ⁾the l part number fingerprint content received is $X_l^{(i_1,...,i_l)}=S_l+W_l^{(i_1,...,i_l)}.$

Here, digital finger-print concrete form be:

$W_l^{(i_1,...,i_l)}=\sqrt{p_{1.l}}{a}_l^{i_1}+\sqrt{p_{2,l}}{a}_l^{i_1,i_2}+...+\sqrt{p_{l,l}}{a}_l^{i_1,...,i_l}------\left(14\right)$

During collusion attack, suppose total K conspirator, SC is the detection index pointing to conspirator, and collusion attack person divides L for each grouping for 1≤l < L, known digital content assailant in SCl by produce l part and conspire copied content, wherein, nl is the additive noise introduced by collusion attack person.Suppose that conspiring the digital copies content obtained is V=V1 Ω V2 ... Ω VL.

The detection actualizing technology of CDMA/TDMA hybrid modulation fingerprint

In reception test side, the known digital content by secret key decryption is V, and for 1≤l < L, detecting step is as follows:

1) first, detector extracts Yl from Vl.

2) first order of fingerprint tree detects: at the 1st grade, detector makes the fingerprint { Y of extraction _l} _{l=1 ..., L}with each D1 fingerprint base be correlated with and calculate ASSOCIATE STATISTICS:

$T^{i_1}=\frac{\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}<Y_k,{a_k}^{i_1}>}{\sqrt{\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}{\left|\right|{a_k}^{i_1}\left|\right|}^2}},i_1=\mathrm{1,2},...,D_1-------\left(15\right)$

Detector estimates that the first order collusion crime region exported is:

$\mathrm{GR}\left(1\right)=\left\{\right[i_1]:T^{i_1}>h_1------\left(16\right)$

Wherein, h ₁for the preset thresholding of tree structure first order fingerprint detection.

3) 2nd≤l < L level of fingerprint tree detects: the collusion region that known detector exported in the past is GR (l-1), for each [i ₁, i ₂..., i _l-1] ∈ GR (l-1), detector calculates ASSOCIATE STATISTICS:

$T^{i_1,...,i_{l-1},i_l}=\frac{\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}<Y_k,{a_k}^{i_1,...,i_{l-1},i_l}>}{\sqrt{\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}{\left|\right|{a_k}^{i_1,...,i_{l-1},i_l}\left|\right|}^2}},i_1=\mathrm{1,2},...,D_l------\left(17\right)$

And point out that collusion attack region is: $\mathrm{GR}\left(l\right)=\left\{\right[i_1,...,i_l]:[i_1,...,i_{l-1}]\&Element;\mathrm{GR}(l-1),T^{i_1,...,i_l}>h_l.$ Wherein, h _lfor the preset thresholding of tree structure l level fingerprint detection.

4) last, collusion attack person's index that detector exports is:

(5) MPEG-2 digital finger-print modulation technique and transmission actualizing technology thereof

Based on CDMA/TDMA hybrid modulation technology, Fig. 9 is illustrated as a kind of MPEG-2 video request program (VOD) multicast transmission actualizing technology.Suppose K ^mfor PKI, for grouping user key, K ⁽ⁱ⁾for user u ⁽ⁱ⁾private key.

Fig. 9 a) in server transmitting terminal, digital finger-print embed and encryption distribute process as follows:

A) according to the tree structure of digital finger-print, by formula $W_l^{(i_1,...,i_l)}=\sqrt{p_{1,l}}{a}_l^{i_1}+\sqrt{p_{2,l}}{a}_l^{i_1,i_2}+...+\sqrt{p_{l,l}}{a}_l^{i_1,...,i_l}$ Generate user u ⁽ⁱ⁾digital fingerprint data W ⁽ⁱ⁾.

B) MPEG-2 compression bit stream is divided into two parts: part 1 comprises motion vector, quantizing factor and other side information; Part 2 comprises DCT coefficient and VLD (variable normal decoding).

