CN103701939A - Data exchange system and method thereof - Google Patents

Abstract

The invention provides a data exchange method. The data exchange method comprises the following steps: a first node in network acquires a trust vector set of a plurality of second nodes which are to be subjected to data interaction by feedback nodes, determines the trust values of the plurality of second nodes according to the trust vector set, determines interaction object nodes which are to be subjected to data interaction from the plurality of second nodes according to the trust values, and performs data interaction with the determined interaction object nodes. The invention also provides a data interaction system. According to the invention, the erroneous judgment times can be reduced effectively, thereby improving the reliability and usability of data interaction of network nodes.

Description

Data exchange system and method thereof

Technical field

The present invention relates to network information processing technology, relate in particular to a kind of data interaction system and method thereof in peer-to-peer network.

Background technology

In peer to peer environment, between the main body of not knowing each other mutually, often need to carry out data interaction and complete a certain activity, such as network trading, pp file transmission download etc.In network, data volume is very huge, and true and false data mix.How the unworthy or false data of filtering from mass data, obtain and utilize valuable True Data, for effective utilization of peer-to-peer network, are very important.

The service transacting of peer to peer environment of take is example, and the subject identity of service can be divided into service requester and ISP.When a requestor faces a plurality of alternative ISP, credible, reliable ISP of requestor's how to confirm also obtains quality services to it, is the key technical problem in peer-to-peer network.

Service evaluation data for ISP are to determine a basis for estimation of ISP's the credible and degree of reliability.But, the complexity of society, the virtual property of network and non-direct contact etc., make to have authentic assessment data, incomplete objective appraisal data and the falseness various situations such as evaluating data (hereinafter to be referred as " unreal evaluating data ") maliciously for ISP's evaluating data.If left, unreal evaluating data exists, and people just cannot distinguish or confirm ISP's real service quality so, causes the chaotic of network trading and service environment and destroys.Therefore,, in web services environment, be necessary unreal evaluating data to filter.

In the problem of filtering at unreal evaluating data at present, mostly adopted the filter method based on irrelevance, the main thought of these class methods is: the evaluating data difference degree with evaluating data person to same ISP, go to portray the size of irrelevance, think that the large node of irrelevance is exactly malicious node.This hypothesis and actual conditions are not inconsistent, and easily cause erroneous judgement.In fact, directly simply divide node into honest node and malicious node, can not summarize the behavioural characteristic of node in real system completely, can cause erroneous judgement.For example: if the service that certain node provides is an e-book, some users like this book, and some users but do not like, this species diversity is that subjectivity by user produces mostly; But if to slander this e-book be a virus document to certain node, giving goes on business comments, and this is likely malice evaluating data.Therefore subjective assessment data and malice evaluating data very likely can show consistency, make both evaluating data similarities very large, cause algorithm erroneous judgement.Equally, part of nodes is reluctant to provide valuable evaluating data to certain reciprocal process, and the evaluating data of selective system acquiescence, presents passivity.This evaluating data itself without malice, but in the process of filtering at evaluating data, with the collusion of malicious node formation advantage, also can make the algorithm based on evaluating data irrelevance lose efficacy.Therefore, the effect of existing unreal evaluating data filter method is unsatisfactory.

Summary of the invention

The object of the present invention is to provide data exchange system and method thereof, improve the validity of data interaction.

According to an aspect of the present invention, provide data interactive method, having comprised: in the first node acquisition network in network, feedback node will be to having carried out the trust vector set of a plurality of Section Points of data interaction; First node according to trust vector set determine the trust value of a plurality of Section Points; First node, according to definite trust value, is determined the interactive object node that will carry out data interaction from a plurality of Section Points; First node and definite interactive object node carry out data interaction.

In some embodiments, wherein trust vector set according to feedback node the evaluating data vector set to each Section Point

and the feedback degree of belief of each feedback node determines, wherein

one of feedback node r _ito Section Point p _jevaluating data vector; And/or the trust value of wherein determining Section Point comprises: to each the node r in trust vector set _i, select to retain evaluating data and the corresponding node that confidence level is higher; With according to the evaluating data of the respective nodes of selecting, determine the trust value of Section Point.

In some embodiments, select to retain evaluating data that confidence level is higher and the step of corresponding node can comprise: each node in the set of identification trust vector

whether comprise insecure evaluating data; With node corresponding to unreliable evaluating data in the set of deletion trust vector.

