CN104092564A - Cloud storage service credit evaluation method - Google Patents

Abstract

A cloud storage service credit evaluation method comprises two stages and six steps. At the first stage, server node credit evaluation is conducted, and the method includes the specific steps that first, the direct trust value TD (Ci, Sj) of Ci to Sj is calculated; second, the recommendation trust value TR (Ci, Sj) of Ci to Sj is calculated; third, the finial trust value T (Ci, Sj) of Ci to Sj is calculated; fourth, the system trust value T (Sj) of a server node Sj is calculated at the stage. At the second stage, user node credit evaluation is conducted, and the method includes the specific steps that fifth, the system trust value T (Ci) of a user node Ci is calculated; sixth, trust value updating is carried out. According to the cloud storage service credit evaluation method, technological evidence is used as a trust evaluation source, credit evaluation is conducted on the user node based on user behavior, the data integrity verification result is used as a CSP storage service trust characteristic so that the CSP storage service can be measured, and therefore the objective fairness and verifiability of the trust measurement are guaranteed.

Description

A kind of cloud stores service credit assessment method

Technical field

The present invention relates to a kind of cloud stores service credit assessment method, under cloud storage environment, by data integrity, detect server node and user node are carried out to credit rating, belong to cloud computing field.

Background technology

Cloud storage is a kind of online memory module, be that user's (client) and server (high in the clouds) are by certain agreement, the outsourcing data of oneself are stored in to high in the clouds, and this emerging storage mode, has flexibility, low cost, the extensibility of cloud computing.User can access high in the clouds whenever and wherever possible, obtains the data of oneself; User pays according to the own actual memory space using, and has reduced the maintenance of data and the cost of purchase memory device, and can expand according to the needs of storage., there is increasing cloud service provider (Cloud Service Provider, CSP) in the development along with cloud storage, a large amount of dissimilar cloud stores service is being provided.Because user is stored in high in the clouds by data, lost physically the control to data, the data that are stored in high in the clouds may be tampered, and delete etc.Therefore, the quality of the prestige of cloud service provider and the cloud storage that they provide becomes the emphasis that user is concerned about, so the cloud stores service that need to provide according to CSP is carried out just credit rating to them.

The method of the existing credit appraisal for cloud stores service is divided into two kinds substantially: a kind of is by the capacity of cloud storage, speed, and the features such as price are analyzed contrast, propose corresponding evaluation method and mark to CSP; Another kind be user used CSP to provide cloud service after give a mark or vote, then these subjective assessments are carried out to polymerization analysis.There are some defects in these evaluation methods: 1) user is subjective assessment to the evaluation of CSP to a certain extent, merely subjective assessment carried out to polymerization analysis and will certainly have evaluation prejudice; 2) only considered the evaluation of user to CSP, and mostly ignored user, also may have the situation of dishonest conduct, for example malice is evaluated, false ballot, lazy node etc.; 3) do not relate to the core of CSP stores service tolerance: the integrality that is stored in high in the clouds data.So propose a kind of cloud stores service credit assessment method of objective and fair, seem particularly important.

In order to realize the fair appraisal to cloud stores service, the present invention utilizes objective technology evidence as the source of trust evaluation, and the behavior based on user is simultaneously carried out credit rating to user node.We utilize the core feature that cloud stores service is measured: the result of data integrity checking is measured CSP stores service as the trust feature of CSP stores service, thereby guarantee the objectivity of trust metrics.By analysis user, the participative behavior in trust evaluation reciprocal process is evaluated the prestige of user node simultaneously, to prevent maliciously evaluating and detect lazy node etc., guarantees the fairness of credit rating, objectivity and verifiability.

Summary of the invention

(1) goal of the invention

The object of the invention is to propose a kind of cloud stores service credit assessment method, according to the core feature of CSP stores service tolerance, the result of data integrity checking carries out as the source of trust evaluation the credit appraisal that can verify to CSP; Meanwhile, the behavior based on user is carried out credit rating to user node, can effectively resist false evaluation, lazy node, guarantees fairness, objectivity and the verifiability of credit rating.