C) for the DCT coefficient not embedding fingerprint, VLC coding is adopted.For the DCT coefficient embedding fingerprint, first carry out inverse quantization, if it belongs to Sl, grouping in the corresponding finger print data digital finger-print that adopts SS (spread spectrum) mode to embed be then, carry out VLC to all DCT coefficient comprising finger print data to form the hybrid digital content comprising digital finger-print and be

D) DCT coefficient of non-embedded finger print data uses K ^mencryption and with motion vector, side information etc. together with multicast to all users.For 1≤l < L, use encrypt also multicast to grouping user digital content use private key K ⁽ⁱ⁾encrypt also clean culture to user u ⁽ⁱ⁾.

B at Fig. 9) in server receiving terminal, digital finger-print embed and encryption distribute process as follows:

A) according to key K ^m, K ⁽ⁱ⁾the digital content data of deciphering multicast and clean culture wherein, V may be subject to collusion attack, also may not have.

B) MPEG-2 decompression is carried out: first inverse variable-length decoding (i.e. VLC is carried out to the data of deciphering ^-1decoding); Then inverse quantization (i.e. Q is carried out ^-1); Obtain non-overlapped 8 × 8DCT block of digital picture and video source data, and according to key obtain the position of the DCT coefficient embedding digital finger-print.Finally, in conjunction with motion compensation and inverse discrete cosine transformation (i.e. IDCT) operation recovery MPG-2 sequence of video images.

D) combine orthogonal matrix according to the CDMA/TDMA of formula (11) and demodulation is carried out to digital fingerprint data.

E) the symmetrical structure fingerprint based on Fig. 6 " father-child node " relation is set, and utilizes mould 2 to operate, and calculates odd even, extract the digital finger-print watermark data Y embedded to the sub-band coefficients of every grade of tree structure ⁽ⁱ⁾.

F) similarity measurement of digital finger-print is carried out with normalizated correlation coefficient defined formula (1).

If coefficient of similarity η > T (detection threshold of setting), then digital fingerprint data detects.Obviously, η more illustrates Y close to 1 ⁽ⁱ⁾more be approximately equal to original fingerprint data .

G) according to statistic correlation function formula (15)-contingent collusion attack situation of (17) detector tree, and the last index detecting collusion attack is exported.

Three. the drift compensation method of digital finger-print CDMA/TDMA hybrid modulation

When digital finger-print transmits, the finger print data that former frame embeds may be diffused into next frame, and the digital finger-print between different frame accumulates and can affect the quality of the video frame images of receiving terminal reconstruct.Therefore, this patent is embedded into digital finger-print in the motion vector of reference frame, and is transferred to each user by unicast tunnel.Its implementation procedure as shown in Figure 10.

The implementation procedure of drift compensation method is as follows: a) for the DCT coefficient that each MPEG-2 encodes, if it does not modulate embedding digital finger-print, so carries out Variable Length Code VLC operation, and pass through K together with motion vector, quantizing factor and side information ^mencryption, all users are given in multicast.

B) for the DCT coefficient that each MPEG-2 encodes, if its modulation embedded in digital finger-print, so first inverse quantization Q is carried out ^-1, then to user u ⁽ⁱ⁾modulation embeds corresponding digital fingerprint data and adds corresponding drift compensation signal data, is formed like this and comprises drift-compensated digital finger-print coefficient.Then quantification Q and Variable Length Code VLC is carried out.

C) last, pass through key K ⁽ⁱ⁾also clean culture is encrypted to user i to blended data.

