CN103152436A - P2P (peer-to-peer) internet trust cloud model computing method based on interest group - Google Patents
P2P (peer-to-peer) internet trust cloud model computing method based on interest group Download PDFInfo
- Publication number
- CN103152436A CN103152436A CN 201310110038 CN201310110038A CN103152436A CN 103152436 A CN103152436 A CN 103152436A CN 201310110038 CN201310110038 CN 201310110038 CN 201310110038 A CN201310110038 A CN 201310110038A CN 103152436 A CN103152436 A CN 103152436A
- Authority
- CN
- China
- Prior art keywords
- node
- group
- trust
- cloud
- interest
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
Images
Landscapes
- Computer And Data Communications (AREA)
Abstract
The invention discloses a P2P (peer-to-peer) internet trust cloud model computing method based on the interest group. The P2P internet trust cloud model computing method consists of an interest group diving and message transferring process, a trust cloud construction process and a node credibility evaluation process, wherein the internet is divided into different groups according to the interests of each group in the interest group diving and message transferring process, and a message transferring mode between nodes in groups and between groups is respectively determined; in the trust cloud construction process, three digital characteristics of the cloud model, i.e. expectation, entropy and excess entropy, are utilized to express a node trust relationship, and an integrated trust cloud consisting of a direct trust cloud and a recommendation trust cloud is constructed; and in the node credibility evaluation process, the node integrated trust value is evaluated by giving priority to a direct experience trust value computing method. According to the method disclosed by the invention, the problem that the direct trust relationship is difficult to establish due to excessive message transferring and asymmetric interest in the P2P internet is solved, and the risk for the node to obtain the unreliable recommendation information can be lowered.
Description
Technical field
The invention belongs to the trust model studying technological domain, relate more specifically to a kind of in the P2P network based on the trust cloud model computational methods of interest group.
Background technology
From being born from the Internet, P2P has just existed, and it is origin and the basis of the Internet.As far back as 1979, the typical case that Truscott and Ellis have just developed based on P2P uses: network, it played Information Communication as a kind of exchange way on the Internet.In recent years, the P2P network relies on the advantages such as its unique distributivity, self-organization to shoot up on the internet, becomes the important component part of the Internet, and the P2P network does not rely on Centroid, can effectively utilize the various slack resources in network, aspect a lot of, good development be arranged.Yet the equity due to the P2P network, the node network of can freely coming in and going out, cause and have more potential safety hazard in network, such as internet worm is attacked, inconsistent attack of behavior etc. is for these problems, a lot of scholars are studied, research finds, the height by degree of belief in trust model comes the identification malicious node, helps to reduce the propagation of malicious file in network; By add incentive mechanism in trust model, can make node better participate in network, provide the resource of use, the access that solves in network is attacked; By trust model, can well wish by inconsistent its essence of doing evil of covering of behavior in identification network, the assailant of deception close together node can be so that cooperate between node more.
Research for the trust model in the P2P network at present has a lot, and each model has shortcoming and advantage separately, and wherein that comparatively classical is XRep, EigenTrust, PeerTrust.
the XRep model adopts the method for mean value calculation to obtain the trust value of node, model does not filter trust information, think that all trust information are all reliable, Weight in calculating for degree of belief is not specifically considered (Damiani E, Vimercati DC, Paraboschi S, Samarati P, Violante F. A reputation-based approach for choosing reliable resources in peer-to-peer networks. Proceedings of 9th ACMP Conf on Computer and Communications Security (CCS'02). Washington DC, USA, ACMP Press, 2002:207 ~ 216.).
the EigenTrust model is by the unique degree of belief of the iterative computation node overall situation, each transaction all needs the iteration of whole network to cause the cost of communication larger, the convergence of calculating is a problem, and model is used as reliable recommended node to the high node of degree of belief, there is irrational place, be exactly not necessarily a honest people like a conscientious people of work, what is said or talked about differs that to establish a capital be reliable (Sepandar DK, Mario TS, Hector GM.The EigenTrust algorithm for reputation management in P2P networks.Proceedings of the 12th Int'l Conf.on World Wide Web. Budapest:ACM Press, 2003:640 ~ 651.).
The PeerTrust model considers affect the factor of node confidence from many aspects: conclude the business evaluation, number of transaction, exchange hour, transaction limit, provide the confidence level of Feedback Evaluation node etc.This algorithm attack tolerant is stronger, but do not consider to malicious node punishment, calculate convergence rate and utilize sparse data to calculate similarity and may bring problem (the Xiong L such as larger error, Liu L. PeerTrust:Supporting Reputation-Based Trust for Peer-to-Peer Electronic Communities.IEEE Transaction on knowledge data engineering, 2004,16 (7): 843 ~ 857.).
From the computational methods of degree of belief, at present have several different methods in trust model, such as based on cloud model, Based on Probability, based on fuzzy theory etc.
Beth proposes trust model and is based on experience and probability statistics, model is illustrated the description of the trusting relationship theory by experience, the computational methods of degree of belief have been provided, simultaneously, proposed to use the method for expressing of probable value that the possibility that the inter-entity execution is concluded the business is described, model does not solve the problem of obtaining of initial trust, do not consider fuzzy behaviour (the Beth T of trust itself, Boreherding M, Klein B. Valuation of Trust in open network. Proceedings of the European Symposium on Research in Computer Security (ESORICS). New York:Springer-Verlag, 1994:3 ~ l8.).
