CN101765231B - Wireless sensor network trust evaluating method based on fuzzy logic - Google Patents

Abstract

The invention relates to a wireless sensor network trust evaluating method based on a fuzzy logic, which comprises the following steps that: (1) an evaluating node initializes the direct trust vectors of evaluated nodes; (2) the evaluating node periodically updates the direct trust vectors of the evaluated nodes by using a fuzzy inference method; (3) the evaluating node collects the indirect trust vectors of recommended nodes on the evaluated nodes; (4) by utilizing the direct trust vectors and the indirect trust vectors of the evaluated nodes, the evaluating node calculates the comprehensive trust vectors of the evaluated nodes; and (5) according to the comprehensive trust vectors of the evaluated nodes, the evaluating node carries out the judgement of the trust classification and the decision of the network collaboration on the evaluated nodes. The invention realizes the trust quantification by adopting the trust value definition based on a fuzzy set theory and using a fuzzy inference algorithm, thereby effectively handling the problem of the subjective fuzziness of the trust and simulating the subjective cognitive process of the trust inference. Compared with the traditional algorithm, the wireless sensor network trust evaluating method has higher sensitivity, accuracy and generality.

Description

A kind of wireless sensor network trust evaluating method based on fuzzy logic

Technical field

The present invention relates to the trust evaluation in wireless sensor network trust the administrative skill field, particularly wireless sensor network, a kind of wireless sensor network trust evaluating method based on fuzzy logic.

Background technology

Wireless sensor network integrates sensor technology, micro electro mechanical system (MEMS) technology, embedded computing technique, wireless communication technology and distributed information processing, utilize the sensor node that is deployed in a large number in the monitored area, perception information in the collection network coverage, by the communication of multi-hop, the information after collecting, handling is offered the terminal use.Wireless sensor network does not need the network support fixed, has characteristics such as rapid deployment, survivability be strong, is widely used in a plurality of fields such as military surveillance, environmental monitoring, Industry Control, medical monitoring.

Wireless sensor network usually is applied to battlefield investigation or security monitoring, is deployed under the environment that lacks basic maintenance.Sensor node is very easily captured by the enemy, and is transformed into malicious node to network implementation attack and destruction.Traditional wireless sensor network security mechanism depends on key code system, grasps effective network security key owing to be captured node, and secret key safety mechanism can't solve its internal security problem that causes.And trust management system does not rely on netkey, can assess the credibility of node by observing behavior and the feature of node, and searches the enforcement of malicious node and the measure of adjustment network security by this, has effectively remedied the deficiency of secret key safety mechanism.Trust evaluation is core and the key of trust management, has determined the validity of whole trust management system.

Existing wireless sensor network trust evaluating method is broadly divided into three classes: 1. based on the trust evaluation of classic scheme.These class methods are added up corresponding trust-factor by the diverse network behavior of observing node, and with the weighted average of all trust-factor trust value as node.As, people such as Crosby propose to add up the trust-factor such as the rate of giving out a contract for a project, forward rate and packet percentage of head rice of node, obtain the node trust value by weighted average; 2. based on the trust evaluation of Bayes principle.This class methods are subjective hypothesis node prior distribution of trusting earlier, and the posteriority that utilizes the statistics of meshed network behavior quality that node is trusted distributes and carries out Bayesian inference, and the mathematic expectaion that the gained posteriority is distributed is as the node trust value.As, the RFSN trust evaluation framework that people such as Ganeriwal propose; 3. based on the trust evaluation of experience.These class methods are utilized the trust value of certain experience formula computing node according to the statistics of meshed network behavior performance.As, people such as Hur are by the trust value of the authenticity computing node of analysis node Monitoring Data.The simple statistical method of this three class methods utilization has realized the analysis of uncertainty to the node trust, satisfied the lightweight requirement of sensor node, but the uncertainty that will trust is equal to randomness fully, has ignored subjective ambiguity, has influenced the accuracy of trust evaluation.In addition, in the trust evaluation based on classic scheme, the weighted average of trust-factor finally makes trust management lower to the single malicious node sensitivity of attack form, and its trust-factor is calculated, and weight is also very difficult to be determined; In the method for evaluating trust based on Bayes principle, the subjectivity of prior distribution hypothesis has increased the weight of the uncertainty of trusting, and the quality classification simple to the meshed network behavior finally also makes trust management lower to the single malicious node sensitivity of attack form; Only investigate the single trust-factor of node based on the method for evaluating trust of experience, have one-sidedness, versatility is poor.