In a word, key point of the present invention is as follows:

(1) the tree structure digital finger-print embedding/etection theory model under collusion attack condition.Scientific practice proves, general Mathematical Modeling will have important directive significance for appropriate design and the concrete algorithm of evaluation analysis.Tree structure system of fingerprints theoretical model under this patent research collusion condition, be from information hiding system model general to specifically, again to concrete relation.Each stages such as detection after process of desynchronizing, digital finger-print embedding, collusion are analyzed and researched, effectively describing the mapping space size that desynchronizes, the copy visual quality that desynchronizes, copy the system internal relation under the multi-constraint condition such as copy quality and conspirator's quantity after visual quality, each stage execution speed, collusion after adding fingerprint, embedding means based theoretical for finding digital finger-print safely and effectively in the application.

(2) according to the not at the same level of tree structure and different branch, the mathematical method that the modulation of different-energy fingerprint embeds and the ability detecting collusion attack is adopted.Digital video generally exists in a compressed format in storage and transmitting procedure, considers that combining fingerprint modulation embedded technology with the DCT domain of compressed video by CDMA/TDMA has important practical significance to resistance against colluders in actual applications.Fingerprint embedding characteristic in compression domain is different from orthogonal spectrum expansion embedding parameter.Take compression domain DCT domain CDMA/TDMA to combine fingerprint and modulate the fingerprint energy that embedded technology effectively make use of branch not at the same level, different; selecting compression domain to embed coefficient and can ensure that execution speed can ensure again visual quality and the security performance that desynchronizes, is important key technology and the protection point of this patent Real Time Compression territory resistance against colluders.

(3) in order to improve the video quality of receiving terminal reconstruct, under the prerequisite not increasing communications cost, the fingerprint drift compensation method that the present invention proposes.

Claims (6)

1. realize a method for the Video Multicast transmission of digital finger-print encryption, it is characterized in that: the described method realizing the Video Multicast transmission of digital finger-print encryption comprises digital finger-print encrypted transmission and digital finger-print and encrypts and receive,

Described digital finger-print encrypted transmission comprises the following step:

1) the carrier information source data of image and video are compressed through Moving Picture Experts Group-2, form MPEG-2 master data S _l;

2) the digital fingerprint data W of image or video _l ⁽ⁱ⁾utilize DGHM multi-wavelet transformation to set to process, form a kind of m ultiwavelet tree;

3) according to the importance ranking of digital finger-print, the different pieces of information of digital fingerprint image characteristic point is become the fingerprint vector base of different-energy according to the different decomposition of m ultiwavelet tree " father-child node " relation and form digital fingerprint data

4) the energy size that the probability decision digital finger-print according to the different branches of tree structure, collusion attack occurring mutually embeds and relation thereof;

5) in compression domain, according to key K ^m, K ⁽ⁱ⁾by CDMA/TDMA quadrature modulation matrix, digital finger-print modulation is embedded into S _lin, form the blended data comprising carrier master data and digital finger-print wherein, K ^mfor PKI, K ⁽ⁱ⁾for user u ⁽ⁱ⁾private key;

6) data flow forming transmission is encrypted respectively, finally by transmission to the data of multicast and clean culture;

Described digital finger-print encryption reception comprises the following steps:

1) according to key K ^m, K ⁽ⁱ⁾the compression bit flow data V of deciphering multicast and clean culture, wherein V may be subject to collusion attack, also may not have;

2) carry out MPEG-2 decompression, Recovery image and video source data, meanwhile, combine orthogonal matrix according to CDMA/TDMA and carry out demodulation;

3) DGHM decomposition is carried out to image and video data, and the different frequency sub-bands coefficient on same spatial location at different levels and direction is divided into same group and forms one and have the symmetrical junction paper mulberry of " father-child node " relation;

4) odd even is calculated to the sub-band coefficients of every grade of tree structure, extract the digital finger-print watermark data Y embedded ⁽ⁱ⁾;

5) detect with normalizated correlation coefficient, coefficient correlation is defined as follows:

If η > is T (detection threshold of setting), then digital fingerprint data detects, and wherein T is the detection threshold arranged, and η, more close to 1, illustrates Y ⁽ⁱ⁾more be approximately equal to original fingerprint data W _l ⁽ⁱ⁾;