The trust model of the propositions such as Tang Wen is to set up on the basis of fuzzy theory, this model applies to fuzzy theory in trust model, by degree of membership, ambiguity is described, provided simultaneously the method that conceptional tree this thought illustrates trust type, but model has been ignored the randomness (Tang Wen that trusts, Chen Zhong. based on the Subjective Trust Management Model research of Fuzzy Set Theory. Journal of Software, 2003,14 (8): 1401 ~ 1408.).
Zhang Weiguang etc. have proposed the Trust Valuation Model based on cloud model, this model applies to this thought of cloud model in subjective trust research, preferably resolve ambiguity and the randomness of trust, provided the calculating of direct trust by reverse cloud algorithm, provided simultaneously the algorithm of recommendation trust, make the subjective trust reasoning be easy to understand, but model is not considered the various malicious act (Zhang Guangwei in network, Kang Jianchu, Meng Xiangyi etc. the subjective trust based on cloud model represents research. computer science, 2006,33(11): 158 ~ 161.).
Trust model research has obtained progress so far aspect a lot, such as the algorithm of model, and the storage of the data of model, fail safe of model etc., these progress are all the results that every scholar makes great efforts, but also have some problems:
(1) Computational Complexity.Through years of researches, there has been the multidigit scholar to provide trust model separately, each model computational methods used are also different, but, the computation complexity of part trust model is larger, makes to trust the computational efficiency step-down, has reduced the practicality of model.
(2) recommend reliability disadvantages.There is a large amount of dishonest recommended nodes in network, such as exaggerate, slander, cooperation deception node, make the recommendation information that obtains unreliable, thereby affect the result of trust evaluation, the trust model of most is all to estimate by computing node the confidence level that the behavior similarity judges recommended node, this is with regard to there being such problem, if evaluation information is less, just there is larger error in the confidence level that obtains by similitude.
(3) " punishment is rewarded " mechanism of trust model.Have than multi-model in the many trust models that exist at present and do not provide corresponding " award " mechanism, provide service in network thereby cause most node to be unwilling to participate in, because this does not have too large meaning for their part, can not get a lot of benefits, this just cause service is provided in network node seldom, simultaneously, for malicious node, also need one effective " punishment " mechanism that it is processed.
(4) foundation of trusting relationship and network traffic problem thereof.Along with the continuous expansion of network size, the chance that transaction occurs between node diminishes gradually, set up direct trusting relationship and will become difficult between node, also can cause message transmission a large amount of in network simultaneously.
Summary of the invention
The object of the invention is to for the trust model in the P2P network provides reliable computational methods, solve the calculation of complex, the accuracy that exist in prior art not high, lack the problem such as rewards and punishments mechanism.
Realize that the object of the invention technology solves bill and is: a kind of P2P network trust cloud model based on interest group calculates and comprises interest group division, interest group message process, trust cloud building process and four steps of node confidence evaluation procedure.
Step 1 interest group is divided: at first determine the number of interest group, each interest group subject of great interest hobby number and network File number of resources; Then determine the group head node, according to the group number of setting, hobby number, file resource number in network, determine the affiliated group of node in network, last non-group of head node searches by query requests the group head node that theme interest is identical with it, and the group head node connects each other;
Step 2 interest group message process: determine the group message transfer mode according to the content of querying node, if theme interest is organized interior inquiry, if less important interest, inquiry between organizing;
Step 3 is trusted the cloud building process: at first obtain node
iWith node
jThe direct interaction number of times
hSecondly compare interaction times
hWith given interaction times threshold value
HIf size is interaction times
hMore than or equal to given interaction times threshold value
HThe time, with node
iTo node
jThe reverse Normal Cloud generating algorithm of direct trust evaluation value input one dimension, obtain directly trusting cloud
If node
iWith node
jThe direct interaction number equal at 0 o'clock, node
iCan only measure node by the recommendation of other nodes
jTrust value, the calculated recommendation node
The recommendation confidence level
, only the recommendation evaluation of estimate of reliable recommended node is inputted the reverse Normal Cloud generating algorithm of one dimension, the recommendation confidence level of reliable recommended node
Thereby, obtain the recommendation trust cloud
If interaction times
hGreater than 0, less than given interaction times threshold value
H, combined joint
iTo node
jDirect trust cloud and recommendation trust cloud jointly determine node
iTo node
jComprehensive trust cloud
Step 4 node confidence evaluation procedure: at first with node
iTo node
jComprehensive trust cloud
Expectation
As node
jTrust value
Then comparison node
jTrust value
With the good turn Evaluation threshold
GTSize, if
Less than the good turn Evaluation threshold
GT, node is not carried out the encouragement and penalty of degree of belief, if
More than or equal to the good turn Evaluation threshold
GT, comparison node
jThe behavior uncertain factor
With Evaluation threshold
GTSize is used for distinguishing malicious node and good will node; If node
jThe behavior uncertain factor
Less than behavior uncertain factor threshold value
, to node
jReward, improve its degree of belief, if node
jThe behavior uncertain factor
Greater than behavior uncertain factor threshold value
, to node
jPunish, reduce its degree of belief.
