CN106296219A - A kind of method trusted based on AIFS weight calculation - Google Patents

Abstract

The present invention relates to a kind of method trusted based on AIFS weight calculation.A cloud service provider is selected to interact；User asks a grade of service to cloud service provider according to demand；Wherein, the grade of service of cloud service provider includes n service indication, and is expressed as by the form of one group of service indication vector；Then, AIFS weighing computation method is used to determine the weight of service indication vector；Cloud service provides according to the requested grade of service, provides cloud service to this request；This service is made evaluation according to the cloud service obtained by user, obtains the actual measured value of service indication vector；And then according to the weight of the service indication vector determined, calculate user's evaluation of estimate to cloud service provider；The cloud service evaluation that cloud service provider is provided by all users determines the degrees of comparison of cloud service provider, i.e. carries out global cloud service provider's reputation and calculates, it is thus achieved that the overall credit value of cloud service provider.The present invention gives one and user is evaluated cloud service provider service result cloud service provider the most objective, believable optimum selecting method.

Description

A kind of method trusted based on AIFS weight calculation

Technical field

The present invention relates to a kind of method trusted based on AIFS weight calculation.

Background technology

Trust and play an important role in business cloud environment.It is one of ultimate challenge of facing of cloud.Trust makes User can select optimal resource in the cloud infrastructure of an isomery.Safety of cloud service is assessed also before selecting cloud service It it is one of them necessary aspect.Therefore, trust model Trust Model (TM) takes as a security intensity assessment and classification Business cloud application and service.It can set up changing of the safety of cloud service, discovery deficiency and cloud infrastructure as a benchmark Kind.Document [1] research finds, safety problem is inseparable with trust problem, and trust model TM is an effective method, Increasing focus of attention propose method and be applied to cloud computing security fields, wherein, select believable cloud service in recent years Provider (Cloud Service Provider, CSP) is also study hotspot^[2].Document [3] introduces a new trust mould Type, according to authority and the ability of current cloud resource provider in past.Trust value computing uses four parameters, such as availability, Reliability, week transfer efficient and data integrity, describe the quality of service requirement that the preparation of how SLA combines User and the ability of cloud resource provider, the final concept proposing similar trust model.Document [3] is under cloud manufacturing environment The problem causing user to be difficult to obtain suitable cloud service because there is the same or analogous manufacture cloud service of a large amount of function, it is proposed that Problem has been carried out abstract by a kind of manufacture cloud service system of selection based on credible evaluation: by reliability, and availability is ageing, Price and credibility include credible feature set in, and consider evaluation time, and the credibility of estimator tentatively solves credible cloud service The problem selected.Document [4] proposes in service-oriented cloud computing QoS evaluation methodology accurately: use Fuzzy Synthetic Decision Assessment cloud service provider, according to the hobby of user, then uses QoS Monitoring Data cloud clothes on the basis of the cloud model of cloud computing The uncertainty of business.Finally, fuzzy logic control QoS evaluation result is drawn.In the research that cloud service provider selects, document [5] being combined with trust model by service level criteria SLA first, comprehensive many property calculation CSP trusts, and selects believable for user CSP provides solution.

Document [6] proposes a framework and algorithm in terms of cloud service selects cloud and user's service preferences, by application The expertise of reasoning, the Combinatorial Optimization combination of fuzzy logic, solve the involutory cloud service select permeability accommodating hobby of user.

Report [7] is pointed out, the security attribute in cloud environment can be characterized as a series of necessary autonomous system variable.These Variable is generally by the particular variables needed for autonomous system.Each security factor, based on fuzzy logic, is decomposed into by document [8] Pass judgment on collection, establish the method that users to trust is followed the trail of by CSP, user behavior is analyzed, follow the trail of such that it is able to stable supervision With the hacker found out in cloud environment.Above it can be seen that the advantage of fuzzy algorithmic approach be the impact that can reduce uncertain factor with And self-recision: multiattribute service value can be adapted to admirably, such as: price, expense, service quality, network delay etc..

Combining it, document above mainly studies the relevant issues that cloud service selects, but does not utilizes fuzzy from weight calculation angle Theory solves the trust problem of cloud service.Therefore, portray relevant to Fuzzy Set Theory herein by trust metrics model Safety problem, utilize AIFS (Antanassov intuitionistic fuzzy set) fuzzy set preferably to process The ambiguity of problem, uncertainty, set forth herein a kind of trust model selected based on AIFS optimal weights.Finally for Family selects credible cloud service provider problem to provide solution and emulation experiment.

List of references:

[1]Bera S,Misra S,Rodrigues J P C.Cloud computing applications for smart grid:A survey[J].IEEE Transactions on Parallel and Distributed Systems, 2015,26(5):1477-1494.

[2]Manuel P.A trust model of cloud computing based on Quality of Service[J].Annals of Operations Research,2015,233(1):281-292.

[3] Tao Fei, Zhang Lin, Guo Hua, etc. cloud manufacturing feature and cloud service combination key issue research [J]. Automated library system Manufacture system, 2011,17 (3): 477-486.