(2) technical scheme

In order to achieve the above object, the source that the present invention has used the result of data integrity checking to calculate as trust value, its technical scheme is as follows.

The cloud storage trust model the present invention relates to comprises two network entities: 1) store data in individual or the enterprise institution in high in the clouds, i.e. user; 2) there is the cloud service provider of special resource and computing capability.

According to the execution phase, it can divide two stages: the stage one in the present invention: server node credit rating is the CSP stores service credit appraisal stage; Stage two: user node credit rating is the user node credit rating stage based on user behavior.

A kind of cloud stores service of the present invention credit assessment method, the technical scheme of the method is as follows:

Stage one: server node credit rating: comprise step 1～step 4, respectively computing client end node C _ito server node S _jdirect trust value (only pass through C _iself and S _jthe trust value that interactive computing obtains) T _d(C _i, S _j) and recommendation trust (by other nodes and S _jthe trust value that interactive computing obtains) T _r(C _i, S _j); By direct trust value and recommendation trust, calculate C _ito S _jfinal trust value T (C _i, S _j); The all client node of polymerization is to S again _jtrust value obtain server node S _jsystem trust value T (S _j).

Step 1: calculate C _ito S _jdirect trust value T _d(C _i, S _j): to carry out each time the result (provide { 0,1} evaluates, and 1 represents to be verified, and data are complete, and 0 represents that checking do not pass through, and data are imperfect) of data integrity checking, be input, the desired value characteristic of utilizing beta to distribute completes C _ito S _jdirect trust value tolerance.

Step 2: calculate C _ito S _jrecommendation trust T _r(C _i, S _j): by with server node S _jmutual other client nodes C crossing _r∈ ID (S _j, C _i) to server node S _jthe comprehensive measurement of trusting degree calculate C _ito S _jrecommendation trust: we are with C _rto S _jdirect trust value T _d(C _r, S _j) as C _rto S _jtrusting degree, trust weight is defined as to C _rand C _ithe similarity of the server node of their common mutual mistake being carried out to the result of data integrity checking, the data integrity evaluation result that they provide is more similar, and the trusting degree between them is just higher, with root mean square RMS (C _r, C _i) identify.

Step 3: calculate C _ito S _jfinal trust value T (C _i, S _j): the direct trust value that step 1 and 2 is obtained and recommendation trust carry out polymerization and calculate C _ito S _ifinal trust value.

Step 4: calculation server node S _jsystem trust value T (S _j): by all client node C _i(i=1to m) is to S _jtrust value T (C _i, S _j) carry out polymerization, its weight definition is C _itrust value T (C _i), finally obtain all S _jthe system trust value of (j=1to n).

Stage two: user node credit rating: the credit value of the server node calculating according to the first stage, arranges a decision threshold T _thjudge S _jwhether be honest server node.Then according to the behavior of user node, its prestige is added up, accumulative total rule is: carrying out integrity verification result with honest server node is 1 and to carry out integrity verification result with dishonest server node be 0 to be all considered as dishonest behavior, interbehavior trust value accumulative total; Other are all considered as dishonest conduct, and interbehavior trust value does not add up.The trust value of server node and user node is carried out to iteration renewal simultaneously, make it reach stable state.

Step 5: calculate user node C _isystem trust value T (C _i): first determine and C _iset I (the C of the mutual server node of crossing _i), then judge C _iwith I (C _i) in the history of each element interactions, according to mutual trust accumulative total rule, trust value add up, then will trust aggregate-value and server node and utmost good faith/non-honest degree of closeness P _jcarry out product weighting, just can calculate user node C _isystem trust value T (C _i).