The present invention compared with prior art, its significant advantage is: (1) has been used for reference in the cognitive custom of human psychology and has preferentially been adopted the thought that direct experience is carried out the decision node degree of belief, makes that transaction count surpasses the interaction times trust threshold between two nodes
HThe time, node just can be trusted judgement by self, does not need recommendation information, trusts thereby simplified the complexity of calculating; (2) propose a kind of method of new calculating " recommendation confidence level ", filtered out incredible recommendation information, made model can resist better the malicious act of dishonest recommended node; (3) use and to characterize probabilistic two parameter-entropys and super entropy in traditional cloud model, introduce reward the factor and penalty factor respectively to the good will node implement to reward, to the malicious node administer doses of punishment, make distinguishing to become and be more prone at the good and evil node; (4) introduce " theme interest ", " less important interest " concept makes the interest grouping more reasonable, can better solve in trust model to be difficult to form direct trusting relationship because of the interest difference, simultaneously, effectively reduces message transmission number in network.
Below in conjunction with accompanying drawing, the present invention is described in further detail.
Description of drawings
Fig. 1 trust value computational process
Fig. 2 grouping process figure
Fig. 3 rear P2P network fundamental diagram that divides into groups
Fig. 4 comprehensively trusts the cloud flow chart
Fig. 5 degree of belief is rewarded the punishment flow chart
Fig. 6 average message complexity (AMC) is with the Changing Pattern of network size
The average interaction times of Fig. 7 (ANT) is with the Changing Pattern of network size
Fig. 8 unsuccessfully downloads the Changing Pattern of sum (TFN) with the emulation cycle
Successful download rate under Fig. 9 cooperation deception malicious node CDMP exists
Successful download rate under Figure 10 strategy type malicious node SMP exists
Embodiment
In conjunction with Fig. 1, the present invention is based on the P2P network trust cloud model computational methods of interest group, step is as follows:
The first step, interest group are divided and and message process.At first set forth the interest group partition process in conjunction with Fig. 2:
Step 1: note
Be the interest set of P2P network, each node
iThere is its interested content type to be designated as
, wherein
,
The node sum that has in the expression network, but each node has a most interested content type, and namely each node has a theme interest, is intended for the foundation of dividing interest group.Except theme interest, each node can also have a plurality of interested less important themes, and lower than theme interest, each less important interest also can be according to according to prioritization to the interested degree of less important interest for node; Therefore, each node is preserved a hobby list.
Step 2: when initial, according to aspect performances such as the processing of node i, storage, bandwidth, determine whether it is the group head node, if in these areas, the better performances of node is defined as organizing head node, otherwise is non-group of head node; In order to prevent single point failure, also a plurality of groups of head nodes can be arranged in each group, in case the group head node deviated from network that connects, can the selection group in other group head node.
Step 3: if this node is non-group of head node, carry out table of query and routing QRT ground exchange by query routing protocol QRP with the group head node, non-group of head node sends corresponding QRT to the group head node, the group head node is integrated the QRT that receives and the QRT of himself, simultaneously with network in other group head nodes carry out QRT and exchange.
Step 4: if can not find the identical group head node of theme interest, can according to the sequence of less important interest, be connected into less important interest group;
Step 5: if this node is the group head node, its can be with it nearest group head node connects.
Then, provide the message process in network after grouping in conjunction with Fig. 3, the message transmission can be divided in group inquiry and assembly and inquire about.
(1) inquiry in the group
If inquiry in group, the content of the querying node interest that is the theme.At first, send a query requests to the group head node, the group head node sends in group other non-group of head node according to concrete query contents; After non-group of head node received query requests, according to the situation of itself, if there is corresponding content, this request is responded; After query node is received response, connect with it, realize resource-sharing.
(2) inquiry between the group
If the content that node will be inquired about is less important interest content, organize head node and send to the group head node of other groups that are attached thereto in the mode of broadcasting, in this way, query requests is handed on always, until find the group head node of the less important interest group that needs inquiry, and then be transmitted to non-group of head node of this group by the group head node of less important interest group.Concrete step is as follows: at first the group head node to this group sends a query requests, and the group head node of this group finds it is not the content of this group according to concrete query contents, relies on QRT to send to the group head node of less important interest; Secondly, less important interest group head node sends to according to the content of inquiry the non-group of head node that is attached thereto in group; Then, after non-group of head node of less important interest received query requests, according to the situation of itself, if there is corresponding content, this request is responded; At last, after query node is received response, connect with it, realize resource-sharing.