[4]Wang S,Liu Z,Sun Q,et al.Towards an accurate evaluation of quality of cloud service in service-oriented cloud computing[J].Journal of Intelligent Manufacturing,2014,25(2):283-291.

[5] high cloud fine jade, Shen Beijun, Kong Huafeng. the cloud computing trust model [J] evaluated with user based on SLA. computer work Journey, 2012,38 (7): 28-30.

[6]Dastjerdi A V,Buyya R.Compatibility-aware cloud service composition under fuzzy preferences of users[J].IEEE Transactions on Cloud Computing,2014,2(1):1-13.

[7]Feng D G,Zhang M,Zhang Y,et al.Study on cloud computing security [J].Journal of software,2011,22(1):71-83.

[8]Jaiganesh M,Aarthi M,Kumar A V A.Fuzzy ART-based user behavior trust in cloud computing[M].Artificial Intelligence and Evolutionary Algorithms in Engineering Systems.Springer India,2015:341-348.

[9]Xu Z,Cai X.Intuitionistic fuzzy information aggregation[M] .Intuitionistic Fuzzy Information Aggregation.Springer Berlin Heidelberg, 2012:1-102.

[10]Wang W,Liu X.Intuitionistic fuzzy information aggregation using Einstein operations[J].IEEE Transactions on Fuzzy Systems,2012,20(5):923-938.

[11] Zhang Lin, Rao Kaili, Wang Ruchuan. based on the dynamic trust evaluation model evaluating credibility under cloud computing environment [J]. communication journal, 2013:31-37..

Summary of the invention

It is an object of the invention to provide a kind of method trusted based on AIFS weight calculation, user is evaluated cloud by the method Service provider service result is the most objective, credible.

For achieving the above object, the technical scheme is that a kind of method trusted based on AIFS weight calculation, including Following steps,

S1: select a cloud service provider to interact；

S2: user asks grade of service l to cloud service provider according to demand；Wherein, the clothes of cloud service provider Business grade x includes n service indication k_i ^l, and be expressed as by the form of one group of service indication vector According to the difference of grade of service x, k_i ^lService indication value the most different；

S3: use AIFS weighing computation method to determine the weight of service indication vector in step S2；

S4: cloud service provides according to requested grade of service l, provides cloud service to this request；

This service is made evaluation according to the cloud service obtained by S5: user, obtains the actual measurement of service indication vector Value；And then the weight of the service indication vector determined according to step S3, calculate user's evaluation of estimate to cloud service provider；

The cloud service evaluation that cloud service provider is provided by S6: all users determines the degrees of comparison of cloud service provider, I.e. carry out global cloud service provider's reputation to calculate, it is thus achieved that the overall credit value of cloud service provider.

In an embodiment of the present invention, described step S3 use AIFS weighing computation method determine that in step S2, service refers to The weight of mark vector to implement process as follows:

Definition Einstein is long-pending, Einstein and respectively with T_ε(a,b)、S_ε(a, b) represents:

$T_{\mathrm{\ϵ}}(a,b)=\frac{a\·b}{1+(1-a)\·(1-b)}---\left(1\right)$

$S_{\mathrm{\ϵ}}(a,b)=\frac{a+b}{1+a\·b}---\left(2\right)$

Wherein, a, b ∈ [0,1], (a b) is support amount and the opposition amount of fuzzy set；

Defining one group of IFS set expression is α=(μ_α, v_α), μ_αFor the support to α variable, v_αFor not propping up α variable Hold degree；The IFS decision value making A, B be decision object x, and A=(μ_A(x),v_A(x)), B=(μ_B(x),v_B(x))

Can obtain according to formula (1), (2), Einstein of AIFS is long-pending, Einstein and computing can be expressed as follows:

$A\⊗B=\{T_{\mathrm{\ϵ}}\left({\mathrm{\μ}}_A\right(x),v_A(x\left)\right),S_{\mathrm{\ϵ}}\left({\mathrm{\μ}}_A\right(x),v_A(x\left)\right)\}$

A B={S_ε(μ_A(x),v_A(x)),T_ε(μ_A(x),v_A(x))}

That is:

$A\⊗B=\{\frac{{\mathrm{\μ}}_A\left(x\right)\·{\mathrm{\μ}}_B\left(x\right)}{1+(1-{\mathrm{\μ}}_A(x\left)\right)\·(1-{\mathrm{\μ}}_B(x\left)\right)},\frac{v_A\left(x\right)+v_B\left(x\right)}{1+v_A\left(x\right)\·v_B\left(x\right)}\}---\left(3\right)$

$A\⊕B=\{\frac{{\mathrm{\μ}}_A\left(x\right)+{\mathrm{\μ}}_B\left(x\right)}{1+{\mathrm{\μ}}_A\left(x\right)\·{\mathrm{\μ}}_B\left(x\right)},\frac{v_A\left(x\right)\·v_B\left(x\right)}{1+(1-v_A(x\left)\right)\·(1-v_B(x\left)\right)}\};---\left(4\right)$