Step 6: belief updating: in step 4, we will use the user node C also not calculating _isystem trust value T (C _i), still be provided with T (C _i) initial value T _default.In this step, we need to calculate the server node of stable state and the trust value of user node according to above calculation procedure the mode by iterative computation, constantly the system trust value of server node and user node is carried out to iterative computation, make its value reach convergence.

Wherein, at " the calculating C described in step 1 _ito S _jdirect trust value T _d(C _i, S _j) ", its computational process is as follows: n wherein _ijfor client node C _iwith server node S _jcarry out the total degree of integrity verification; F _pDP(C _i, S _j, k) be client node C _iwith server node S _jcarry out the result of the k time integrity verification; VEF (C _i, S _j, k) be client node C _iwith server node S _jenvirment factor while carrying out the k time integrity verification.

Wherein, at " the calculating C described in step 2 _ito S _jrecommendation trust T _r(C _i, S _j) ", its computational process is as follows: $T_R(C_i,S_j)={\mathrm{\Σ}}_{C_r\∈\mathrm{ID}(S_j,C_i)}{T}_D(C_r,S_j)\·\frac{\mathrm{RMS}(C_r,C_i)}{{\mathrm{\Σ}}_{C_q\∈\mathrm{ID}(S_j,C_i)}\mathrm{RMS}(C_q,C_i)},$ ID (S wherein _j, C _i) be I (S _j) remove C in set _ithe set of node; RMS (C _r, C _i) be root mean square, be used for being expressed as C _ito C _rtrusting degree, its computational methods are as follows: $\mathrm{RMS}(C_r,C_i)=1-\sqrt{\frac{{\mathrm{\Σ}}_{y\∈\mathrm{SS}(C_r,C_i)}{(\frac{{\mathrm{\Σ}}_{k=1}^{N(C_r,y)}{F}_{\mathrm{PDP}}(C_r,y,k)}{N(C_r,y)}-\frac{{\mathrm{\Σ}}_{k=1}^{N(C_i,y)}{F}_{\mathrm{PDP}}(C_i,y,k)}{N(C_i,y)})}^2}{\left|\mathrm{SS}(C_r,C_i)\right|}},$ SS (C wherein _r, C _i) be and client node C _rand C _iall carried out the server node set of integrity verification.

Wherein, at " the calculating C described in step 3 _ito S _jfinal trust value T (C _i, S _j) ", its computational process is as follows: T (C _i, S _j)=α T _d(C _i, S _j)+β T _r(C _i, S _j), wherein α and β are respectively the weight of direct trust value and recommendation trust.

Wherein, at " the calculation server node S described in step 4 _jsystem trust value T (S _j) ", its computational process is as follows: $T\left(S_j\right)={\mathrm{\Σ}}_{i=1}^{m}T(C_i,S_j)\·\frac{T\left(C_i\right)}{{\mathrm{\Σ}}_{j=1}^{m}T\left(C_j\right)}.$

Wherein, at " the calculating user node C described in step 5 _isystem trust value T (C _i) ", its computational process is as follows: i (C wherein _i) be and C _ithe set of the mutual server node of crossing; ? be a set, in set, the number of element is | I (C _i) |, representative is and C respectively _ithe numbering of mutual server node; a _ijfor client node C _iwith server node S _jcan be to T (C in interaction results _i) accumulative total trust value; P _jfor server node and utmost good faith/non-honest degree of closeness.

(3) advantage and effect

A kind of cloud stores service of the present invention credit assessment method, its advantage with effect is: 1) compare with existing Cloud Server credit rating model, this model not only can carry out credit rating to Cloud Server node, also can carry out credit rating to user node, can avoid false and evaluate, resist lazy node; 2) different from the reputation model based on subjective assessment, the present invention is based on objective technology evidence, and the result that data integrity detects is carried out the calculating of trust value, makes model have stronger objectivity and fairness; 3) model finally can reach stable state, and the trust value of each node calculating all reaches convergence, makes model have correctness and stability; 4) be provided with communication environment factor variable, for the significance level difference of data integrity proof procedure each time, the weight of this communication be weighted, can effectively avoid malice evaluation.