In scientific research, the blooming that is subjected to human subjective cognitive ability restriction and produces, U.S. cybernetist Zadeh has proposed fuzzy set theory as far back as nineteen sixty-five, he has set forth the implication of ambiguity exactly, formulated the mathematical method (degree of membership, membership function, fuzzy subset etc.) of dividing ambiguity, for the development of fuzzy theory necessary base.Fuzzy theory has caused that the experts and scholars of various fields pay close attention to, and develops very fastly, has derived a series of new branch of science such as fuzzy reasoning, fuzzy control, fuzzy statistics.Expression and processing that fuzzy set theory and fuzzy reasoning method are respectively subjective trust provide strong tool.In the trust evaluation of wireless sensor network, use fuzzy theory, can accurately express the subjective ambiguity of trust, the subjective cognitive process of reasoning is trusted in simulation, solves randomness and subjective uncertainty problem that node is trusted simultaneously, improves accuracy, versatility and the sensitivity of trust evaluation.

Summary of the invention

The purpose of this invention is to provide a kind of wireless sensor network trust evaluating method based on fuzzy logic, solve the problems referred to above that existing method for evaluating trust exists, realize the accurate assessment of subjective trust in the wireless sensor network.

A kind of wireless sensor network trust evaluating method based on fuzzy logic of the present invention adopts distributed algorithm, requires the assessment that adjacent node is trusted mutually in the network, and the main body of assessment is called the assessment node, and object is called evaluated node;

This appraisal procedure of the present invention, trust value employing confidence level and two kinds of representations of trust vector to node specifically are defined as:

If node j is the neighbor node of node i, node i is carried out trust evaluation to node j; Wherein, i and j represent the ID of node in the network, are natural number;

Confidence level t: the assessment node i is to the confidence level t of evaluated node j _IjExpression, interval is [0,1], t _IjBe worth more big, the assessment node i more high to the trusting degree of evaluated node j;

Trust vector T: trust the definition of vector based on fuzzy set theory.The trusted degree of node is divided into " insincere ", " uncertain ", " more credible " and " definitely credible " four ranks; Domain at node credibility marks off and four fuzzy subset L that rank is corresponding _T1, L _T2, L _T3And L _T4, and set up corresponding membership function μ _T1(t), μ _T2(t), μ _T3(t) and μ _T4(t); Consider the restriction of node resource, membership function adopts triangle or trapezoidal function form; Then the vector representation of node trust is T=(v ₁, v ₂, v ₃, v ₄), the individual component v of k (k=1,2,3,4) of vector _kThe expression node is to the degree of membership of k level of trust; Among the present invention, DT, IT and CT represent respectively to assess node to the direct trust vector of evaluated node, trust vector sum indirectly and comprehensively trust vector; If the assessment node i is to the confidence level t of evaluated node j _Ij=t ^*, then assess node i to the vectorial DT of direct trust of evaluated node j _Ij=(μ _T1(t ^*), μ _T2(t ^*), μ _T3(t ^*), μ _T4(t ^*)).

A kind of wireless sensor network trust evaluating method based on fuzzy logic of the present invention, concrete steps are as follows:

Step 1: directly trusting of the evaluated node of assessment node initializing is vectorial, and concrete grammar is:

Owing to lack prior information, the assessment node i can't be determined the confidence level of evaluated node j fully, and the component assignment that therefore will directly trust corresponding " uncertain " level of trust in the vector is 1, and other component assignment is 0, then directly trusts the initial value of vector ${\mathrm{DT}}_{\mathrm{ij}}^0=\left(\mathrm{0,1,0,0}\right).$

Step 2: the assessment node uses fuzzy reasoning method to be updated periodically the direct trust vector of evaluated node, and the concrete practice is:

(1) establishing the assessment node i constantly finishes 0 evaluated node j is directly trusted vectorial DT _IjInitialization, then assess node i and observe the network behavior of evaluated node j at l τ to (l+1) τ period (l 〉=0, τ is the update cycle of directly trusting vector), and add up evaluated node j trust-factor γ constantly at (l+1) τ ₁, γ ₂..., γ _mThe changes delta γ of (trust-factor of establishing node has m) ₁, Δ γ ₂..., Δ γ _mThe assessment node i adopts the communication pattern that mixes reception by the observation of monitor channel realization to evaluated node j network behavior;

(2) (l+1) τ constantly, the assessment node i dynamically arranges corresponding renewal weight according to the variation tendency of evaluated node j trust-factor, establishes trust-factor γ _s(s=1,2 ..., m) be changed to Δ γ _s, then it upgrades weights W _sValue is:

Wherein, 0＜W _L＜＜W _H＜1;

(3) the assessment node i is upgraded all trust-factor values of evaluated node j:

${\mathrm{\γ}}_s^{l+1}=(1-W_i)\×{\mathrm{\γ}}_s^l+W_s\×\mathrm{\Δ}{\mathrm{\γ}}_s---\left(2\right)$

Wherein, γ _s ^lAnd γ _s ^L+1Be respectively l τ and (l+1) the τ trust-factor value of evaluated node j constantly;