6) according to statistics, relevance function detector tree generation collusion attack at different levels, finally obtains digital fingerprint data,

Described digital finger-print CDMA/TDMA hybrid modulation matrix P=P ^jointbe a upper triangular matrix:

Known { p _{l, l}} _{l=1 ..., L}and P ^jointtime, we can find N ₁, N ₂..., N _lmeet relation:

P ^Joint[N ₁N ₂…N _L] ^T＝N[ρ ₁ρ ₂…ρ _L] ^T

Wherein p _{l, L}> ρ _l, p _{1, l}, p _{2, l}..., p _{l, l}be respectively tree structure Data Modulation coefficient at different levels; ρ ₁, ρ ₂..., ρ _lfor the probability of collusion attack appears in tree structure node at different levels.

2. according to claim 1 realize digital finger-print encryption Video Multicast transmission method, it is characterized in that described step 4) with step 6) in collusion attack comprise the following steps:

If SC detects and points out the index of collusion attack probability of happening, the suspicious copied content that definition collusion attack function produces is k is conspirator, and g (.) is collusion function, then for the 1≤l≤L level of tree, detect collusion process as follows from V:

A) the fingerprint tree first order detects: detector sets up the digital finger-print Y and each D1 fingerprint that extract in the first order statistic correlation:

Wherein, || a|| is the Euclid norm of a, and first order detection is conspired the region of distorting and is h1 is first order predicted estimate thresholding;

B) fingerprint tree 2nd≤l≤L level detects: the estimation region of known (1-1) level collusion attack is GR (1-1), for any [i ₁..., i _l-1] result of ∈ GR (l-1), 2nd≤l≤L level detection statistics is:

Detect the region pointed to collusion and occur h _lit is the prediction thresholding of 1 grade of fingerprint detection;

C) detector estimates that the index exporting collusion generation is set to

3. the method realizing the Video Multicast transmission of digital finger-print encryption according to claim 1, it is characterized in that: described digital finger-print CDMA/TDMA combines orthogonal matrix to carry out being provided with fingerprint drift compensation method in demodulation, be exactly embed digital fingerprint data in the motion vector of MPEG-2 video sequence reference frame, described fingerprint drift compensation method, contain the digital finger-print secret information that in reference frame, motion vector embeds, and validated user can only be transferred to by unicast tunnel, specifically comprise the following steps:

A) for the DCT coefficient that each MPEG-2 encodes, if it does not modulate embedding digital finger-print, so carry out Variable Length Code VLC operation, and pass through K together with motion vector, quantizing factor and side information ^mencryption, all users are given in multicast;

B) for the DCT coefficient that each MPEG-2 encodes, if its modulation embedded in digital finger-print, so first inverse quantization Q is carried out ^-1, then to user u ⁽ⁱ⁾modulation embeds corresponding digital fingerprint data and adds corresponding drift compensation signal data, is formed like this and comprises drift-compensated digital finger-print coefficient, then carry out quantification Q and Variable Length Code VLC;

C) last, pass through key K ⁽ⁱ⁾also clean culture is encrypted to user i to blended data.

4. the method realizing the Video Multicast transmission of digital finger-print encryption according to claim 1, is characterized in that in the embedding of described CDMA/TDMA hybrid modulation fingerprint, known P ^joint{ N _l} _{l=1 ..., L}, in the 1≤l≤L level of fingerprint tree structure, for fingerprint base there is relation:

Here, gaussian distributed and separate, length is Nk, and is embedded in Sk, and symbol " ∏ " represents the series connection factor,

For user u ⁽ⁱ⁾the part 1 digital finger-print content received is here, digital finger-print concrete form be:

During collusion attack, suppose total K conspirator, SC is the detection index pointing to conspirator, and collusion attack person divides L for each grouping for 1≤l < L, known digital content at SC ₁in assailant by produce part 1 collusion copied content, wherein, n _lbe the additive noise introduced by collusion attack person, suppose that conspiring the digital copies content obtained is V=V ₁Ω V ₂... Ω V _l.