In above-mentioned steps, according to the evaluating data of the node of selecting, determine that the step of the trust value of Section Point can comprise:

According to

exceptional value threshold range is set, deletes in set higher than the evaluating data of anomaly evaluation high thresholdhigh with lower than the evaluating data of anomaly evaluation threshold value lower limit; To the set after suppressing exception thresholding, extract any two groups of evaluating datas and do and differ from and divide into groups; For the evaluating data of grouping, be chosen in each node to more one group of proportion in the trust evaluation set TrustTable of Section Point; With one group of evaluating data to selection, get weighted average, obtain feeding back degree of belief Γ _f, from TrustTable inquiry, obtain direct degree of belief Γ _d, according to direct degree of belief Γ _dwith feedback degree of belief Γ _f, determine the trust value trust of Section Point _j.

In some embodiments, can also comprise: the trust value trust that further judges Section Point _jwhether be greater than predetermined value, if judgment result is that trust value is greater than predetermined value, first node determines that this Section Point is interactive object node, and carries out data interaction with interactive object node.

According to a further aspect in the invention, a kind of data interaction system is provided, comprise: the first node in network, a plurality of feedback nodes that can communicate by letter with first node, and can with the Section Point of first node and feedback node data interaction, wherein, first node comprises degree of belief determining unit, in acquisition network, feedback node will be to carrying out the trust vector set of a plurality of Section Points of data interaction, and according to trust vector, the trust value of a plurality of Section Points is determined in set; With data interaction unit, according to definite trust value, from a plurality of Section Points, determine the interactive object node that will carry out data interaction, and carry out data interaction with definite interactive object node.

In some embodiments, degree of belief determining unit according to feedback node the evaluating data vector set to each Section Point

and the feedback degree of belief of each feedback node determines, wherein

one of feedback node r _ito Section Point p _jevaluating data vector.Wherein, degree of belief determining unit also can be to each the node r in trust vector set _i, select to retain evaluating data and the corresponding node that confidence level is higher; With according to the evaluating data of the respective nodes of selecting, determine the trust value of Section Point.

In some embodiments, degree of belief determining unit can be identified each node in trust vector set

whether comprise insecure evaluating data, and delete node corresponding to unreliable evaluating data in trust vector set.

In some embodiments, degree of belief determining unit is also carried out following operation:

According to exceptional value threshold range is set, deletes in set higher than the evaluating data of anomaly evaluation high thresholdhigh with lower than the evaluating data of anomaly evaluation threshold value lower limit; To the set after suppressing exception thresholding, extract any two groups of evaluating datas and do and differ from and divide into groups; For the evaluating data of grouping, be chosen in each node to more one group of proportion in the trust evaluation set TrustTable of Section Point; With one group of evaluating data to selection, get weighted average, obtain feeding back degree of belief Γ _f, from TrustTable inquiry, obtain direct degree of belief Γ _d, according to direct degree of belief Γ _dwith feedback degree of belief Γ _f, determine the trust value trust of Section Point _j.

In some embodiments, data interaction unit can judge the trust value trust of Section Point _jwhether be greater than predetermined value, if judgment result is that trust value is greater than predetermined value, first node determines that this Section Point is interactive object node, and carries out data interaction with interactive object node.

In some embodiments, data interaction system also comprises trust table data cell, stores the trust value of first node to Section Point, and after data interaction, upgrades trust value, and trust value is sent to first node and feedback node.

Erroneous judgement number of times when the present invention can effectively be reduced in and carry out data interaction between network node, thus the mutual reliabilty and availability of network node data promoted.

Accompanying drawing explanation

Fig. 1 is the data interaction architectural schematic of one embodiment of the present invention;

Fig. 2 is the method flow schematic diagram of the data interaction of one embodiment of the present invention.

Fig. 3 is the data interaction system schematic diagram of one embodiment of the present invention.

Embodiment

Below in conjunction with drawings and the specific embodiments, the present invention is described in further detail.

As shown in Figure 1, data interactive method provided by the invention can be applied in peer to peer environment.Hereinafter, service requester is as first node 100, and ISP is as first node 300, and for service requester (first node 100), providing is feedback node 200 to ISP's (first node 300) evaluating data feedback.The node type that is appreciated that first node 100, first node 300 and feedback node 200 is not what fix, can be according to the interior perhaps other factors alternative types of data interaction.