Accompanying drawing explanation

Fig. 1 is FB(flow block) of the present invention

Fig. 2 is system architecture diagram of the present invention

Embodiment

The present invention includes two entities, user and cloud server.Below with reference to accompanying drawing, described data are held to effective verification method and be described in detail, Fig. 1 is FB(flow block) of the present invention; Fig. 2 is system architecture diagram of the present invention;

Main symbol and algorithmic translation:

(1) F _pDP(C _i, S _j, k): client node C _iwith server node S _jcarry out the k time PDP (Provable Data Possession, the data property held proves) the result;

(2) N (C _i, S _j): client node C _iwith server node S _jcommunicate the total degree of (PDP checking), can note by abridging as n _ij;

N _ij1: client node C _iwith server node S _jcarrying out PDP checking, to provide the result be 1 number of times

N _ij2: client node C _iwith server node S _jcarrying out PDP checking, to provide the result be 0 number of times

N _ij3: client node C _iwith server node S _jtherefore carrying out PDP checking, to provide the result be that the number of times of NULL has: n _ij=n _ij1+ n _ij2+ n _ij3;

(3) VEF (C _i, S _j, k): Verification Environment Factor, client node C _iwith server node S _jenvirment factor while carrying out the k time PDP checking.Effect: can preventing malice attack, for example very sincere in little transaction, promotes own prestige, thereby and in concluding the business greatly non-honest profit;

(4) I (S _j): with server node S _jcarried out the set of the client node of PDP checking;

(5) ID (S _j, C _i): I (S _j) remove C in set _ithe set of node;

(6) RMS (C _r, C _i): root-mean-square, root mean square, is used for comparison node C _iand C _revaluation difference to the server node of their common mutual mistake, is expressed as C _ito C _rtrusting degree, if evaluation situation is similar, their root-mean-square value is high, trusting degree is high, otherwise trusting degree is low;

(7) SS (C _r, C _i): with client node C _rand C _iall carried out the server node set of PDP checking;

(8) I (C _i): with C _ithe set of the mutual server node of crossing;

(9) i.e. definition be a set, in set, the number of element is | I (C _i) |, representative is and C respectively _ithe numbering of mutual server node;

(10) $a_{\mathrm{ij}}=\left\{\begin{array}{ccc}\frac{n_{\mathrm{ij}1}}{n_{\mathrm{ij}}},& \mathrm{if}& T\left(S_j\right)\&GreaterEqual;T_{\mathrm{th}}\\ \frac{n_{\mathrm{ij}2}}{n_{\mathrm{ij}}},& \mathrm{if}& T\left(S_j\right)<T_{\mathrm{th}}\end{array}\right.$ : define a _ijfor client node C _iwith server node S _jcan be to T (C in interaction results _i) accumulative total trust value;

(11) $P_j=\left\{\begin{array}{ccc}T\left(S_j\right),& \mathrm{if}& T\left(S_j\right)\&GreaterEqual;T_{\mathrm{th}}\\ 1-T\left(S_j\right),& \mathrm{if}& T\left(S_j\right)<T_{\mathrm{th}}\end{array}\right.$ : P _jfor server node and utmost good faith/non-honest degree of closeness.

Technical scheme can be divided into server node credit rating and two stages of user node credit rating.A kind of cloud stores service of the present invention credit assessment method, the method concrete steps (comprising specific formula for calculation and algorithm) are as follows:

The 1.CSP stores service credit appraisal stage

Step 1: calculate C _ito S _jdirect trust value T _d(C _i, S _j)

Calculating direct trust value should be by client node C _iwith server node S _jinteractive history obtain, due to client node C _iwith server node S _jmay be alternately repeatedly, so will carry out each time result (provide { 0, the 1} evaluation) F of PDP checking _pDP(C _i, S _j, k) be multiplied by verification environment factor VEF (C each time _i, S _j, k), then carry out polymerization and normalized.If refusal provides PDP the result, { 0,1} evaluates, so this result F _pDP(C _i, S _j, k) be designated as 0, that is to say while being only verified client node C _ito server node S _jdirect trust value just add up, checking not by or refusal provide the result, direct trust value does not add up.