Second step is comprehensively trusted the cloud building process, sets forth cloud model and human psychology cognitive custom in conjunction with Fig. 4 and combines and provided the process that computing node is comprehensively trusted cloud, and its method is as follows:
Step 1: structure is directly trusted cloud.Work as node
iWant decision node
jDegree of belief the time, obtain node
iWith node
jBetween
hInferior direct interaction experiences, direct interaction experiences is expressed as node
iTo node
jThe satisfaction score value that service is provided, score value is between interval [0,1.0], wherein score value 0 expression node provides the poorest service, and is very dissatisfied; Score value 0.25 expression node provides transaction unsuccessful, and is dissatisfied; Score value 0.5 expression node provides service common, and is general satisfied; Score value 0.75 expression node provides Transaction Success, and is satisfied; Score value 1.0 expression nodes provide transaction very successful, feel quite pleased.Suppose that it is 0.75 and 1.0 service that the good will node only provides satisfaction, it is 0.0 and 0.25 service that satisfaction is provided when malicious node is done evil, and provides 0.75 and 1.0 service when not doing evil, the 0.5 initial trust degree as node.If
, with node
iTo node
jThe reverse Normal Cloud generating algorithm of direct trust evaluation value input one dimension obtain directly trusting cloud
, the reverse Normal Cloud generative process of concrete one dimension is as follows:
Step 2: structure recommendation trust cloud.If node
iWith node
jBetween without direct interaction experiences, namely the direct interaction number of times equals at 0 o'clock, works as node
iWhen wanting to obtain the recommendation trust of node j, node
iNeed to be according to node
jThe recommended node set obtain at no distant date all once with node
jThe node of interactive history experience was arranged to node
jThe evaluation of estimate of service is provided.But in order to guarantee the reliability of recommended node, need to be each recommended node
All preserve historical recommendation statistical information, comprised the successful referral number of times
, unsuccessfully recommend number of times
With the recommendation confidence level
, query node
iAt every turn when adopting recommendation information, if the confidence level of this recommended node
, think node
Recommendation information reliable, can adopt, otherwise, do not adopt, namely only will recommend confidence level
The reverse Normal Cloud generating algorithm of recommendation evaluation of estimate input one dimension of node, obtain the recommendation trust cloud
In recommendation trust cloud construction process, the recommendation confidence level of recommended node calculates as follows:
Step 2.1: node
iWith node
jAfter completing once transaction, according to the service quality that gets, with reference to the satisfaction standards of grading, provide evaluation of estimate, in conjunction with original evaluating data, the method by mean value calculation obtains node
jDirect degree of belief
Step 2.3: calculate
With
The absolute value of difference, if
, wherein
For recommending the confidence level threshold value, be generally service quality maximum difference good in trust evaluation, successful recommendation number of times
Add 1, otherwise, failed recommendation number of times
Add 1;
Step 3: the comprehensive trust of structure cloud.Work as node
iWith node
jThe direct interaction number of times
hMore than or equal to given interaction times threshold value
HThe time, think node
iAccording to own existing historical experience, fully can decision node
j, directly trust cloud
Be exactly comprehensively to trust cloud
Work as node
iWith node
jThe direct interaction number of times
hEqual at 0 o'clock, node
iCan only obtain node by recommended node
jTrust evaluation, this moment is comprehensive trusts cloud
Deteriorate to the recommendation trust cloud
Work as node
iWith node
jThe direct interaction number of times
hArrive 0
HIn the time of in scope, comprehensively trust cloud and jointly determined by direct trust cloud and recommendation trust cloud, need this moment first to calculate and directly trust cloud
, then calculated recommendation is trusted cloud
, trust cloud with these two kinds at last and carry out comprehensively obtaining comprehensively trusting cloud
The formula of above computational methods is as follows:
Wherein
Be the logic multiply operator,
Be the logic add operator,
Embodied the lay particular stress on degree of node self to direct trust and recommendation trust, its size can be regulated, corresponding
,
,
Computing formula is:
In the 3rd step, the node trust value is estimated.After the comprehensive trust cloud that gets node, just can reward punishment to node according to trust value and the behavior uncertain factor of node, in conjunction with Fig. 5, the node trust value computational methods that punitive measures is rewarded in concrete fusion are as follows:
Node
iAccording to the node that gets
jComprehensive trust cloud
, will expect
As node
jTrust value
If,
Less than the good turn Evaluation threshold
GT, node is not carried out the encouragement and penalty of degree of belief; If
More than or equal to the good turn Evaluation threshold
GT, computing node at first
jThe behavior uncertain factor
If, greater than behavior uncertain factor threshold value
, node is punished, reduce its degree of belief, the punishment formula is:
, wherein, wherein
Penalty factor,
The satisfaction that the service that can provide service node according to model is provided of value is regulated, thereby determines the punishment dynamics size to node,
Be worth littlely, the degree of belief that calculates is less, and is just larger to the punishment dynamics of node; If node
jThe behavior uncertain factor less than behavior uncertain factor threshold value
, node is rewarded, improve its degree of belief, the award formula is:
, wherein
For rewarding the factor, be used for regulating model to the award degree of node confidence, when the nodes ' behavior uncertain factor does not surpass behavior uncertain factor threshold value, can give to a certain degree award to node, to promote the degree of belief of node.