(4) the assessment node i is according to γ ₁ ^L+1, γ ₂ ^L+1..., γ ₃ ^L+1, fuzzy reasoning is from the confidence level t to evaluated node j in (l+1) τ moment _Ij ^L+1, and with t _Ij ^L+1The membership function of each level of trust correspondence of substitution, can obtain assessing node i to evaluated node j at (l+1) τ vectorial DT of direct trust constantly _Ij ^L+1:

${\mathrm{DT}}_{\mathrm{ij}}^{l+1}=({\mathrm{\μ}}_{T1}\left(t_{\mathrm{ij}}^{l+1}\right),{\mathrm{\μ}}_{T2}\left(t_{\mathrm{ij}}^{l+1}\right),{\mathrm{\μ}}_{T3}\left(t_{\mathrm{ij}}^{l+1}\right),{\mathrm{\μ}}_{T4}\left(t_{\mathrm{ij}}^{l+1}\right))---\left(3\right)$

Wherein, the assessment node i is according to the trust-factor value of evaluated node j, and fuzzy reasoning self to the concrete grammar of evaluated node j confidence level is:

1. assess node i and before trust evaluation, set up the fuzzy inference rule of node credibility in advance, and the fuzzy implication relation R between extraction trust-factor and the node credibility _γ-t, be specially:

A. the fuzzy inference rule of node credibility must be set up before trust evaluation in advance, and concrete method for building up is:

At first, with trust-factor γ _s(s=1,2 ..., m) be divided into " bad ", " generally " and " good " be totally three good and bad ranks, and set up corresponding fuzzy subset r at its domain _{S, 1}, r _{S, 2}, r _{S, 3}And membership function μ _{Rs, 1}(γ _s), μ _{Rs, 2}(γ _s), μ _{Rs, 3}(γ _s); Consider the restriction of node resource, membership function adopts triangle or trapezoidal function form;

Secondly, according to general knowledge and the experience of trusting reasoning, (Q=3 builds together to set up the fuzzy inference rule of node credibility ^mThe bar rule), the standard of foundation is:

When arbitrary trust-factor was " bad ", then the confidence level of node was " insincere ";

When the trust-factor of minority is " good ", and when not having trust-factor to be " bad ", then the confidence level of node is " uncertain ";

When most of trust-factor are " good ", and when not having trust-factor to be " bad ", then the confidence level of node is " more credible ";

When all trust-factor were " good ", then the confidence level of node was " definitely credible ";

B. assess node i according to fuzzy inference rule, extract fuzzy implication relation R between trust-factor and the node credibility _γ-tConcrete grammar be:

At first, the implication of extracting under the wall scroll fuzzy rule concerns T _α(α=1,2 ..., Q):

Wherein, μ _R1 ^α(γ ₁), μ _R2 ^α(γ ₂) ..., μ _Rm ^α(γ _m) be other membership function of the good and bad level of each trust-factor under the α bar fuzzy rule, μ _T ^α(t) be the membership function of α bar fuzzy rule lower node level of trust;

Secondly, the implication relation under all fuzzy rules is carried out comprehensively obtaining implication relation R _γ-t:

2. assess node i with the trust-factor value γ of evaluated node j ₁ ^*, γ ₂ ^*..., γ _m ^*With R _γ-tIt is synthetic to carry out reasoning, obtains the fuzzy output μ of node credibility _T ^*(t):

3. assessing node i utilizes gravity model appoach with μ _T ^*(t) the confidence level t to evaluated node j is tried to achieve in reverse gelatinization _Ij:

$t_{\mathrm{ij}}=\mathrm{COG}=\underset{{\mathrm{\μ}}_T^{*}}{\∫}{\mathrm{\μ}}_T^{*}\left(t\right)\·\mathrm{tdt}/\underset{{\mathrm{\μ}}_T^{*}}{\∫}{\mathrm{\μ}}_T^{*}\left(t\right)\mathrm{dt}---\left(7\right)$

Step 3: when needs carried out network cooperation with evaluated node, the assessment node was collected recommended node to the indirect trust vector of evaluated node, and concrete grammar is:

When the assessment node i need determine whether to carry out network cooperation with evaluated node j, broadcast the querying command that evaluated node is trusted vector indirectly towards periphery; Recommended node is after receiving querying command, and self is vectorial as trusting vector indirectly to directly trusting of assessment node j, sends to the assessment node i; Here, recommended node can only be the common neighbor node of assessment node and evaluated node.