5. the method realizing the Video Multicast transmission of digital finger-print encryption according to claim 4, it is characterized in that in the detection of described CDMA/TDMA hybrid modulation fingerprint, in reception test side, the known digital content by secret key decryption is V, for 1≤l < L, detecting step is as follows:

1) first, detector is from V ₁middle extraction Y ₁;

2) first order of fingerprint tree detects: at the 1st grade, detector makes the fingerprint { Y of extraction _l} _{l=1 ..., L}with each D1 fingerprint base be correlated with and calculate ASSOCIATE STATISTICS:

Detector estimates that the first order collusion crime region exported is:

Wherein, h ₁for the preset thresholding of tree structure first order fingerprint detection;

3) 2nd≤l < L level of fingerprint tree detects: the collusion region that known detector exported in the past is GR (1-1), for each [i ₁, i ₂..., i _l-1] ∈ GR (l-1), detector calculates ASSOCIATE STATISTICS:

And point out that collusion attack region is: wherein, h _lfor the preset thresholding of tree structure the 1st grade of fingerprint detection;

4) last, collusion attack person's index that detector exports is:

6. the method realizing the Video Multicast transmission of digital finger-print encryption according to claim 5, it is characterized in that in described digital finger-print encrypted transmission, the process that digital finger-print embeds and encryption distributes is as follows:

A) according to the tree structure of digital finger-print, by formula generate user u ⁽ⁱ⁾digital fingerprint data W ⁽ⁱ⁾;

B) MPEG-2 compression bit stream is divided into two parts: part 1 comprises motion vector, quantizing factor and other side information, and part 2 comprises DCT coefficient and VLD, the wherein variable normal decoding of VLD;

C) for the DCT coefficient not embedding fingerprint, adopt VLC coding, for the DCT coefficient embedding fingerprint, first carry out inverse quantization, if it belongs to S _l, grouping in the corresponding finger print data digital finger-print that adopts SS (spread spectrum) mode to embed be then, carry out VLC to all DCT coefficient comprising finger print data to form the hybrid digital content comprising digital finger-print and be

D) DCT coefficient of non-embedded finger print data uses K ^mencryption and with motion vector, side information together with multicast to all users, for 1≤l < L, use encrypt also multicast to grouping user , digital content use private key K ⁽ⁱ⁾encrypt also clean culture to user u ⁽ⁱ⁾;

During described digital finger-print encryption receives, the process that digital finger-print embeds and encryption distributes is as follows:

A) according to key K ^m, K ⁽ⁱ⁾the digital content data of deciphering multicast and clean culture wherein, V may be subject to collusion attack, also may not have;

B) MPEG-2 decompression is carried out: first inverse variable-length decoding (i.e. VLC is carried out to the data of deciphering ^-1decoding); Then inverse quantization Q is carried out ^-1, obtain non-overlapped 8 × 8DCT block of digital picture and video source data, and according to key obtain the position of the DCT coefficient embedding digital finger-print, finally, in conjunction with motion compensation and inverse discrete cosine transformation operation recovery MPG-2 sequence of video images;

D) combine orthogonal matrix according to the CDMA/TDMA of formula (11) and demodulation is carried out to digital fingerprint data;

E) the symmetrical structure fingerprint based on Fig. 6 " father-child node " relation is set, and utilizes mould 2 to operate, and calculates odd even, extract the digital finger-print watermark data Y embedded to the sub-band coefficients of every grade of tree structure ⁽ⁱ⁾;

F) similarity measurement of digital finger-print is carried out with normalizated correlation coefficient defined formula (1);

G) according to statistic correlation function formula (15)-(17), the contingent collusion attack situation of detector tree, and export the last index detecting collusion attack.