Suppose there is real number λ, then have a following multinomial multiplying:

$\mathrm{\λ}\·A=\{\frac{{\[1+{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}-{\[1-{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}}{{\[1+{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}+{\[1-{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}},\frac{2\·{\[v_A\left(x\right)\]}^{\mathrm{\λ}}}{{\[2-v_A\left(x\right)\]}^{\mathrm{\λ}}+{\[v_A\left(x\right)\]}^{\mathrm{\λ}}}\}---\left(5\right)$

Owing to there being n service indication, i.e. n decision attribute, orderFor AIFS collection value, wherein, j= 1,2,...,n；W=(w₁,w₂,...,w_n)^TForWeight vectors, and Σ w_j=1, according to (3), (4), (5) polymerizing value of weight can be obtained, and polymerizing value is also an IFS value:

$IFW=\{\frac{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{w_j}-{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{w_j}}{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{w_j}+{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{w_j}},\frac{2\·{\Π}_{j=1}^n{v_{a_j}}^{w_j}}{{\Π}_{j=1}^n{(2-v_{a_j})}^{w_j}+{\Π}_{j=1}^n{v_{a_j}}^{w_j}}\}---\left(6\right)$

By formula (6), the support of the IFS value of each decision scheme that cloud service provider policymaker provides can be calculated Amount, the maximum scheme of support amount is optimal case, and the vectorial weight proposed using the program is as the power of service indication vector Weight.

In an embodiment of the present invention, described step S5 to implement process as follows:

This service is made evaluation according to the cloud service obtained by user, obtains service indication vectorActual measured value (a₁,a₂,a₃,a₄,...,a_n), thus can service inclined difference sigma:

$\mathrm{\σ}=\underset{i=1}{\Σ}({k_i}^l-a_i)\·w_i---\left(7\right)$

If service indication k_i ^lActual measured value higher than service indication value, then k_i ^l-a_i=0；

User evaluates the not amount of support of cloud service provider and calculates with statistical method:

User evaluates the IFS of cloud service provider and gathers (μ_a,v_a)=(1-v_a,v_a) (9)。

In an embodiment of the present invention, described step S6 to implement process as follows:

For overall N number of user, can be considered that the secondary AIFS polymerization that each user's weight is 1/N calculates, by each use (the μ at family_a,v_a) substitute in formula (6), available global cloud service provider's reputation value T_CSP:

$T_{CSP}=\{\frac{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{1/N}-{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{1/N}}{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{1/N}+{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{1/N}},\frac{2\·{\Π}_{j=1}^n{v_{a_j}}^{1/N}}{{\Π}_{j=1}^n{(2-v_{a_j})}^{1/N}+{\Π}_{j=1}^n{v_{a_j}}^{1/N}}\}---\left(10\right)$

The T that will obtain_CSPSupport amount percentage, available final cloud service provider overall reputation value is

$100*\frac{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{1/N}-{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{1/N}}{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{1/N}+{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{1/N}}---\left(11\right).$

Compared to prior art, the method have the advantages that

The method of the present invention, by setting up rational trust model and method for evaluating trust, provides a kind of credible cloud clothes The system of selection of business provider, the method can dynamically adjust user's power in global assessment from the angle of overall situation prestige Weight, rationally passes judgment on the prestige of CSP, reaches to select as far as possible the purpose of credible CPS；Test result indicate that, this method can be preferable Ground solves credible CSP select permeability

Accompanying drawing explanation

Fig. 1 is trust model schematic diagram of the present invention.

Fig. 2 is overall situation reputation comparison diagram.

Fig. 3 is accuracy rate comparison diagram.

Detailed description of the invention

Below in conjunction with the accompanying drawings, technical scheme is specifically described.

A kind of method trusted based on AIFS weight calculation of the present invention, comprises the steps,

S1: select a cloud service provider to interact；

S2: user asks grade of service l to cloud service provider according to demand；Wherein, the clothes of cloud service provider Business grade l includes n service indication k_i ^l, and be expressed as by the form of one group of service indication vector According to the difference of grade of service l, k_i ^lService indication value the most different；

S3: use AIFS weighing computation method to determine the weight of service indication vector in step S2；

S4: cloud service provides according to requested grade of service l, provides cloud service to this request；

This service is made evaluation according to the cloud service obtained by S5: user, obtains the actual measurement of service indication vector Value；And then the weight of the service indication vector determined according to step S3, calculate user's evaluation of estimate to cloud service provider；

The cloud service evaluation that cloud service provider is provided by S6: all users determines the degrees of comparison of cloud service provider, I.e. carry out global cloud service provider's reputation to calculate, it is thus achieved that the overall credit value of cloud service provider.