Step 4: the assessment node is comprehensive weight with the confidence level of recommended node, and indirect trust vector and direct trust vector are synthesized, and obtains the comprehensive trust vector of evaluated node, and concrete grammar is:

If the assessment node i receives the individual recommended node of p (p 〉=0) to the indirect trust vector IT of evaluated node j _1j, IT _2j..., IT _PjThe assessment node i is with self confidence level t to these recommended nodes _I1, t _I2..., t _IpBe comprehensive weight, this p is trusted vector indirectly and directly trust vectorial DT _IjSynthesize, calculate the vectorial CT of comprehensive trust to evaluated node j _Ij:

${\mathrm{CT}}_{\mathrm{ij}}=\frac{{\mathrm{DT}}_{\mathrm{ij}}+\underset{u=1}{\overset{p}{\mathrm{\Σ}}}{t}_{\mathrm{iu}}\·{\mathrm{IT}}_{\mathrm{uj}}}{1+\underset{u=1}{\overset{p}{\mathrm{\Σ}}}{t}_{\mathrm{iu}}}---\left(8\right)$

Step 5: the assessment node is vectorial according to comprehensive trusts of evaluated node, judges the level of trust of evaluated node, and adjusts network cooperation behavior between self and evaluated node flexibly, and concrete grammar is:

The assessment node i will comprehensively be trusted vectorial CT _IjThe middle corresponding level of trust of largest component is as the level of trust under the evaluated node j; According to the level of trust of evaluated node j, whether the assessment node i determines the standard of carrying out network cooperation with evaluated node j to be: if evaluated node j belongs to " credible anything but " rank, then assess the node i refusal and cooperate with it; If evaluated node j belongs to " uncertain " rank, then assess node i and only carry out the lower network cooperation of risk with it; If evaluated node j belongs to " more credible " rank, then assess node and carry out the general network cooperation of risk with it; If evaluated node j belongs to " definitely credible " rank, then assess node i is carried out any risk with it network cooperation.

The invention has the advantages that:

(1) the present invention adopts fuzzy set theory that the node trust is carried out fuzzy classification, utilizes Fuzzy Logic Reasoning Algorithm to realize the quantification that node is trusted, and the subjective cognitive process of reasoning is trusted in simulation, has higher accuracy than conventional method;

(2) trust fuzzy inference rule based on the node of strictness, the malicious node single to the attack form has higher sensitivity than conventional method;

(3) at different network environments, can set up corresponding fuzzy rule according to practical experience, have stronger versatility.

Description of drawings

Fig. 1 is the wireless sensor network trust evaluating method of a kind of fuzzy logic of the present invention, the flow chart of trust evaluation;

Fig. 2 is for directly trusting the flow chart of vector period renewal among the present invention;

Fig. 3 is the schematic diagram of the evaluated node credibility of assessment node fuzzy reasoning among the present invention;

Fig. 4 obtains the schematic diagram of indirect trust for assessment node among the present invention;

Embodiment

The present invention is described in further detail below in conjunction with accompanying drawing.

A kind of wireless sensor network trust evaluating method based on fuzzy logic of the present invention adopts distributed algorithm, the assessment that requires the adjacent node in the network to trust mutually, and the main body of assessment is called the assessment node, and object is called evaluated node;

A kind of wireless sensor network trust evaluating method based on fuzzy logic of the present invention has provided confidence level and the expression mode of trusting two kinds of node trust values of vector, specifically is defined as:

If node j is the neighbor node of node i, node i is carried out trust evaluation to node j; Wherein, i and j represent the ID of node in the network, are natural number;

Confidence level t: the assessment node i is to the confidence level t of evaluated node j _IjExpression, interval is [0,1], t _IjBe worth more big, the assessment node i more high to the trusting degree of evaluated node j;

Trust vector T: trust the definition of vector based on fuzzy set theory.The trusted degree of node is divided into " insincere ", " uncertain ", " more credible " and " definitely credible " four ranks; Domain at node credibility marks off and four fuzzy subset L that rank is corresponding _T1, L _T2, L _T3And L _T4, and set up corresponding membership function μ _T1(t), μ _T2(t), μ _T3(t) and μ _T4(t); Consider the restriction of node resource, membership function adopts triangle or trapezoidal function form; Then the vector representation of node trust is T=(v ₁, v ₂, v ₃, v ₄), the individual component v of k (k=1,2,3,4) of vector _kThe expression node is to the degree of membership of k level of trust; Among the present invention, DT, IT and CT represent respectively to assess node to the direct trust vector of evaluated node, trust vector sum indirectly and comprehensively trust vector; If the assessment node i is to the confidence level t of evaluated node j _Ij=t ^*, then assess node i to the vectorial DT of direct trust of evaluated node j _Ij=μ _T1(t ^*), μ _T2(t ^*), μ _T3(t ^*), μ _T4(t ^*));

A kind of wireless sensor network trust evaluating method based on fuzzy logic of the present invention, the assessment node is realized by following steps to the trust evaluation flow process of evaluated node as shown in Figure 1:

Step 1: directly trusting of the evaluated node of assessment node initializing is vectorial, and concrete grammar is:

Owing to lack prior information, the assessment node i can't be determined the trusted degree of evaluated node j fully, and the component assignment that therefore will directly trust corresponding " uncertain " level of trust in the vector is 1, and other component assignment is 0, then directly trusts the initial value of vector ${\mathrm{DT}}_{\mathrm{ij}}^0=\left(\mathrm{0,1,0,0}\right).$

Step 2: the assessment node use fuzzy reasoning method be updated periodically evaluated node directly trust vector flow process as shown in Figure 2, concrete steps are:

(1) establishing the assessment node i constantly finishes 0 evaluated node j is directly trusted vectorial DT _IjInitialization, then assess node i and observe the network behavior of evaluated node j is observed to (l+1) τ period (l 〉=0, τ is the update cycle of directly trusting vector) at l τ, and add up evaluated node j trust-factor γ constantly at (l+1) τ ₁, γ ₂..., γ _mThe changes delta γ of (trust-factor of establishing node has m, and initial value is 0) ₁, Δ γ ₂..., Δ γ _mThe assessment node i adopts the communication pattern that mixes reception by the observation of monitor channel realization to evaluated node j network behavior;

(2) (l+1) τ constantly, the assessment node i dynamically arranges corresponding renewal weight according to the variation tendency of evaluated node j trust-factor, establishes trust-factor γ _s(s=1,2 ..., m) be changed to Δ γ _s, then it upgrades weights W _sValue is:

0＜W wherein _L＜＜W _H＜1;

(3) the assessment node i is upgraded all trust-factor values of evaluated node j:

${\mathrm{\γ}}_s^{l+1}=(1-W_i)\×{\mathrm{\γ}}_s^l+W_s\×\mathrm{\Δ}{\mathrm{\γ}}_s---\left(2\right)$

Wherein, γ _s ^lAnd γ _s ^L+1Be respectively l τ and (l+1) the τ trust-factor value of evaluated node j constantly;

(4) the assessment node i is according to γ ₁ ^L+1, γ ₂ ^L+1..., γ ₃ ^L+1, fuzzy reasoning is from the confidence level t to evaluated node j in (l+1) τ moment _Ij ^L+1, and with t _Ij ^L+1The membership function of each level of trust correspondence of substitution, can obtain assessing node i to evaluated node j at (l+1) τ vectorial DT of direct trust constantly _Ij ^L+1:

${\mathrm{DT}}_{\mathrm{ij}}^{l+1}=({\mathrm{\μ}}_{T1}\left(t_{\mathrm{ij}}^{l+1}\right),{\mathrm{\μ}}_{T2}\left(t_{\mathrm{ij}}^{l+1}\right),{\mathrm{\μ}}_{T3}\left(t_{\mathrm{ij}}^{l+1}\right),{\mathrm{\μ}}_{T4}\left(t_{\mathrm{ij}}^{l+1}\right))---\left(3\right)$

Wherein, the assessment node is according to the trust-factor value of evaluated node, the principle of the evaluated node credibility of fuzzy reasoning as shown in Figure 3, concrete grammar is:

1. according to the fuzzy inference rule of node credibility, the assessment node i is extracted the fuzzy implication relation R between trust-factor and the node credibility _γ-t, be specially:

A. fuzzy inference rule is set up well before trust evaluation in advance, and concrete method for building up is:

At first, with trust-factor γ _s(s=1,2 ..., m) be divided into " bad ", " generally " and " good " be totally three good and bad ranks, and set up corresponding fuzzy subset r at its domain _{S, 1}, r _{S, 2}, r _{S, 3}And membership function μ _{Rs, 1}(γ _s), μ _{Rs, 1}(γ _s), μ _{Rs, 1}(γ _s); Consider the restriction of node resource, membership function adopts triangle or trapezoidal function form;

Secondly, according to general knowledge and the experience of trusting reasoning, (Q=3 builds together to set up the fuzzy inference rule of node credibility ^mThe bar rule), the standard of foundation is:

When arbitrary trust-factor was " bad ", then the confidence level of node was " insincere ";

When the trust-factor of minority is " good ", and when not having trust-factor to be " bad ", then the confidence level of node is " uncertain ";

When most of trust-factor are " good ", and when not having trust-factor to be " bad ", then the confidence level of node is " more credible ";

When all trust-factor were " good ", then the confidence level of node was " definitely credible ";

B. assess node according to fuzzy inference rule, extract fuzzy implication relation R between trust-factor and the node credibility _γ-tConcrete grammar be:

At first, the implication of extracting under the wall scroll fuzzy rule concerns T _α(α=1,2 ..., Q):

Wherein, μ _R1 ^α(γ ₁), μ _R2 ^α(γ ₂) ..., μ _Rm ^α(γ _m) be other membership function of the good and bad level of each trust-factor under the α bar fuzzy rule, μ _T ^α(t) be the membership function of α bar fuzzy rule lower node level of trust;