In this programme, we utilize beta to distribute to measure C _ito S _jdirect trust value, it is the posterior probability of binary file that beta distributes, what its expectation was described is the probability that after observing, result system event occurs, so the desired value that we distribute with beta is as internodal direct trust value.We use represent the parameter alpha in beta distribution _ij, use n _ij-α _ijrepresent the parameter beta in beta distribution _ij, the expectation of the beta probability distribution of carrying out with such parameter is expressed as client node C _ito server node S _jdirect trust value.

$\begin{array}{c}f\left(X\right|\mathrm{\&alpha;},\mathrm{\&beta;})=\frac{X^{{\mathrm{\&alpha;}}_{\mathrm{ij}}-1}{(1-X)}^{{\mathrm{\&beta;}}_{\mathrm{ij}}-1}}{B({\mathrm{\&alpha;}}_{\mathrm{ij}},{\mathrm{\&beta;}}_{\mathrm{ij}})},\mathrm{where}{a}_{\mathrm{ij}}=\underset{k=1}{\overset{n_{\mathrm{ij}}}{\mathrm{\&Sigma;}}}{F}_{\mathrm{PDP}}(C_i,S_j,k)\&CenterDot;\mathrm{VEF}(C_i,S_j,k),{\mathrm{\&beta;}}_{\mathrm{if}}\\ =n_{\mathrm{ij}}-{\mathrm{\&alpha;}}_{\mathrm{ij}}\end{array}$

B (α wherein _ij, β _ij) be beta function, its expression formula is so client node C _ito server node S _jdirect trust value be:

$T_D(C_i,S_j)=E\left(X\right)=\frac{{\mathrm{\&alpha;}}_{\mathrm{ij}}}{{\mathrm{\&alpha;}}_{\mathrm{ij}}+{\mathrm{\&beta;}}_{\mathrm{ij}}}=\frac{1}{n_{\mathrm{ij}}}\underset{k=1}{\overset{n_{\mathrm{ij}}}{\mathrm{\&Sigma;}}}{F}_{\mathrm{PDP}}(C_i,S_j,k)\&CenterDot;\mathrm{VEF}(C_i,S_j,k)$

Step 2: calculate C _ito S _jrecommendation trust T _r(C _i, S _j)

Computing client end node C _ito server node S _jrecommendation trust should by with server node S _jmutual other client nodes C crossing _r∈ ID (S _j, C _i) to server node S _jtrusting degree comprehensive, so we are with C _rto S _jdirect trust value as C _rto S _jtrusting degree, weight should be C so _ito C _rtrusting degree, and do not communicate by letter mutually between client node, therefore its trusting degree can be by analyzing client node C _rand C _ithe similarity of the evaluation of the server node of their common mutual mistake being carried out to the result of PDP checking compares, if they are very similar to the result that these server nodes carry out PDP checking, trust value between them is just high so, on the contrary with regard to low (Things of a kind come together).If be made with like this benefit, be exactly that a node belongs to malicious node set, and a node belongs to good will node set, so they to provide the similarity of result just very low, trust value between them is just very low, has so just avoided to a certain extent the dishonest recommendation between these two nodes.Computing formula is as follows:

$T_R=(C_i,S_j)={\mathrm{\&Sigma;}}_{C_r\&Element;\mathrm{ID}(S_j,C_i)}{T}_D(C_r,S_j)\&CenterDot;\frac{\mathrm{RMS}(C_r,C_i)}{{\mathrm{\&Sigma;}}_{C_q\&Element;\mathrm{ID}(S_j,C_i)}\mathrm{RMS}(C_q,C_i)}$