In described step S3, employing AIFS weighing computation method determines the concrete of the weight that in step S2, service indication is vectorial Realize process as follows:

Definition Einstein is long-pending, Einstein and respectively with T_ε(a,b)、S_ε(a, b) represents:

$T_{\mathrm{\ϵ}}(a,b)=\frac{a\·b}{1+(1-a)\·(1-b)}---\left(1\right)$

$S_{\mathrm{\ϵ}}(a,b)=\frac{a+b}{1+a\·b}---\left(2\right)$

Wherein, a, b ∈ [0,1], (a b) is support amount and the opposition amount of fuzzy set；

Defining one group of IFS set expression is α=(μ_α, v_α), μ_αFor the support to α variable, v_αFor not propping up α variable Hold degree；The IFS decision value making A, B be decision object x, and A=(μ_A(x),v_A(x)), B=(μ_B(x),v_B(x))

Can obtain according to formula (1), (2), Einstein of AIFS is long-pending, Einstein and computing can be expressed as follows:

$A\⊗B=\{T_{\mathrm{\ϵ}}\left({\mathrm{\μ}}_A\right(x),v_A(x\left)\right),S_{\mathrm{\ϵ}}\left({\mathrm{\μ}}_A\right(x),v_A(x\left)\right)\}$

A B={S_ε(μ_A(x),v_A(x)),T_ε(μ_A(x),v_A(x))}

That is:

$A\⊗B=\{\frac{{\mathrm{\μ}}_A\left(x\right)\·{\mathrm{\μ}}_B\left(x\right)}{1+(1-{\mathrm{\μ}}_A(x\left)\right)\·(1-{\mathrm{\μ}}_B(x\left)\right)},\frac{v_A\left(x\right)+v_B\left(x\right)}{1+v_A\left(x\right)\·v_B\left(x\right)}\}---\left(3\right)$

$A\⊕B=\{\frac{{\mathrm{\μ}}_A\left(x\right)+{\mathrm{\μ}}_B\left(x\right)}{1+{\mathrm{\μ}}_A\left(x\right)\·{\mathrm{\μ}}_B\left(x\right)},\frac{v_A\left(x\right)\·v_B\left(x\right)}{1+(1-v_A(x\left)\right)\·(1-v_B(x\left)\right)}\};---\left(4\right)$

Suppose there is real number λ, then have a following multinomial multiplying:

$\mathrm{\λ}\·A=\{\frac{{\[1+{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}-{\[1-{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}}{{\[1+{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}+{\[1-{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}},\frac{2\·{\[v_A\left(x\right)\]}^{\mathrm{\λ}}}{{\[2-v_A\left(x\right)\]}^{\mathrm{\λ}}+{\[v_A\left(x\right)\]}^{\mathrm{\λ}}}\}---\left(5\right)$

Owing to there being n service indication, i.e. n decision attribute, orderFor AIFS collection value, wherein, j= 1,2,...,n；W=(w₁,w₂,...,w_n)^TForWeight vectors, and ∑ w_j=1, according to (3), (4), (5) polymerizing value of weight can be obtained, and polymerizing value is also an IFS value:

$IFW=\{\frac{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{w_j}-{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{w_j}}{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{w_j}+{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{w_j}},\frac{2\·{\Π}_{j=1}^n{v_{a_j}}^{w_j}}{{\Π}_{j=1}^n{(2-v_{a_j})}^{w_j}+{\Π}_{j=1}^n{v_{a_j}}^{w_j}}\}---\left(6\right)$

By formula (6), the support of the IFS value of each decision scheme that cloud service provider policymaker provides can be calculated Amount, the maximum scheme of support amount is optimal case, and the vectorial weight proposed using the program is as the power of service indication vector Weight.

Described step S5 to implement process as follows:

This service is made evaluation according to the cloud service obtained by user, obtains service indication vectorActual measured value (a₁,a₂,a₃,a₄,...,a_n), thus can service inclined difference sigma:

$\mathrm{\σ}=\underset{i=1}{\Σ}({k_i}^l-a_i)\·w_i---\left(7\right)$

If service indication k_i ^lActual measured value higher than service indication value, then k_i ^l-a_i=0；

User evaluates the not amount of support of cloud service provider and calculates with statistical method:

User evaluates the IFS of cloud service provider and gathers (μ_a,v_a)=(1-v_a,v_a) (9)。

Described step S6 to implement process as follows:

For overall N number of user, can be considered that the secondary AIFS polymerization that each user's weight is 1/N calculates, by each use (the μ at family_a,v_a) substitute in formula (6), available global cloud service provider's reputation value T_CSP:

$T_{CSP}=\{\frac{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{1/N}-{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{1/N}}{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{1/N}+{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{1/N}},\frac{2\·{\Π}_{j=1}^n{v_{a_j}}^{1/N}}{{\Π}_{j=1}^n{(2-v_{a_j})}^{1/N}+{\Π}_{j=1}^n{v_{a_j}}^{1/N}}\}---\left(10\right)$

The T that will obtain_CSPSupport amount percentage, available final cloud service provider overall reputation value is

$100*\frac{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{1/N}-{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{1/N}}{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{1/N}+{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{1/N}}---\left(11\right).$

Below for the present invention implement process.