Secondly, the implication relation under all fuzzy rules is carried out comprehensively obtaining implication relation R _γ-t:

2. assess node i with the trust-factor actual value γ of evaluated node j ₁ ^*, γ ₂ ^*..., γ _m ^*With R _γ-tIt is synthetic to carry out reasoning, obtains the fuzzy output μ of node credibility _T ^*(t):

3. assessing node i utilizes gravity model appoach with μ _T ^*(t) the confidence level t to evaluated node j is tried to achieve in reverse gelatinization _Ij:

$t_{\mathrm{ij}}=\mathrm{COG}=\underset{{\mathrm{\μ}}_T^{*}}{\∫}{\mathrm{\μ}}_T^{*}\left(t\right)\·\mathrm{tdt}/\underset{{\mathrm{\μ}}_T^{*}}{\∫}{\mathrm{\μ}}_T^{*}\left(t\right)\mathrm{dt}---\left(7\right)$

Step 3: when needs carried out network cooperation with evaluated node, the assessment node was collected recommended node to the principle of the indirect trust vector of evaluated node as shown in Figure 4, and concrete grammar is:

When the assessment node need carry out network cooperation with evaluated node, broadcast the querying command that evaluated node is trusted vector indirectly towards periphery; Recommended node is after receiving querying command, and self is vectorial as trusting vector indirectly to directly trusting of assessment node, sends to the assessment node; Here, recommended node can only be the common neighbor node of assessment node and evaluated node.

Step 4: the assessment node is comprehensive weight with the confidence level of recommended node, will trust vector indirectly and synthesize with direct trust vector, obtains the comprehensive trust vector of evaluated node, and concrete grammar is:

If the assessment node i receives the individual recommended node of p (p 〉=0) to the indirect trust vector IT of evaluated node j _1j, IT _2j..., IT _PjThe assessment node i is with self confidence level t to these recommended nodes _I1, t _I2..., t _IpBe comprehensive weight, this p is trusted vector indirectly and directly trust vectorial DT _IjSynthesize, calculate the vectorial CT of comprehensive trust to evaluated node j _Ij:

${\mathrm{CT}}_{\mathrm{ij}}=\frac{{\mathrm{DT}}_{\mathrm{ij}}+\underset{u=1}{\overset{p}{\mathrm{\Σ}}}{t}_{\mathrm{iu}}\·{\mathrm{IT}}_{\mathrm{uj}}}{1+\underset{u=1}{\overset{p}{\mathrm{\Σ}}}{t}_{\mathrm{iu}}}---\left(8\right)$

Step 5: the assessment node is vectorial according to comprehensive trusts of evaluated node, judges the level of trust of evaluated node, adjusts self and evaluated internodal network cooperation behavior flexibly, and concrete grammar is:

The assessment node i will comprehensively be trusted vectorial CT _IjThe middle corresponding level of trust of largest component is as the level of trust under the evaluated node j; According to the level of trust of evaluated node j, whether the assessment node i determines the standard of carrying out network cooperation with evaluated node j to be: if evaluated node j belongs to " credible anything but " rank, then assess the node i refusal and cooperate with it; If evaluated node j belongs to " uncertain " rank, then assess node i and only carry out the lower network cooperation of risk with it; If evaluated node j belongs to " more credible " rank, then assess node and carry out the general network cooperation of risk with it; If evaluated node j belongs to " definitely credible " rank, then assess node i is carried out any risk with it network cooperation.

Eventually the above, the present invention proposes a kind of wireless sensor network trust evaluating method based on fuzzy theory, provided the vector form definition of node trust value based on fuzzy set theory, utilize fuzzy reasoning method to quantize node and directly trust vector, and by direct, indirect comprehensive trust value of trusting the synthetic computing node of weighting of vector; Simulated and trusted the subjective cognitive process of deriving, had stronger versatility, and have higher sensitivity and accuracy than conventional method.

It should be noted last that, above embodiment is only unrestricted in order to technical scheme of the present invention to be described, although with reference to preferred embodiment the present invention is had been described in detail, those of ordinary skill in the art is to be understood that, can make amendment or be equal to replacement technical scheme of the present invention, and not break away from the spirit and scope of technical solution of the present invention.