This peer to peer environment does not have core super node, thereby evaluating data is not to launch around ISP.At service requester, obtain in the process of service, service requester active collection Feedback Evaluation data, the Feedback Evaluation data of collecting are filtered and identified, and then ISP's degree of belief is estimated, the ISP node high to degree of belief obtains service, and the actual reciprocal process of this service is made to evaluating data.Meanwhile, dynamically update the trusted status to ISP, for forwarding TrustTable next time, provide trusted status foundation, TrustTable comprises trust vector, service requester, the trust estimation that ISP and service requester are made ISP.

The generation of Feedback Evaluation data and forwarding are carried out as follows.Feedback Evaluation is data from Feedbacks Peers, Feedbacks Peers does not participate in certain other mutual nodes in network, the service node (being the node of request service) that is provided in mutual node feeds back dependent credit message, is mainly trust vector.Feedbacks Peers processes ISP's evaluating data vector set, and wherein evaluating data vector representation is

as mentioned below, by the evaluating data vector calculation vector of establishing trust.Evaluating data vector is stored in LocalHistoryArray, comprises service requester r _i, ISP p _j, and r _ito p _jthe evaluating data ratings that the service providing provides.

Suppose node r _ito p _jset D={d at the evaluating data providing in mutual for nearest h time ₁, d ₂..., d _h, d wherein _k∈ [0,1], k ∈ [1, h], h≤10.Element in set D derives from LocalHistoryArray, arranges in chronological order d ₁from the current once mutual evaluation producing more of a specified duration, d _hexpression is from the evaluation of current recent mutual generation.

represent node r _ito p _jelement in the evaluating data set D providing in mutual for nearest h time merges the evaluating data obtaining.Merge and carry out as follows:

${\mathrm{ratings}}_i^j=\left\{\begin{array}{cc}\frac{\underset{k=1}{\overset{h}{\mathrm{\Σ}}}{d}_k\×\mathrm{\γ}\left(k\right)}{k},& h\≠0\\ 0,& h=0\end{array}\right.$

In conjunction with the cognitive process of psychology of human society, from current, recently should can reflect the authentic assessment data cases of present node to the service of receiving alternately.Therefore, to de novo interbehavior, give more weights, reduce the early weight of interaction experiences, at the superior historical factor gamma (k) decaying in time of evaluating data, γ (k) ∈ [0,1].Wherein h is r _ito same node p _jthe number of evaluating data history vectors.Historical factor gamma (k) computing formula is as follows:

$\mathrm{\γ}\left(k\right)=\left\{\begin{array}{cc}1,& k=h\\ \mathrm{\γ}(k-1)=\mathrm{\γ}\left(k\right)-\frac{1}{h},& 1\≤k\≤h\end{array}\right.$

In peer to peer environment, between node, status is equal to, and does not serve as third party's node of agency, also just there is no reliable feedback person, and broadcast feedback information is unique selection.Each node adopts the mode of fixed time broadcast, forwards the TrustTable record of oneself to credible neighbours.

For the degree of belief to ISP is estimated, calculation services supplier's degree of belief comprises 2 parts: direct degree of belief and feedback degree of belief, and by the estimation with the stack degree of establishing trust of feedback degree of belief result by direct degree of belief.The direct trust evaluation data that directly degree of belief is ISP, after feedback degree of belief is service requester and ISP's interaction data, the trust evaluation data of service requester to ISP's feedback.

The directly calculating of degree of belief:

node r _ito p _jfusion at the evaluating data collection providing in mutual for nearest h time.As described above.

The calculating of feedback degree of belief: gather at trust vector TrustTable in, N wherein _ifor ratings filters the passive evaluating data after (explanation sees below), i ∈ [1, t], S _jto be the subjective assessment data after ratings filtration, j ∈ [1, k], B _kto be the malice evaluating data after ratings filtration, k ∈ [1, m], H _pto be the honest evaluating data after ratings filtration, p ∈ [1, n].Trust vector is weighted and on average obtains feeding back degree of belief.The passive evaluating data of wherein identifying, malice evaluating data is corrupt data, and weight is 0, and subjective assessment data weighting is 0.5, and honest evaluating data weight is 1.Weighted average calculation formula is as follows:

${\mathrm{\Γ}}_f=(\underset{i=1}{\overset{t}{\mathrm{\Σ}}}{N}_i*0+\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{B}_k*0+\underset{j=1}{\overset{l}{\mathrm{\Σ}}}{S}_j*0.5+\underset{p=1}{\overset{n}{\mathrm{\Σ}}}{H}_p*1)/(j+p)$

The estimation of degree of belief: be by direct degree of belief and feedback degree of belief are superposeed, according to formula trust _j=w ₁* Γ _d+ w ₂* Γ _foperation, w ₁and w ₂be respectively weight separately, meet w ₁+ w ₂=1,

w ₁∈[0,1],w ₂∈[0,1]。(w ₁and w ₂be empirical parameter, in this example, adopt w ₁=w ₂=0.5)

Dynamically update trusted status, node completes the Evaluation and Prediction of degree of belief according to Mathematical Modeling, dynamically carries out recalculating of trust value.Node A and Node B mutual (a for example service transacting) before, will be estimated the trust value as the Node B of evaluating data object as the node A of evaluating data main body.If trust _j<0.5, is considered as not physical node, reselects evaluating data object.Otherwise, can complete mutual.After completing alternately, node A need to upgrade the trust value to B in TrustTable.

If this alternately the Feedback Evaluation data based on feedback person as judgement basis, node A will provide the trust state value of the main body of feedback to it in TrustTable inquiry while upgrading it to the trust value of Node B, if sincere node is accepted feedback person's Feedback Evaluation, otherwise refusal feedback person's Feedback Evaluation.

The trust messages of the present invention's feedback is obtained by broadcast.If node is physical node not around, refusal, from the Feedback Evaluation data of this neighbor node, to the not physical node identifying, is revised node trusted status.Specifically, node A carries out filter analysis to the trust state of own neighbor node, if own neighbor node is physical node not, its trust state is set to insincere, and after refusal from the Feedback Evaluation data of this neighbor node.

Node regularly forwards TrustTable to adjacent trusted node, only accepts the Feedback Evaluation data that trusted node provides simultaneously.If receive the feedback that neighbor node provides indiscriminately, be easy to because neighbor node is physical node not, the feedback providing that receives physical node is not also real data not, thus easily generation is refused trust service or is accepted the situation of bad service.Thereby node trusted status has very strong actual use value.

As shown in Figure 2, data filtering method according to an embodiment of the present invention comprises:

Node A searches for the service of obtaining (step S201), wherein nodes set ρ={ p from network ₁, p ₂... p _ncan provide a certain service to node A.Node A picks out node p from set ρ _j(1≤j≤n) (step S202).

It is p that node A extracts ISP from TrustTable _jtrust vector set φ.Each vector in this trust vector set φ

represent r _inode is to p _jthe evaluating data that the service that node provides provides is

(S203).

To the r in the trust vector set φ extracting in step S203 _i, identify it and provide no for passive evaluating data (step S204).If r _iin TrustTable to node p arbitrarily _jthe evaluating data providing is all the same, judges that evaluating data is passive evaluating data, this node r _ifor passive node; Otherwise, carry out next r _ithe judgement of node.So continue, to the last a r _inode judgement is complete.If each node r _iall different to the evaluating data that other node provides arbitrarily in TrustTable, think and there is no passive evaluating data in the evaluating data of this trust vector set φ, in set also without passive node.

If in step S204, the evaluating data in set φ while being judged as non-passive evaluating data, further judge all the other each evaluating datas

value; Otherwise in delete step S204 by set the evaluating data in φ

be judged as the node r of passive evaluating data _i, more further judge other evaluating data

value.In experimentation, with

the concentrated scope of value as range of normal value, by checking, this scope is set to 80%, contributes to the filtration of these exceptional values, and which piece that concentrates on of whole evaluation set can be described, is convenient to like this find out abnormity point, meanwhile, can not delete other evaluations again by mistake.

If wherein 80%

value is all more than or equal to 0.5, and evaluating data exceptional value thresholding is set to first threshold scope, and for example 0.9 and 0.2; If wherein 80%

value is all less than 0.5, and Second Threshold scope is set, and for example 0.8 and 0.1.