1. trust model design

In this trust model, all of local side cloud user the credit rating of CSP (cloud service provider) is passed through poly- Close and be calculated overall CSP prestige.Trust model mentality of designing is as shown in Figure 1.

Trust calculation procedure:

(1) trust model selects a CSP (cloud service provider) to interact；

(2) user asks a grade of service l:CSP grade of service to include n service indication k_i ^l, with one group of service indication The form of vector is expressed asAccording to the difference of grade of service x, k_i ^lService indication value the most different； And use AIFS weighing computation method to determine the weight of service indication vector；

(3) according to the requested grade of service, CSP provides cloud service to this request；

(4) this service is made evaluation according to the cloud service obtained by user: the acquisition of evaluation of estimate is vectorial with service indication Actual measured value calculate；

(5) the cloud service evaluation that this CSP is provided by all users determines the degrees of comparison of CSP.

Herein in scheme, mainly it is made up of three parts:

(1) evidence weight w=(w is trusted₁,w₂,...,w_n)^TAcquisition；

(2) trust the acquisition of evidence, i.e. user and obtain the actual value of service indication vector；

(3) unique user trust value computing to CSP；CSP overall situation reputation calculates.

The set of all of trust evidence is as input, and these evidences export as a trust vector: T=(t₁,t₂, t₃,...,t_n).All of trust value is used 1 to quantify by we, and 1 ... n represents the quantity trusting desired value.The trust of unique user Calculated by the weight distribution of each evidence of this user and obtain.In trust model, user uses the fuzzy algorithmic approach cloud to receiving The satisfaction of service indices calculates the satisfaction of this service.

In the present invention program, how the weight of conclusion evidence vector is a most important link, rationally determines trust The weight of index, obtained user evaluates satisfaction could be reasonable, and the user of overall credible objective evaluation CSP is evaluating CSP can obtain bigger weight, the most just can reflect objective rational CSP credibility, next trifle will be given The method determining weight vectors.

3. the determination of weight

The determination of 3.1 service indication vector weights

After policymaker makes a policy, provide n dimension attribute weight proportion, and the IFS value of each decision attribute.Can draw IFW polymerizing value, can (support, the opposition degree) of this weight vectors by IFW value.Reach the purpose of Rational Decision, it is achieved Multiple decision schemes are selected the decision scheme that support is the highest.IFW decision method given below:

The core of fuzzy set be characteristic function with 0 or 1 value, this can take any value on closed interval [0,1] The extension of membership function.Zadeh fuzzy set is expanded to AIFS (Antanassov intuitionistic by Antanassov Fuzzy set), membership function is only a monotropic function, and it can not be used in many practical situations, and that expresses props up simultaneously The evidence held and oppose^[9].Document [10] proposes on the basis of AIFS further and defines that Einstein is long-pending, Einstein With, and Einsteinian scalar multiplication operation (Einstein sum, Einstein product, and Einstein scalar Multiplication.), then it is applied in Multiple Attribute Decision Problems according to these operations.

Definition 1: Einstein is long-pending herein, Einstein and respectively with T_ε(a,b)、S_ε(a, b) represents:

$T_{\mathrm{\ϵ}}(a,b)=\frac{a\·b}{1+(1-a)\·(1-b)}---\left(1\right)$

$S_{\mathrm{\ϵ}}(a,b)=\frac{a+b}{1+a\·b}---\left(2\right)$

Wherein, a, b ∈ [0,1], (a b) is support amount and the opposition amount of fuzzy set；

Definition 2: herein, one group of IFS set expression is α=(μ_α, v_α), μ_αFor the support to α variable, v_αFor α is become The not degree of support of amount；The IFS decision value making A, B be decision object x, and A=(μ_A(x),v_A(x)), B=(μ_B(x),v_B(x))

Therefore can obtain according to formula (1), (2), Einstein of AIFS is long-pending, Einstein and computing can be expressed as follows:

$A\⊗B=\{T_{\mathrm{\ϵ}}\left({\mathrm{\μ}}_A\right(x),v_A(x\left)\right),S_{\mathrm{\ϵ}}\left({\mathrm{\μ}}_A\right(x),v_A(x\left)\right)\}$

A B={S_ε(μ_A(x),v_A(x)),T_ε(μ_A(x),v_A(x))}

That is:

$A\⊗B=\{\frac{{\mathrm{\μ}}_A\left(x\right)\·{\mathrm{\μ}}_B\left(x\right)}{1+(1-{\mathrm{\μ}}_A(x\left)\right)\·(1-{\mathrm{\μ}}_B(x\left)\right)},\frac{v_A\left(x\right)+v_B\left(x\right)}{1+v_A\left(x\right)\·v_B\left(x\right)}\}---\left(3\right)$