Claims (1)

1. the wireless sensor network trust evaluating method based on fuzzy logic adopts distributed algorithm, requires the assessment that adjacent node is trusted mutually in the network, and the main body of assessment is called the assessment node, and object is called evaluated node;

Trust value to node adopts confidence level and trusts two kinds of representations of vector, specifically is defined as:

If node j is the neighbor node of node i, node i is carried out trust evaluation to node j; Wherein, i and j represent the ID of node in the network, are natural number;

Confidence level t: the assessment node i is to the confidence level t of evaluated node j _IjExpression, interval is [0,1], t _IjBe worth more big, the assessment node i more high to the trusting degree of evaluated node j;

Trust vector T: trust the definition of vector based on fuzzy set theory; The trusted degree of node is divided into " insincere ", " uncertain ", " more credible " and " definitely credible " four ranks; Domain at node credibility marks off and four fuzzy subset L that rank is corresponding _T1, L _T2, L _T3And L _T4, and set up corresponding membership function μ _T1(t), μ _T2(t), μ _T3(t) and μ _T4(t); Consider the restriction of node resource, membership function adopts triangle or trapezoidal function form; Then the vector representation of node trust is T=(v ₁, v ₂, v ₃, v ₄), k component v of vector _kThe expression node is to the degree of membership of k level of trust; Wherein, k=1,2,3,4; DT, IT and CT represent respectively to assess node to the direct trust vector of evaluated node, trust vector sum indirectly and comprehensively trust vector; If the assessment node i is to the confidence level t of evaluated node j _Ij=t ^*, then assess node i to the vectorial DT of direct trust of evaluated node j _Ij=(μ _T1(t ^*), μ _T2(t ^*), μ _T3(t ^*), μ _T4(t ^*));

It is characterized in that: the concrete steps of this appraisal procedure are as follows:

Step 1: directly trusting of the evaluated node of assessment node initializing is vectorial, and concrete grammar is:

Owing to lack prior information, the assessment node i can't be determined the confidence level of evaluated node j fully, and the component assignment that therefore will directly trust corresponding " uncertain " level of trust in the vector is 1, and other component assignment is 0, then directly trusts the initial value of vector ${\mathrm{DT}}_{\mathrm{ij}}^0=\left(\mathrm{0,1,0,0}\right);$

Step 2: the assessment node uses fuzzy reasoning method to be updated periodically the direct trust vector of evaluated node, and the concrete practice is:

(1) establishing the assessment node i constantly finishes 0 evaluated node j is directly trusted vectorial DT _IjInitialization, then assess node i and observe the network behavior of evaluated node j at l τ to (l+1) τ period, l 〉=0 wherein, τ is the update cycle of directly trusting vector, and adds up evaluated node j trust-factor constantly at (l+1) τ

Variation

Wherein, the trust-factor of node j has m, and the initial value of each trust-factor is 0; The assessment node i adopts the communication pattern that mixes reception by the observation of monitor channel realization to evaluated node j network behavior;

(2) (l+1) τ constantly, the assessment node i dynamically arranges corresponding renewal weight according to the variation tendency of evaluated node j trust-factor, establishes trust-factor

Be changed to S=1 wherein, 2 ..., m; Then it upgrades weights W _sValue is:

Wherein, 0＜W _L＜＜W _H＜1;

(3) the assessment node i is upgraded all trust-factor values of evaluated node j:

${\mathrm{\γ}}_s^{l+1}=(1-W_s)\×{\mathrm{\γ}}_s^l+W_s\×\mathrm{\Δ}{\mathrm{\γ}}_s---\left(2\right)$

Wherein,

With

Be respectively l τ and (l+1) the τ trust-factor value of evaluated node j constantly;

(4) assessment node i basis

Fuzzy reasoning is from the confidence level to evaluated node j in (l+1) τ moment

And will

The membership function of each level of trust correspondence of substitution, it is vectorial in (l+1) τ direct trust constantly to evaluated node j to obtain assessing node i

${\mathrm{DT}}_{\mathrm{ij}}^{l+1}=({\mathrm{\μ}}_{T1}\left(t_{\mathrm{ij}}^{l+1}\right),{\mathrm{\μ}}_{T2}\left(t_{\mathrm{ij}}^{l+1}\right),{\mathrm{\μ}}_{T3}\left(t_{\mathrm{ij}}^{l+1}\right),{\mathrm{\μ}}_{T4}\left(t_{\mathrm{ij}}^{l+1}\right))---\left(3\right)$

Wherein, the assessment node i is according to the trust-factor value of evaluated node j, and fuzzy reasoning self to the concrete grammar of evaluated node j confidence level is:

1. assess node i and before trust evaluation, set up the fuzzy inference rule of node credibility in advance, and the fuzzy implication relation between extraction trust-factor and the node credibility

Be specially:

A. the fuzzy inference rule of node credibility must be set up before trust evaluation in advance, and concrete method for building up is:

At first, with trust-factor

Be divided into " bad ", " generally " and " good " be totally three good and bad ranks, and set up corresponding fuzzy subset r at its domain _{S, 1}, r _{S, 2}, r _{S, 3}And membership function Consider the restriction of node resource, membership function adopts triangle or trapezoidal function form;