According to the setting of step S204, delete in set φ higher than the evaluating data of the anomaly evaluation data gate limit value upper limit with lower than the evaluating data of anomaly evaluation data gate limit value lower limit

thereby deleted abnormal evaluating data (S205).The evaluating data staying may include part subjective assessment data, honest evaluating data and malice evaluating data.Because not all subjective node is all abnormal, most of or honest, thereby thereby identify subjective node by setting threshold scope.The evaluating data that subjective node provides, can be subject to the restriction of the subjective factors such as the ken of node own, to same service, may have larger difference.Therefore, in this example, extra high evaluating data and low especially evaluating data are all rejected away, distinguished point is deleted, contribute to the follow-up processing to data, can not affect because of distinguished point the calculating of mean value, thereby reach the object of identification.

Evaluating data in the definite set of step S205 includes honest evaluating data, malice evaluating data, part subjective assessment data.Not deleting all subjective assessment data, why delete those abnormal nodes, is because above-mentioned abnormal nodes can cause very large impact for choosing of standard.

Extract any two evaluating datas do poor.Two evaluating datas (a, b), suppose that wherein a is standard, are the poor number Sab that obtains.From evaluating data set, extract an evaluating data c again, be poor relatively Sac with standard a, if Sac<Sab, a and c be a set, otherwise c and b are a set; From evaluating data set, extract an evaluating data d again, be poor Sad, if Sad<min{Sac; Sab}, if Sac<Sab, d and c are in a set; if Sac>Sab, d and b are in a set; If Sad>max{Sac, Sab}, if Sac<Sab, d and b are in a set, if Sac>Sab, d and c are in a set; If Sad is situated between, judgement more close who, corresponding and who belong to a set together.So until all extracted evaluating data.

Two groups of evaluating datas that are divided into like this, the difference of corresponding evaluating data can meet the difference of respectively organizing desk evaluation data and be tending towards 0.Because, if along with the number of evaluating increases, evaluate on the same group set and should derive from same class service requester, only there is random error in the evaluation that same ISP is provided, because of but can meet the difference of respectively organizing desk evaluation data and trend towards 0, if the satisfied difference of respectively organizing desk evaluation data trends towards 0, illustrate in two groups of evaluation set that are divided into, each group evaluation is all to derive from the same class attribute service evaluation person of ((sincerity) with a high credibility or (maliciously) with a low credibility), realize honest evaluation and malice and evaluated the classification of two class data, be divided into A and B group (step S206).

To the A of step S206 and B group evaluating data, select to retain one group of data that evaluation proportion in TrustTable of providing is large as the evaluation (S207) of believable node (service requester).One group of evaluating data that step S207 is retained processes to obtain feeding back degree of belief Γ _f, direct degree of belief Γ _dfrom TrustTable inquiry, obtain estimation egress p _jtrust value trust _j(step S208), computational methods are as the estimation step of above-mentioned degree of belief.

If the node p of estimation _jtrust value trust _j>=0.5, node A and p _jstart transaction.Otherwise, reselect node, turn back to step S202 (step S209).

Node A and node p _jcomplete after transaction, provide evaluating data, preserve evaluating data (step S210).Again to p _jfeedback node 200 carries out trusted status classification, and overall process finishes.

In the analogue system of example, be provided with 42 nodes, in honest evaluating data: subjective assessment data: malice evaluating data: the ratio that passive evaluating data is 7:7:4:3 generates, and each node independently completes interactive history at random.On the evaluating data basis generating, a pair of node (the A of random choose, B), wherein the trust value of node A estimation Node B, moves respectively the inventive method, the filter method based on irrelevance and does not have filter algorithm to calculate trust value, the distance that irrelevance refers between evaluating data is algorithm of the present invention, do not have filter algorithm to refer to: a pair of node of random choose, without method of the present invention, filter evaluating data, directly initiate request, calculate False Rate, take False Rate as performance index.False Rate is that node A allows node A to the Node B that poor service is provided, initiate the shared proportion of number of times of request service by the definite trust value of said method of the present invention mistakenly.Test has been carried out 10 times altogether, experimental condition: require above-mentioned node to generate by corresponding proportion, ratio refers to node ratio, the generation of evaluating data is random, each node independently completes interactive history at random, the generation that is to say evaluating data is random, can guarantee that like this each node equiprobability of experiment participates in mutual.Strengthen credibility, generate in proportion node, independently complete at random interactive history, filter evaluating data.Do accordingly 1000 and take turns, count False Rate, said process is repeated 10 times, average, obtain final False Rate 5.1%, 20.4%, 50.9% and illustrated that improvement algorithm can effectively reduce erroneous judgement number of times, thereby can promote the reliabilty and availability of system.