At present, we have verified superperformance of the present invention by simulated experiment and contrast experiment, wherein the contrast experiment to as if the PeerTrust model.Having defined 3 kinds of nodes in experiment, is respectively good will node and dishonest recommended node and malicious node.Good will node Good Peer wherein, note by abridging and be GP: only provide the node of good will service, namely the service quality of node is 0.75 and 1.0; Malicious node Cooperation Deceive Malicious Peer is cheated in cooperation, notes by abridging to be CDMP: during as query node, for the node in its group, the trust evaluation that provides is all 1.0, and for the node outside group, the trust evaluation that provides is all 0.0; During as service node, the behavior that the node in its group is done is all good will, to the node outside its group, carries out malicious act with the probability of mrate, and it is 0.0 and 0.25 service that service namely is provided; Strategy type malicious node Strategic Malicious Peer, brief note is for SMP: make the degree of belief of self always can be lower than a certain value T by certain strategy, when the degree of belief of finding self during lower than T, begin to carry out service quality and be the behavior of 1.0 good will, when higher than or when equaling T, begin to do evil, it is 0.0 and 0.25 service that service quality namely is provided.Further, in order to assess the performance with contrast model, 3 evaluation indexes have been defined, successfully download rate, average message complexity, convergence rate.Wherein, successfully download rate Successful Download Rate notes by abridging and is SDR, refers in process of exchange, and the number of times that the good will node is successfully downloaded accounts for the ratio of the total download time of all good will nodes, is formulated as
Average message complexity Average Message Complexity notes by abridging and is AMC, is illustrated in system's running, on average passes through the message number of individual node.Convergence rate can embody by two aspects: on average interaction times and unsuccessfully download sum.Wherein, average interaction times Average Transaction Number notes by abridging and is ATN, and the expression model is reaching a stability through what all after dates, and namely failed download time is tending towards 0; Total Total Failure Download Number is downloaded in failure, notes by abridging to be TFN, represents the total failed download number of all good will nodes in the single cycle.
Fig. 6 is for 500,1000,2000,3000 heterogeneous networks scales, network is divided into 10 groups, in the situation that ttl value is 6, average message complexity AMC after grouping in the front CMTrust network of IGCMTrust and grouping relatively, after the grouping, the average message complexity in network obviously tails off as seen from Figure 6, this is due to the grouping posterior nodal point if inquire about theme interest, and it is inquiry in group, and query context reduces greatly.
Fig. 7 is under the diverse network scale, and malicious node is the SMP of T=0.8, the ATN test result in accounting 50% situation, and test with PeerTrust.As can be seen from Figure 8 the ATN of IGCMTrust compares with PeerTrust, obvious advantage is arranged, variation along with network size, the ATN of IGCMTrust is smaller, and PeerTrust is 3000 o'clock in network size, its ATN has reached 100, and this has illustrated that IGCMTrust is better than PeerTrust in the performance aspect average interaction times ATN.
Fig. 8 is that network size is 1000, malicious node is the SMP of T=0.8, the total TFN test result of failed download in 500 situations is arranged, see on the whole, the TFN of IGCMTrust is all the time lower than PeerTrust, simultaneously, reduce much since the TFN of the 7th cycle IGCMTrust quickly, and stably keep down, make node download to the malicious file number of times and become less, this explanation IGCMTrust trust model can suppress the malicious act of SMP quickly, and the validity of model is embodied.
Fig. 9 and Figure 10 are that network size is 1000, and the good will node is 500, and malicious node is 500, the uncertain threshold value of behavior
Be 0.13, penalty factor
Be 0.5, reward the factor
Be 0.5, the interaction times trust threshold
HBe 10, the degree that lays particular stress on of direct trust and recommendation trust
Be 0.5, recommend the confidence level threshold value
Be 0.25, good will Node evaluation threshold value
GTBe in 0.75 situation when having cooperation deception malicious node CDMP and tactful type malicious node SMP in network, the test result of the successful download rate SDR in network.
As seen from Figure 9, when having a large amount of CDMP in network, IGCMTrust can resist this CDMP's " cooperation deception " behavior preferably, its SDR is higher than the PeerTrust trust model, this is because IGCMTrust is different from the method that PeerTrust weighs recommended node credibility, and PeerTrust adopts similitude is obtained the multiple evaluation value in the not too suitable the present invention of recommendation confidence level situation; In figure, the No-Trust model is selected responsive node at random, so its SDR is in 0.5 left and right; Can find out in addition, when CDMP take mrate as 0.3 when doing evil, also namely when doing evil than small probability, the SDR of IGCMTrust still can maintain a higher level, this is because the IGCMTrust model has been introduced the recommendation confidence level
RTD, filtered out the recommendation information of incredible malicious node, and divided into groups to make the increased frequency of concluding the business between node, more easily find malicious node, thereby effectively guaranteed the accuracy that degree of belief is calculated.
As seen from Figure 10, when having a large amount of SMP in network, IGCMTrust can be good at resisting " fluctuation " behavior of this SMP, and its SDR is higher than the PeerTrust trust model, and the SDR of No-Trust remains on 0.5 left and right.Can find out, when T is respectively 0.8 and 0.85, when also namely the degree of belief of this SMP may be higher than the degree of belief of GP, the SDR of IGCMTrust still maintains higher level, and that the SDR of PeerTrust has just begun is lower, SDR uprises gradually subsequently, last and IGCMTrust is similar, this is because IGCMTrust has introduced the behavior uncertain factor, make malicious node in a single day do evil, can reach the purpose that malicious node is punished by the behavior uncertain factor, thereby guarantee the SDR of network integral body.