$A\⊕B=\{\frac{{\mathrm{\μ}}_A\left(x\right)+{\mathrm{\μ}}_B\left(x\right)}{1+{\mathrm{\μ}}_A\left(x\right)\·{\mathrm{\μ}}_B\left(x\right)},\frac{v_A\left(x\right)\·v_B\left(x\right)}{1+(1-v_A(x\left)\right)\·(1-v_B(x\left)\right)}\};---\left(4\right)$

Suppose there is real number λ, Wang, Liu define multinomial multiplying:

$\mathrm{\λ}\·A=\{\frac{{\[1+{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}-{\[1-{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}}{{\[1+{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}+{\[1-{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}},\frac{2\·{\[v_A\left(x\right)\]}^{\mathrm{\λ}}}{{\[2-v_A\left(x\right)\]}^{\mathrm{\λ}}+{\[v_A\left(x\right)\]}^{\mathrm{\λ}}}\}---\left(5\right)$

Definition 3: have n service indication in the present invention, therefore be provided with n decision attribute, orderFor AIFS Collection value, wherein, j=1,2 ..., n；W=(w₁,w₂,...,w_n)^TForWeight vectors (∑ w_j= 1), the polymerizing value of weight can be obtained according to (3), (4), (5), and polymerizing value is also an IFS value:

$IFW=\{\frac{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{w_j}-{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{w_j}}{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{w_j}+{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{w_j}},\frac{2\·{\Π}_{j=1}^n{v_{a_j}}^{w_j}}{{\Π}_{j=1}^n{(2-v_{a_j})}^{w_j}+{\Π}_{j=1}^n{v_{a_j}}^{w_j}}\}---\left(6\right)$

Assume that service indication is (k₁Price, k₂Expense, k₃Service quality, k₄Network delay), its weight be (w1, w2, w3, w4)。

Policymaker provides M decision scheme (subscript represents Protocol Numbers):

Such as, first decision scheme is

Can obtain the value of the respectively IFW of each scheme according to formula (6), the result obtained also is an IFS value, then than The support amount of each IFS value compared with in M decision scheme, the scheme of support amount maximum is then optimal case, proposes with the program Vector weight is as the weight of service indication vector.

3.2 users evaluation of estimate to CSP

After user asks a grade of service, it is thus achieved that service terminates, service indication vector will be obtainedActual measured value (a₁,a₂,a₃,a₄...), user to the evaluation of estimate of CSP by servicing departure σ It is calculated.

Definition 4: service departure σ: the grade of service value of user's request and the deviation of actual measured value,

$\mathrm{\σ}=\underset{i=1}{\Σ}({k_i}^x-a_i)\·w_i---\left(7\right)$

If in certain index, the actual measured value of user is higher than service indication value (meeting the service request of user), then k_i ^x-a_i=0；

Definition 5: user evaluates the IFS of CSP and gathers (μ_a,v_a)

User evaluates the not amount of support of cloud service provider and calculates with statistical method:

User evaluates the IFS of cloud service provider and gathers (μ_a,v_a)=(1-v_a,v_a) (9)

3.3 CSP overall situation reputations calculate

During overall situation reputation calculates, if T_CSPOverall credit value for this CSP.For overall N number of user, can be considered each User's weight is that the secondary AIFS polymerization of 1/N calculates, by (the μ of each user_a,v_a) substitute in formula (6), available overall situation CSP Reputation value T_CSP:

$T_{CSP}=\{\frac{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{1/N}-{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{1/N}}{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{1/N}+{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{1/N}},\frac{2\·{\Π}_{j=1}^n{v_{a_j}}^{1/N}}{{\Π}_{j=1}^n{(2-v_{a_j})}^{1/N}+{\Π}_{j=1}^n{v_{a_j}}^{1/N}}\}---\left(10\right)$

The T that will obtain_CSPSupport amount percentage, available final cloud service provider overall reputation value is

$100*\frac{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{1/N}-{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{1/N}}{{\Π}_{j=1}^n{(1+{\mathrm{\μ}}_{a_j})}^{1/N}+{\Π}_{j=1}^n{(1-{\mathrm{\μ}}_{a_j})}^{1/N}}---\left(11\right).$

4. emulation and experiment:

(1) research of CSP prestige index change: single CSP is set in experiment and provides a user with service 100 times, number of users Amount is 10.The number of times having Fig. 1 abscissa quantity in 100 times provides a user with and does not meets the service that user requires, is finally given 100 users average to the global assessment value of this CSP.To using trust model scheme herein and not using herein in experiment Trust model scheme contrast, result as shown in Figure 2:

(2) experiment is to select credible cloud service provider accuracy rate as comparison object, compares use under identical condition The scheme of trust model TM and do not use the scheme of trust model TM, wherein the quantity of user is set to 100.Its corresponding result It is presented in Fig. 2.In most of the cases, user selects credible cloud service provider CSP accuracy to be to be above 90%.Only When CSP quantitative proportion be up to 50% and above time, this ratio drops to minimum about 85% (when malice discusses maximum), this Be improve main cause be reputation model make well-deserved reputation client trust scoring take higher weight the overall situation reputation Calculating.Therefore, the cloud service that the supplier of malice provides seldom is selected by user.