Secondly, according to general knowledge and the experience of trusting reasoning, set up the fuzzy inference rule of node credibility, Q=3 builds together ^mThe bar rule, the standard of foundation is:

When arbitrary trust-factor was " bad ", then the confidence level of node was " insincere ";

When the trust-factor of minority is " good ", and when not having trust-factor to be " bad ", then the confidence level of node is " uncertain ";

When most of trust-factor are " good ", and when not having trust-factor to be " bad ", then the confidence level of node is " more credible ";

When all trust-factor were " good ", then the confidence level of node was " definitely credible ";

B. assess node i according to fuzzy inference rule, extract fuzzy implication relation between trust-factor and the node credibility

Concrete grammar is:

At first, the implication of extracting under the wall scroll fuzzy rule concerns R _α, α=1,2 ..., Q

$R_{\mathrm{\α}}={\mathrm{\μ}}_{r1}^{\mathrm{\α}}\left({\mathrm{\γ}}_1\right)^{\mathrm{\μ}}_{r2}^{\mathrm{\α}}\left({\mathrm{\γ}}_2\right)^...^{\mathrm{\μ}}_{\mathrm{rm}}^{\mathrm{\α}}\left({\mathrm{\γ}}_m\right)^{\mathrm{\μ}}_T^{\mathrm{\α}}\left(t\right)---\left(4\right)$

Wherein, Be other membership function of the good and bad level of each trust-factor under the α bar fuzzy rule,

It is the membership function of α bar fuzzy rule lower node level of trust;

Secondly, the implication relation under all fuzzy rules is carried out comprehensively obtaining implication relation

2. assess node i with the trust-factor value of evaluated node j

With

It is synthetic to carry out reasoning, obtains the fuzzy output of node credibility

3. assessing node i utilizes gravity model appoach to incite somebody to action

The confidence level t to evaluated node j is tried to achieve in the reverse gelatinization _Ij:

$t_{\mathrm{ij}}=\mathrm{COG}=\underset{{\mathrm{\μ}}_T^{*}}{\∫}{\mathrm{\μ}}_T^{*}\left(t\right)\·\mathrm{tdt}/\underset{{\mathrm{\μ}}_T^{*}}{\∫}{\mathrm{\μ}}_T^{*}\left(t\right)\mathrm{dt}---\left(7\right)$

Step 3: when needs carried out network cooperation with evaluated node, the assessment node was collected recommended node to the indirect trust vector of evaluated node, and concrete grammar is:

When the assessment node i need determine whether to carry out network cooperation with evaluated node j, broadcast the querying command that evaluated node is trusted vector indirectly towards periphery; Recommended node is after receiving querying command, and self is vectorial as trusting vector indirectly to directly trusting of assessment node j, sends to the assessment node i; Here, recommended node can only be the common neighbor node of assessment node and evaluated node;

Step 4: the assessment node is comprehensive weight with the confidence level of recommended node, and indirect trust vector and direct trust vector are synthesized, and obtains the comprehensive trust vector of evaluated node, and concrete grammar is:

If the assessment node i receives p recommended node to the indirect trust vector IT of evaluated node j _1j, IT _2j..., IT _PjWherein, p 〉=0; The assessment node i is with self confidence level t to these recommended nodes _I1, t _I2..., t _IpBe comprehensive weight, this p is trusted vector indirectly and directly trust vectorial DT _IjSynthesize, calculate the vectorial CT of comprehensive trust to evaluated node j _Ij:

${\mathrm{CT}}_{\mathrm{ij}}=\frac{{\mathrm{DT}}_{\mathrm{ij}}+\underset{u=1}{\overset{p}{\mathrm{\Σ}}}{t}_{\mathrm{iu}}\·{\mathrm{IT}}_{\mathrm{uj}}}{1+\underset{u=1}{\overset{p}{\mathrm{\Σ}}}{t}_{\mathrm{iu}}}---\left(8\right)$

Step 5: the assessment node is vectorial according to comprehensive trusts of evaluated node, judges the level of trust of evaluated node, and adjusts network cooperation behavior between self and evaluated node flexibly, and concrete grammar is:

The assessment node i will comprehensively be trusted vectorial CT _IjThe middle corresponding level of trust of largest component is as the level of trust under the evaluated node j; According to the level of trust of evaluated node j, whether the assessment node i determines the standard of carrying out network cooperation with evaluated node j to be: if evaluated node j belongs to " credible anything but " rank, then assess the node i refusal and cooperate with it; If evaluated node j belongs to " uncertain " rank, then assess node i and only carry out the lower network cooperation of risk with it; If evaluated node j belongs to " more credible " rank, then assess node and carry out the general network cooperation of risk with it; If evaluated node j belongs to " definitely credible " rank, then assess node i is carried out any risk with it network cooperation.

Images