Trust model IGCMTrust computational methods of the present invention are better than PeerTrust trust model and the random No-Trust trust model of selecting node to download aspect successfully download rate.
Claims (5)
1. the P2P network trust cloud model computational methods based on interest group, is characterized in that comprising interest group division, interest group message process, trust cloud building process and node confidence evaluation procedure, and concrete steps are as follows:
The described interest group partition process of step 1 is as follows: at first determine the number of interest group, each interest group subject of great interest hobby number and network File number of resources; Then determine the group head node, according to the group number of setting, hobby number, file resource number in network, determine the affiliated group of node in network, last non-group of head node searches by query requests the group head node that theme interest is identical with it, and the group head node connects each other;
The described interest group message process of step 2 is as follows: determine the group message transfer mode according to the content of querying node, if theme interest is organized interior inquiry, if less important interest, inquiry between organizing;
The described trust cloud of step 3 building process is as follows: at first obtain node
iWith node
jThe direct interaction number of times
hSecondly compare interaction times
hWith given interaction times threshold value
HIf size is interaction times
hMore than or equal to given interaction times threshold value
HThe time, with node
iTo node
jThe reverse Normal Cloud generating algorithm of direct trust evaluation value input one dimension, obtain directly trusting cloud
If node
iWith node
jThe direct interaction number equal at 0 o'clock, node
iCan only measure node by the recommendation of other nodes
jTrust value, the calculated recommendation node
The recommendation confidence level
, only the recommendation evaluation of estimate of reliable recommended node is inputted the reverse Normal Cloud generating algorithm of one dimension, wherein the recommendation confidence level of reliable recommended node
Thereby, obtain the recommendation trust cloud
If interaction times
hGreater than 0, less than given interaction times threshold value
H, combined joint
iTo node
jDirect trust cloud and recommendation trust cloud jointly determine node
iTo node
jComprehensive trust cloud
The described node confidence evaluation procedure of step 4 is as follows: at first with node
iTo node
jComprehensive trust cloud
Expectation
As node
jTrust value
Then comparison node
jTrust value
With the good turn Evaluation threshold
GTSize, if
Less than the good turn Evaluation threshold
GT, node is not carried out the encouragement and penalty of degree of belief, if
More than or equal to the good turn Evaluation threshold
GT, comparison node
jThe behavior uncertain factor
With Evaluation threshold
GTSize is used for distinguishing malicious node and good will node; If node
jThe behavior uncertain factor
Less than behavior uncertain factor threshold value
, to node
jReward, improve its degree of belief, if node
jThe behavior uncertain factor
Greater than behavior uncertain factor threshold value
, to node
jPunish, reduce its degree of belief.
2. P2P network trust cloud model computational methods based on interest group according to claim 1 is characterized in that in described step 1, the interest group partiting step is as follows:
No. ID, step 1.1 setting group head node needs the group number of dividing into groups
n, the hobby number that each node has
m(
m<=
n), All Files sum in network;
Step 1.2 is according to No. ID, the group head node of setting, determine the group head node, according to the group number of setting, the hobby number, number of files in network is determined the affiliated group of node in network, the interest that node has and the quantity of documents that has thereof, wherein hobby can divide be the theme interest and less important interest, and the group that No. ID of theme interest is assigned to node is identical No. ID;
Non-group of head node of step 1.3 searched No. ID identical group head node of theme interest with it by group head node query requests;
After step 1.4 group head node is received a group node query requests, if find corresponding theme interest with own identical, with the neighbor node of this non-group of head node as oneself, non-group of head node also will be organized head node as the neighbor node of oneself simultaneously;
Step 1.5 group head node connects each other.
3. P2P network trust cloud model computational methods based on interest group according to claim 1 is characterized in that in described step 2, interest group message process step is as follows:
1 group of interior inquiry of step 2.
If be inquiry in group, the content of the querying node interest that is the theme, can inquire about by the following steps:
Step 2.1.1 sends a file polling request to the group head node, and the group head node sends in group other non-group of head node according to concrete query contents;
After non-group of head node of step 2.1.2 received the file polling request, according to the situation of itself, if there is corresponding content, this request is responded;
After step 2.1.3 query node is received response, connect with it, carry out file and download;
Inquire about between 2 groups of steps 2.
If the content that node will be inquired about is less important interest content, can inquire about by the following steps:
At first step 2.2.1 sends query requests between a group to the group head node of this group, and the group head node of this group finds it is not the content of this group according to concrete query contents, sends to the group head node of less important interest;
The less important interest group head node of step 2.2.2 sends to according to the content of inquiry the non-group of head node that is attached thereto in group;
Non-group of head node of the less important interest of step 2.2.3 received between group after query requests, according to the situation of itself, if there is corresponding content, this request responded;
After step 2.2.4 query node is received response, connect with it, carry out file and download.