According to the research of pertinent literature [11], parameter is arranged such as table 1 below:

Table 1 parameter arranges table

In experiment two, the purpose of this experiment is to compare user to propose service request to terminating service and carrying out service After evaluation, have employed trust model and the credible CSP accuracy rate being provided without trust model compares；Wherein, there is void in 40% user The false behavior evaluated, can be modeled as the Sybil attack that CSP manufactures, and the multiple users self created improve self assessment, but adopt After TM, can effectively reduce false evaluation proportion in overall situation reputation is evaluated so that the result that overall situation reputation is evaluated is more For objective, the credible CSP accuracy rate that user can choose is higher.

5. conclusion

Trust model increasingly receives publicity in cloud computing.Trust is an important composition during relation development Part.This is the preposition major issue in cloud computing safety problem.The meaning of research trust model is: will for safety problem It is easier to take suitable safety measure, and makes correct decision, this paper presents one and dynamically adjust based on AIFS weight Trust model, by setting up rational trust model and method for evaluating trust, be given a kind of credible cloud service provider choosing Selection method, the method can dynamically adjust user's weight in global assessment from the angle of overall situation prestige, rationally pass judgment on The prestige of CSP, reaches to select as far as possible the purpose of credible CPS.Test result indicate that, this method can preferably solve credible CSP select permeability.

Being above presently preferred embodiments of the present invention, all changes made according to technical solution of the present invention, produced function is made With during without departing from the scope of technical solution of the present invention, belong to protection scope of the present invention.

Claims (4)

1. the method trusted based on AIFS weight calculation, it is characterised in that: comprise the steps,

S1: select a cloud service provider to interact；

S2: user asks grade of service l to cloud service provider according to demand；Wherein, the service etc. of cloud service provider Level x includes n service indication k_i ^l, and be expressed as by the form of one group of service indication vectorAccording to The difference of grade of service x, k_i ^lService indication value the most different；

S3: use AIFS weighing computation method to determine the weight of service indication vector in step S2；

S4: cloud service provides according to requested grade of service l, provides cloud service to this request；

This service is made evaluation according to the cloud service obtained by S5: user, obtains the actual measured value of service indication vector；Enter And the weight of the service indication vector determined according to step S3, calculate user's evaluation of estimate to cloud service provider；

The cloud service evaluation that cloud service provider is provided by S6: all users determines the degrees of comparison of cloud service provider, i.e. enters Row global cloud service provider's reputation calculates, it is thus achieved that the overall credit value of cloud service provider.

A kind of method trusted based on AIFS weight calculation the most according to claim 1, it is characterised in that: described step S3 Middle employing AIFS weighing computation method determine the weight of the vector of service indication in step S2 to implement process as follows:

Definition Einstein is long-pending, Einstein and respectively with T_ε(a,b)、S_ε(a, b) represents:

$T_{\mathrm{\ϵ}}\left(a,b\right)=\frac{a\·b}{1+\left(1-a\right)\·\left(1-b\right)}---\left(1\right)$

$S_{\mathrm{\ϵ}}(a,b)=\frac{a+b}{1+a\·b}---\left(2\right)$

Wherein, a, b ∈ [0,1], (a b) is support amount and the opposition amount of fuzzy set；

Defining one group of IFS set expression is α=(μ_α, v_α), μ_αFor the support to α variable, v_αFor α variable do not supported journey Degree；The IFS decision value making A, B be decision object x, and A=(μ_A(x),v_A(x)), B=(μ_B(x),v_B(x)) according to formula (1), (2) can obtain, Einstein of AIFS is long-pending, Einstein and computing can be expressed as follows:

$A\⊗B=\{T_{\mathrm{\ϵ}}\left({\mathrm{\μ}}_A\right(x),v_A(x\left)\right),S_{\mathrm{\ϵ}}\left({\mathrm{\μ}}_A\right(x),v_A(x\left)\right)\}$

$A\⊕B=\{S_{\mathrm{\ϵ}}\left({\mathrm{\μ}}_A\right(x),v_A(x\left)\right),T_{\mathrm{\ϵ}}\left({\mathrm{\μ}}_A\right(x),v_A(x\left)\right)\}$

That is:

$A\⊗B=\{\frac{{\mathrm{\μ}}_A\left(x\right)\·{\mathrm{\μ}}_B\left(x\right)}{1+(1-{\mathrm{\μ}}_A(x\left)\right)\·(1-{\mathrm{\μ}}_B(x\left)\right)},\frac{v_A\left(x\right)+v_B\left(x\right)}{1+v_A\left(x\right)\·v_B\left(x\right)}\}---\left(3\right)$