According to a further aspect in the invention, a kind of data interaction system is provided, comprise as shown in Figure 3: the first node 100 in network, a plurality of feedback nodes 200 that can communicate by letter with first node 100, and can with the Section Point 300 of first node 100 and feedback node 200 data interactions.Wherein, first node 100 comprises degree of belief determining unit 101, obtains the 200 pairs of trust vector set that will carry out a plurality of Section Points 300 of data interaction of feedback node in network, and according to trust vector, the trust value of a plurality of Section Points 300 is determined in set; With data interaction unit 102, according to definite trust value, from a plurality of Section Points 300, determine the interactive object node that will carry out data interaction, and carry out data interaction with definite interactive object node.

Degree of belief determining unit 101 is according to the evaluating data vector set of 200 pairs of each Section Points 300 of feedback node

and the feedback degree of belief of each feedback node 200 determines, wherein

one of feedback node 200 r _ito Section Point 300p _jevaluating data vector.Wherein, degree of belief determining unit 101 also can be to each the node r in trust vector set _i, select to retain evaluating data and the corresponding node that confidence level is higher; With according to the evaluating data of the respective nodes of selecting, determine the trust value of Section Point 300.Determine the method for Section Point 300 trust values as described above.

In addition, degree of belief determining unit 101 can be identified each node in trust vector set

whether comprise insecure evaluating data, and delete node corresponding to unreliable evaluating data in trust vector set.RM as described above.

Degree of belief determining unit 101 each node in carry out the set of identification trust vector whether comprise the insecure evaluating data police, can carry out following operation:

According to

exceptional value threshold range is set, deletes in set higher than the evaluating data of anomaly evaluation high thresholdhigh with lower than the evaluating data of anomaly evaluation threshold value lower limit; To the set after suppressing exception thresholding, extract any two groups of evaluating datas and do and differ from and divide into groups; For the evaluating data of grouping, be chosen in each node to more one group of proportion in the trust evaluation set TrustTable of Section Point 300; With one group of evaluating data to selection, get weighted average, obtain feeding back degree of belief Γ _f, from TrustTable inquiry, obtain direct degree of belief Γ _d, according to direct degree of belief Γ _dwith feedback degree of belief Γ _f, determine the trust value trust of Section Point 300 _j.

Data interaction unit 102 can judge the trust value trust of Section Point 300 _jwhether be greater than predetermined value, if judgment result is that trust value is greater than predetermined value, first node 100 determines that this Section Point 300 is interactive object node, and carries out data interaction with interactive object node.

Data interaction system also comprises trust table data cell 103, store the trust value of 100 pairs of Section Points 300 of first node, and after data interaction, upgrade the trust value to the Section Point 300 of this data interaction, and trust value is sent to first node 100 and feedback node 200.

Erroneous judgement number of times when the present invention can effectively be reduced in and carry out data interaction between network node, thus the mutual reliabilty and availability of network node data promoted.

Claims (10)

1. data interactive method, comprising:

In first node in network (100) acquisition network, feedback node (200) will be to carrying out the trust vector set of a plurality of Section Points (300) of data interaction;

Described first node (100) is determined the trust value of described a plurality of Section Point (300) according to described trust vector set;

Described first node (100), according to described trust value, is determined the interactive object node that will carry out data interaction from described a plurality of Section Points (300);

Described first node (100) carries out data interaction with described definite interactive object node.

2. data interactive method according to claim 1, wherein said trust vector set is the evaluating data vector set to Section Point described in each (300) according to described feedback node (200)

and feedback degree of belief of feedback node (200) is determined described in each, wherein

one of described feedback node (200) r _ito described Section Point (300) p _jevaluating data vector; And/or

The trust value of wherein said definite described Section Point (300) comprising:

To node r described in each in described trust vector set _i, select to retain evaluating data and the corresponding node that confidence level is higher; With

According to the evaluating data of the respective nodes of described selection, determine the trust value of described Section Point (300).