4. P2P network trust cloud model computational methods based on interest group according to claim 1, trust cloud building process step in the step 3 under it is characterized in that as follows:
Step 3.1 structure is directly trusted cloud, if node
iWith node
jThere is direct interaction, i.e. the direct interaction number of times
h〉0, node
iObtain node according to own existing historical experience
iTo node
jDirect trust evaluation, by node
iTo node
jThe reverse Normal Cloud generating algorithm of direct trust evaluation value input one dimension, can obtain directly trusting cloud
, wherein
Be respectively expectation, entropy and super entropy;
If step 3.2 structure recommendation trust cloud is node
iWith node
jThere is not direct interaction, i.e. the direct interaction number of times
h=0, node
iCan only by with node
jThere is the recommended node of direct interaction experience to obtain node
iTo node
jTrust value, in order to guarantee recommended node
The recommendation reliability, the definition recommended node the recommendation confidence level
, only will recommend confidence level
The reverse Normal Cloud generating algorithm of recommendation evaluation of estimate input one dimension of recommended node, obtain the recommendation trust cloud
Recommended node wherein
Recommendation confidence level computational process as follows:
Step 3.2.1 node
With node
jAfter completing once transaction, according to the service quality that gets, with reference to the satisfaction standards of grading, provide evaluation of estimate, in conjunction with original evaluating data, the method by mean value calculation obtains node
jDirect degree of belief
Step 3.2.3 calculates
With
The absolute value of difference, if
Successful recommendation number of times
Add 1, otherwise, failed recommendation number of times
Add 1,
For recommending the confidence level threshold value, be service quality maximum difference good in trust evaluation;
Cloud and recommendation trust cloud are directly trusted in step 3.3 combination, if node
iWith node
jThe direct interaction number of times
hMore than or equal to given interaction times threshold value
HThe time, think node
iAccording to own existing historical experience, fully can decision node
j, no longer need to consider the recommendation trust of recommended node, node
iTo node
jComprehensive trust cloud
It is exactly node
iTo node
jDirect trust cloud
If node
iWith node
jThe direct interaction number of times
hEqual at 0 o'clock, node
iCan only be by the recommendation of other nodes, comprehensively trust this moment
Deteriorate to the recommendation trust cloud
If node
iWith node
jThe direct interaction number of times
hArrive 0
HIn the time of in scope, comprehensively trust cloud and jointly determined by direct trust cloud and recommendation trust cloud, the formula of computational methods is as follows:
Wherein
Be the logic multiply operator,
Be the logic add operator,
Embodied the lay particular stress on degree of node self to direct trust and recommendation trust, its size can be regulated, corresponding
,
,
Computing formula is:
5. P2P network trust cloud model computational methods based on interest group according to claim 1 is characterized in that in described step 4, the node confidence evaluation procedure is carried out in such a way:
The behavior uncertain factor
, bonding behavior uncertain factor threshold value
, introduce and reward the factor
, node is rewarded, improve its degree of belief, introduce penalty factor
, node is punished, reduce its degree of belief, its method is as follows:
Node
iAccording to the node that gets
jComprehensive trust cloud
, will expect
As node
jTrust value
If,
Less than the good turn Evaluation threshold
GT, node is not carried out the encouragement and penalty of degree of belief; If
More than or equal to the good turn Evaluation threshold
GT, computing node at first
jThe behavior uncertain factor
If, greater than behavior uncertain factor threshold value
, node is punished, reduce its degree of belief, the punishment formula is:
, wherein
Be penalty factor, span is
If node
jThe behavior uncertain factor less than behavior uncertain factor threshold value
, node is rewarded, improve its degree of belief, the award formula is:
, wherein
For rewarding the factor, its span is
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201310110038 CN103152436A (en) | 2013-04-01 | 2013-04-01 | P2P (peer-to-peer) internet trust cloud model computing method based on interest group |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201310110038 CN103152436A (en) | 2013-04-01 | 2013-04-01 | P2P (peer-to-peer) internet trust cloud model computing method based on interest group |
Publications (1)
Publication Number | Publication Date |
---|---|
CN103152436A true CN103152436A (en) | 2013-06-12 |
Family
ID=48550295
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 201310110038 Withdrawn CN103152436A (en) | 2013-04-01 | 2013-04-01 | P2P (peer-to-peer) internet trust cloud model computing method based on interest group |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103152436A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103561047A (en) * | 2013-07-31 | 2014-02-05 | 南京理工大学 | P2P network trust cloud model calculating method based on interest groups |
CN103985018A (en) * | 2014-06-03 | 2014-08-13 | 杭州师范大学 | Method for analyzing networked transaction data collection and transaction completion degree |
CN104022892A (en) * | 2014-03-31 | 2014-09-03 | 东莞中山大学研究院 | Network optimization method based on user interest grouping |
CN105678590A (en) * | 2016-02-07 | 2016-06-15 | 重庆邮电大学 | topN recommendation method for social network based on cloud model |