$\mathrm{RMS}(C_r,C_i)=1-\sqrt{\frac{{\mathrm{\&Sigma;}}_{y\&Element;\mathrm{SS}(C_r,C_i)}{(\frac{{\mathrm{\&Sigma;}}_{k=1}^{N(C_r,y)}{F}_{\mathrm{PDP}}(C_r,y,k)}{N(C_r,y)}-\frac{{\mathrm{\&Sigma;}}_{k=1}^{N(C_i,y)}{F}_{\mathrm{PDP}}(C_i,y,k)}{N(C_i,y)})}^2}{\left|\mathrm{SS}(C_r,C_i)\right|}}$

Step 3: calculate C _ito S _jrecommendation trust T _r(C _i, S _j)

On the basis of step 1 and step 2, direct trust value and recommendation trust are carried out to polymerization, its weight is respectively α and β.At different net environments, the requirement of subjective evaluation be there are differences, α and β also just can suitably adjust according to corresponding environmental condition and requirement.Computing formula is as follows:

T(C _i，S _j)＝α·T _D(C _i，S _j)+β·T _R(C _i，S _j)

Step 4: calculation server node S _jsystem trust value T (S _j)

Final calculation server node S _jtrust value T (S _j): should be by all client node C _i(i=1to m) is to S _jtrust value T (C _i, S _j) carry out polymerization, by its weight definition, be C _itrust value T (C _i), then divided by be normalized, computing formula is as follows:

$T\left(S_j\right)=\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}T(C_i,S_j)\&CenterDot;\frac{T\left(C_i\right)}{{\mathrm{\&Sigma;}}_{j=1}^{m}T\left(C_j\right)}$

2. user node credit rating stage based on user behavior

Step 5: calculate user node C _isystem trust value T (C _i)

In the stage 1, we can calculate the trust value of all server nodes.According to specific network condition, a decision threshold T is set _thand judge: if T is (S _j)>=T _th, think S _jhonest server node, if T is (S _j) < T _th, think S _jit is dishonest server node.Then determine and C _iset I (the C of the mutual server node of crossing _i), then judge C _iwith I (C _i) in the history of each element interactions, in reciprocal process, to only have PDP interaction results with honest server node be 1 and can be accumulated in trust value with the behavior that the PDP interaction results of dishonest server node is 0, and other behaviors are all considered as dishonest conduct and can not be accumulated in trust value.Then we have defined server node S _jdegree of closeness P with utmost good faith/non-honest node _j, then will trust aggregate-value P _jcarry out product weighting, just can calculate user node C _isystem trust value T (C _i).Computing formula is as follows:

$T\left(C_i\right)=\frac{1}{\left|I\left(C_i\right)\right|}\stackrel{q=1}{\mathrm{\&Sigma;}}{a}_{i{\mathrm{\&alpha;}}_{C_i}^q}\&CenterDot;P_{{\mathrm{\&sigma;}}_{C_i}^q}$

Step 6: belief updating

By above calculating, we can find out, the trust value of server node depends on the prestige situation of history mutual information and user node, and meanwhile, the trust value of user node depends on the prestige situation of history mutual information and server node.So we will use the user node C also not calculating in step 4 _isystem trust value T (C _i), still be provided with T (C _i) initial value T _defaultso, carrying out in the computational process of step 4 for the first time T (C _i)=T _default, then by step 4,5, calculate all server nodes and the credit value of user node.But because calculating once can not reach the effect that a stable state is trusted, so in this stage, we need to calculate by the mode of iterative computation the server node of stable state and the trust value of user node, constantly the system trust value of server node and user node is carried out to iterative computation, make its value reach convergence.Server node belief updating algorithm and user node belief updating algorithm are as follows:

Algorithm 1 is for the trust value of iteration update server node, and in order conveniently to carry out calculation specifications, we are by the trust value T (S of server node ₁), T (S ₂) ..., T (S _n) remember into vector T _s={ T (S ₁), T (S ₂) ..., T (S _n) ^t.Algorithm concrete steps are:

1. give the trust value initialize T of user node _default;

2. according to given initial value T _default, by step 4 and step 5, calculate next round

3. constantly carry out iterative computation, until trust value reaches convergence,

Algorithm 2 upgrades the trust value of user node for iteration, in order conveniently to carry out calculation specifications, we are by the trust value T (C of client node ₁), T (C ₂) ..., T (C _m) remember into vector T _c={ T (C ₁), T (C ₂) ..., T (C _m) ^t.Algorithm concrete steps are:

1. give the trust value initialize T of user node _default;

2. according to given initial value T _default, by step 4 and step 5, calculate next round

3. constantly carry out iterative computation, until trust value reaches convergence,

Claims (6)

1. a cloud stores service credit assessment method, is characterized in that: the method concrete steps are as follows:

Stage one: server node credit rating: comprise step 1～step 4, respectively computing client end node C _ito server node S _jdirect trust value T _d(C _i, S _j) and recommendation trust T _r(C _i, S _j); By direct trust value and recommendation trust, calculate C _ito S _jfinal trust value T (C _i, S _j); The all client node of polymerization is to S again _jtrust value obtain server node S _jsystem trust value T (S _j);

Step 1: calculate C _ito S _jdirect trust value T _d(C _i, S _j): the result of carrying out each time data integrity checking of take is input, provide { 0,1} evaluates, and 1 represents to be verified, and data are complete, and 0 represents that checking do not pass through, and data are imperfect, and the desired value characteristic of utilizing beta to distribute completes C _ito S _jdirect trust value tolerance;

Step 2: calculate C _ito S _jrecommendation trust T _r(C _i, S _j): by with server node S _jmutual other client nodes C crossing _r∈ ID (S _j, C _i) to server node S _jthe comprehensive measurement of trusting degree calculate C _ito S _jrecommendation trust: with C _rto S _jdirect trust value T _d(C _r, s _j) as C _rto S _jtrusting degree, trust weight is defined as to C _rand C _ithe similarity of the server node of their common mutual mistake being carried out to the result of data integrity checking, the data integrity evaluation result providing is more similar, and the trusting degree between them is just higher, with root mean square RMS (C _r, C _i) identify;

Step 3: calculate C _ito S _jfinal trust value T (C _i, S _j): the direct trust value that step 1 and 2 is obtained and recommendation trust carry out polymerization and calculate C _ito S _jfinal trust value;

Step 4: calculation server node S _jsystem trust value T (S _j): by all client node C _i(i=1to m) is to S _jtrust value T (C _i, S _j) carry out polymerization, its weight definition is C _itrust value T (C _i), finally obtain all S _jthe system trust value of (j=1to n);

Stage two: user node credit rating: the credit value of the server node calculating according to the first stage, arranges a decision threshold T _thjudge S _jwhether be honest server node; Then according to the behavior of user node, its prestige is added up, accumulative total rule is: carrying out integrity verification result with honest server node is 1 and to carry out integrity verification result with dishonest server node be 0 to be all considered as dishonest behavior, interbehavior trust value accumulative total; Other are all considered as dishonest conduct, and interbehavior trust value does not add up, and the trust value of server node and user node are carried out to iteration renewal simultaneously, make it reach stable state;

Step 5: calculate user node C _isystem trust value T (C _i): first determine and C _iset I (the C of the mutual server node of crossing _i), then judge C _iwith I (C _i) in the history of each element interactions, according to mutual trust accumulative total rule, trust value add up, then will trust aggregate-value and server node and utmost good faith/non-honest degree of closeness P _jcarry out product weighting, just calculate user node C _isystem trust value T (C _i);

Step 6: belief updating: in step 4, use the user node C also not calculating _isystem trust value T (C _i), therefore be provided with T (C _i) initial value Tdefault: in this step, need to calculate the server node of stable state and the trust value of user node according to above calculation procedure the mode by iterative computation, constantly the system trust value of server node and user node is carried out to iterative computation, make its value reach convergence.