$A\⊕B=\{\frac{{\mathrm{\μ}}_A\left(x\right)+{\mathrm{\μ}}_B\left(x\right)}{1+{\mathrm{\μ}}_A\left(x\right)\·{\mathrm{\μ}}_B\left(x\right)},\frac{v_A\left(x\right)\·v_B\left(x\right)}{1+(1-v_A(x\left)\right)\·(1-v_B(x\left)\right)}\};---\left(4\right)$

Suppose there is real number λ, then have a following multinomial multiplying:

$\mathrm{\λ}\·A=\{\frac{{\[1+{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}-{\[1-{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}}{{\[1+{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}+{\[1-{\mathrm{\μ}}_A\left(x\right)\]}^{\mathrm{\λ}}},\frac{2\·{\[v_A\left(x\right)\]}^{\mathrm{\λ}}}{{\[2-v_A\left(x\right)\]}^{\mathrm{\λ}}+{\[v_A\left(x\right)\]}^{\mathrm{\λ}}}\}---\left(5\right)$

Owing to there being n service indication value, i.e. n decision attribute, orderFor AIFS collection value, wherein, j=1, 2,...,n；W=(w₁,w₂,...,w_n)^TForWeight vectors, andAccording to (3), (4), (5) polymerizing value of weight can be obtained, and polymerizing value is also an IFS value:

$IFW=\left\{\frac{{\mathrm{\Π}}_{j=1}^n{\left(1+{\mathrm{\μ}}_{a_j}\right)}^{w_j}-{\mathrm{\Π}}_{j=1}^n{\left(1-{\mathrm{\μ}}_{a_j}\right)}^{w_j}}{{\mathrm{\Π}}_{j=1}^n{\left(1+{\mathrm{\μ}}_{a_j}\right)}^{w_j}+{\mathrm{\Π}}_{j=1}^n{\left(1-{\mathrm{\μ}}_{a_j}\right)}^{w_j}},\frac{2\·{\mathrm{\Π}}_{j=1}^n{v_{a_j}}^{w_j}}{{\mathrm{\Π}}_{j=1}^n{\left(2-v_{a_j}\right)}^{w_j}+{\mathrm{\Π}}_{j=1}^n{v_{a_j}}^{w_j}}\right\}---\left(6\right)$

By formula (6), the support amount of the IFS value of each decision scheme that cloud service provider policymaker provides can be calculated, The maximum scheme of the amount of holding is optimal case, and the vectorial weight proposed using the program is as the weight of service indication vector.

A kind of method trusted based on AIFS weight calculation the most according to claim 2, it is characterised in that: described step S5 To implement process as follows:

This service is made evaluation according to the cloud service obtained by user, obtains service indication vector Actual measured value (a₁,a₂,a₃,a₄,...,a_n), thus can service inclined difference sigma:

$\mathrm{\σ}=\underset{i=1}{\Σ}(k_i^l-a_i)\·w_i---\left(7\right)$

If service indication k_i ^lActual measured value higher than service indication value, then k_i ^l-a_i=0；

User evaluates the not amount of support of cloud service provider and calculates with statistical method:

User evaluates the IFS of cloud service provider and gathers (μ_a,v_a)=(1-v_a,v_a) (9)。

A kind of method trusted based on AIFS weight calculation the most according to claim 3, it is characterised in that: described step S6 To implement process as follows:

For overall N number of user, can be considered that the secondary AIFS polymerization that each user's weight is 1/N calculates, by each user's (μ_a,v_a) substitute in formula (6), available global cloud service provider's reputation value T_CSP:

$T_{CSP}=\left\{\frac{{\mathrm{\Π}}_{j=1}^n{\left(1+{\mathrm{\μ}}_{a_j}\right)}^{1/N}-{\mathrm{\Π}}_{j=1}^n{\left(1-{\mathrm{\μ}}_{a_j}\right)}^{1/N}}{{\mathrm{\Π}}_{j=1}^n{\left(1+{\mathrm{\μ}}_{a_j}\right)}^{1/N}+{\mathrm{\Π}}_{j=1}^n{\left(1-{\mathrm{\μ}}_{a_j}\right)}^{1/N}},\frac{2\·{\mathrm{\Π}}_{j=1}^n{v_{a_j}}^{1/N}}{{\mathrm{\Π}}_{j=1}^n{\left(2-v_{a_j}\right)}^{1/N}+{\mathrm{\Π}}_{j=1}^n{v_{a_j}}^{1/N}}\right\}---\left(10\right)$

The T that will obtain_CSPSupport amount percentage, available final cloud service provider overall reputation value is

$100*\frac{{\mathrm{\Π}}_{j=1}^n{\left(1+{\mathrm{\μ}}_{a_j}\right)}^{1/N}-{\mathrm{\Π}}_{j=1}^n{\left(1-{\mathrm{\μ}}_{a_j}\right)}^{1/N}}{{\mathrm{\Π}}_{j=1}^n{\left(1+{\mathrm{\μ}}_{a_j}\right)}^{1/N}+{\mathrm{\Π}}_{j=1}^n{\left(1-{\mathrm{\μ}}_{a_j}\right)}^{1/N}}---\left(11\right).$