CN105868334A (en) * | 2016-03-28 | 2016-08-17 | 云南财经大学 | Personalized film recommendation method and system based on feature augmentation |
CN106941695A (en) * | 2017-05-19 | 2017-07-11 | 陕西师范大学 | A kind of data distributing method matched based on interest in opportunistic network |
CN107025436A (en) * | 2017-03-13 | 2017-08-08 | 西安电子科技大学 | A kind of self refresh human intrusion detection method based on confidence level |
CN107358533A (en) * | 2017-06-15 | 2017-11-17 | 桂林理工大学 | A kind of user of Web Community recommends method and system |
CN108710685A (en) * | 2018-05-21 | 2018-10-26 | 中国联合网络通信集团有限公司 | The determination method and device of target object confidence level |
CN109151051A (en) * | 2018-09-12 | 2019-01-04 | 南昌航空大学 | Data safety Enhancement Method under a kind of cloud computing environment |
-
2013
- 2013-04-01 CN CN 201310110038 patent/CN103152436A/en not_active Withdrawn
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103561047A (en) * | 2013-07-31 | 2014-02-05 | 南京理工大学 | P2P network trust cloud model calculating method based on interest groups |
CN104022892A (en) * | 2014-03-31 | 2014-09-03 | 东莞中山大学研究院 | Network optimization method based on user interest grouping |
CN103985018A (en) * | 2014-06-03 | 2014-08-13 | 杭州师范大学 | Method for analyzing networked transaction data collection and transaction completion degree |
CN103985018B (en) * | 2014-06-03 | 2017-01-25 | 杭州师范大学 | Method for analyzing networked transaction data collection and transaction completion degree |
CN105678590A (en) * | 2016-02-07 | 2016-06-15 | 重庆邮电大学 | topN recommendation method for social network based on cloud model |
CN105868334B (en) * | 2016-03-28 | 2020-10-30 | 云南财经大学 | Feature incremental type-based personalized movie recommendation method and system |
CN105868334A (en) * | 2016-03-28 | 2016-08-17 | 云南财经大学 | Personalized film recommendation method and system based on feature augmentation |
CN107025436A (en) * | 2017-03-13 | 2017-08-08 | 西安电子科技大学 | A kind of self refresh human intrusion detection method based on confidence level |
CN106941695A (en) * | 2017-05-19 | 2017-07-11 | 陕西师范大学 | A kind of data distributing method matched based on interest in opportunistic network |
CN106941695B (en) * | 2017-05-19 | 2018-06-08 | 陕西师范大学 | It is a kind of in opportunistic network to be based on the matched data distributing method of interest |
CN107358533A (en) * | 2017-06-15 | 2017-11-17 | 桂林理工大学 | A kind of user of Web Community recommends method and system |
CN108710685A (en) * | 2018-05-21 | 2018-10-26 | 中国联合网络通信集团有限公司 | The determination method and device of target object confidence level |
CN109151051A (en) * | 2018-09-12 | 2019-01-04 | 南昌航空大学 | Data safety Enhancement Method under a kind of cloud computing environment |
CN109151051B (en) * | 2018-09-12 | 2020-12-08 | 南昌航空大学 | Data security enhancement method in cloud computing environment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103152436A (en) | P2P (peer-to-peer) internet trust cloud model computing method based on interest group | |
CN103561047A (en) | P2P network trust cloud model calculating method based on interest groups | |
Guo et al. | A classification of trust computation models for service-oriented internet of things systems | |
Liu et al. | Stereotrust: a group based personalized trust model | |
Rosaci | Trust measures for competitive agents | |
Tian et al. | R2Trust, a reputation and risk based trust management framework for large-scale, fully decentralized overlay networks | |
Chen et al. | Optimization of the trust propagation on supply chain network based on blockchain plus | |
Wang et al. | A trusted consensus fusion scheme for decentralized collaborated learning in massive IoT domain | |
CN103279512A (en) | Method for using most influential node in social network to achieve efficient viral marketing | |
Fan et al. | Behavior-based reputation management in P2P file-sharing networks | |
Meng et al. | @ Trust: A trust model based on feedback-arbitration in structured P2P network | |
US8914505B2 (en) | Methods and apparatus for tuning a network for optimal performance | |
Laoutaris et al. | Bounded budget connection (BBC) games or how to make friends and influence people, on a budget | |
Tang et al. | Combining global and local trust for service recommendation | |
CN102006305B (en) | Method for suppressing malicious behavior based on distributed recommendable reputation in peer-to-peer (P2P) network | |
CN116366669A (en) | Consensus method based on reputation value weight balance suitable for crowdsourcing system | |
Li et al. | Federated learning based on Stackelberg game in unmanned-aerial-vehicle-enabled mobile edge computing | |
Wang et al. | Characterizing economic and social properties of trust and reputation systems in P2P environment | |
Wu et al. | Privacy-preserving trust management method based on blockchain for cross-domain industrial IoT | |
CN115640305B (en) | Fair and reliable federal learning method based on blockchain | |
Wang et al. | The comparison of distributed P2P trust models based on quantitative parameters in the file downloading scenarios | |
Wang et al. | A super‐agent‐based framework for reputation management and community formation in decentralized systems | |
Zhang et al. | Collective ratings for online communities with strategic users | |
Nishi et al. | Optimality analysis of locality-aware tit-for-tat-based P2P file distribution | |
CN107018027A (en) | A kind of link prediction method based on Bayesian Estimation and common neighbor node degree |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C04 | Withdrawal of patent application after publication (patent law 2001) | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20130612 |