2. a kind of cloud stores service credit assessment method according to claim 1, is characterized in that: at " the calculating C described in step 1 _ito S _jdirect trust value T _d(C _i, S _j) ", its computational process is as follows: $T_D(C_i,S_j)=\frac{1}{n_{\mathrm{ij}}}{\mathrm{\&Sigma;}}_{k=1}^{n_{\mathrm{ij}}}{F}_{\mathrm{PDP}}(C_i,S_j,k)\&CenterDot;\mathrm{VEF}(C_i,S_j,k),$ N wherein _ijfor client node C _iwith server node S _jcarry out the total degree of integrity verification; F _pDP(C _i, S _j, k) be client node C _iwith server node S _jcarry out the result of the k time integrity verification; VEF (C _i, S _j, k) be client node C _iwith server node S _jenvirment factor while carrying out the k time integrity verification.

3. a kind of cloud stores service credit assessment method according to claim 1, is characterized in that: at " the calculating C described in step 2 _ito S _jrecommendation trust T _r(C _i, S _j) ", its computational process is as follows: $T_R(C_i,S_j)={\mathrm{\&Sigma;}}_{C_r\&Element;\mathrm{ID}(S_j,C_i)}{T}_D(C_r,S_j)\&CenterDot;\frac{\mathrm{RMS}(C_r,C_i)}{{\mathrm{\&Sigma;}}_{C_q\&Element;\mathrm{ID}(S_j,C_i)}\mathrm{RMS}(C_a,C_i)},$ ID (S wherein _j, C _i) be I (S _j) remove C in set _ithe set of node; RMS (C _r, C _i) be root mean square, be used for being expressed as C _ito C _rtrusting degree, its computational methods are as follows:

$\mathrm{RMS}(C_r,C_i)=1-\sqrt{\frac{{\mathrm{\&Sigma;}}_{y\&Element;\mathrm{SS}(C_r,C_i)}{(\frac{{\mathrm{\&Sigma;}}_{k=1}^{N(C_r,y)}{F}_{\mathrm{PDF}}(C_r,y,k)}{N(C_r,y)}-\frac{{\mathrm{\&Sigma;}}_{k=1}^{N(C_i,y)}{F}_{\mathrm{PDP}}(C_i,y,k)}{N(C_i,y)})}^2}{\left|\mathrm{SS}(C_r,C_i)\right|}},$ SS (C wherein _r, C _i) be and client node C _rand C _iall carried out the server node set of integrity verification.

4. a kind of cloud stores service credit assessment method according to claim 1, is characterized in that: at " the calculating C described in step 3 _ito S _jfinal trust value T (C _i, S _j) ", its computational process is as follows: T (C _i, S _j)=α T _d(C _i, S _j)+β T _r(C _i, S _j), wherein α and β are respectively the weight of direct trust value and recommendation trust.

5. a kind of cloud stores service credit assessment method according to claim 1, is characterized in that: at " the calculation server node S described in step 4 _jsystem trust value T (S _j) ", its computational process is as follows:

6. a kind of cloud stores service credit assessment method according to claim 1, is characterized in that: at " the calculating user node C described in step 5 _isystem trust value T (C _i) ", its computational process is as follows: i (C wherein _i) be and C _ithe set of the mutual server node of crossing; ? be a set, in set, the number of element is | I (C _i) |, representative is and C respectively _ithe numbering of mutual server node; a _ijfor client node C _iwith server node S _jcan be to T (C in interaction results _i) accumulative total trust value; P _jfor server node and utmost good faith/non-honest degree of closeness.