3. data interactive method according to claim 2, wherein said selection retains evaluating data that confidence level is higher and the step of corresponding node comprises:

Identify each node in described trust vector set whether comprise insecure evaluating data; With

Delete node corresponding to unreliable evaluating data in described trust vector set, and/or

Described according to the evaluating data of the node of described selection, determine that the step of the trust value of described Section Point (300) comprising:

According to

exceptional value threshold range is set, deletes in set higher than the evaluating data of anomaly evaluation high thresholdhigh with lower than the evaluating data of anomaly evaluation threshold value lower limit;

To the described set after described suppressing exception thresholding, extract any two groups of evaluating datas and do and differ from and divide into groups;

For the evaluating data of described grouping, be chosen in each node to more one group of proportion in the trust evaluation set TrustTable of described Section Point (300); With

To one of described selection group of evaluating data, get weighted average, obtain described feedback degree of belief Γ _f, from described TrustTable inquiry, obtain direct degree of belief Γ _d, according to described direct degree of belief Γ _dwith feedback degree of belief Γ _f, determine the trust value trust of described Section Point (300) _j, and/or

Further judge the trust value trust of described Section Point (300) _jwhether be greater than predetermined value,

If judgment result is that described trust value is greater than predetermined value, described first node (100) determines that this Section Point (300) is described interactive object node, and carries out data interaction with described interactive object node.

4. data interactive method according to claim 2, wherein said evaluating data vector according to node r _ito p _jelement in the evaluating data set D providing in mutual for nearest h time merges and obtains, and described fusion is carried out as follows:

${\mathrm{ratings}}_i^j=\left\{\begin{array}{cc}\frac{\underset{k=1}{\overset{h}{\mathrm{\&Sigma;}}}{d}_k\&times;\mathrm{\&gamma;}\left(k\right)}{k},& h\&NotEqual;0\\ 0,& h=0\end{array}\right.$

Wherein, h is r _ito same node p _jthe number of evaluating data history value, γ (k) is the historical factor, determines as follows:

$\mathrm{\&gamma;}\left(k\right)=\left\{\begin{array}{cc}1,& k=h\\ \mathrm{\&gamma;}(k-1)=\mathrm{\&gamma;}\left(k\right)-\frac{1}{h},& 1\&le;k\&le;h\end{array}\right.$ 。

5. data interactive method according to claim 4, wherein said feedback degree of belief obtains in the following manner:

In the trust vector set TrustTable of described feedback node (200), establish

in, N wherein _ifor passive evaluating data, i ∈ [1, t], S _jto be subjective assessment data, j ∈ [1, k], B _kmalice evaluating data, k ∈ [1, m], H _phonest evaluating data, p ∈ [1, n], described feedback degree of belief Γ _fby the weighted average of described passive evaluating data, malice evaluating data, subjective assessment data and honest evaluating data is obtained.

6. data interactive method according to claim 5, wherein said Γ _fweighted average mode be:

${\mathrm{\&Gamma;}}_f=(\underset{i=1}{\overset{t}{\mathrm{\&Sigma;}}}{N}_i*0+\underset{k=1}{\overset{m}{\mathrm{\&Sigma;}}}{B}_k*0+\underset{j=1}{\overset{l}{\mathrm{\&Sigma;}}}{S}_j*0.5+\underset{p=1}{\overset{n}{\mathrm{\&Sigma;}}}{H}_p*1)/(j+p)$ 。

7. data interactive method according to claim 6, wherein said Section Point (300) degree of belief trust _jobtain in the following manner:

trust _j=w ₁×Γ _d+w ₂×Γ _f

Wherein, w ₁and w ₂be respectively weight coefficient, meet w ₁+ w ₂=1, w ₁∈ [0,1], w ₂∈ [0,1].

8. according to the data interactive method described in claim 1-7 any one, also comprise: described first node (100) and described interactive object node complete after data interaction, generate the evaluating data to described interactive object node.

9. a data interaction system, comprise: the first node in network (100), a plurality of feedback nodes (200) that can communicate by letter with first node (100), and can with the Section Point (300) of described first node (100) and feedback node (200) data interaction

Wherein, described first node (100) comprising:

Degree of belief determining unit (101), in acquisition network, feedback node (200) will be to carrying out the trust vector set of a plurality of Section Points (300) of data interaction, and according to trust vector, the trust value of a plurality of Section Points (300) is determined in set; With

Data interaction unit (102) according to described definite trust value, is determined the interactive object node that will carry out data interaction, and is carried out data interaction with definite interactive object node from described a plurality of Section Points (300).

10. data interaction system according to claim 9, wherein said data interaction system also comprises trust table data cell (103), store the trust value of described first node (100) to described Section Point (300), and after described data interaction, upgrade described trust value, and described trust value is sent to described first node (100) and feedback node (200).

Images