CN101835158B - Sensor network trust evaluation method based on node behaviors and D-S evidence theory - Google Patents

Abstract

The invention discloses a sensor network trust evaluation method based on node behaviors and a D-S evidence theory, comprising the following five steps: 1) designing various trust factor strategies for nodes in a wireless sensor network; 2) setting trust factor weights according to network application scenes and simultaneously calculating the node behavior coefficient mu to obtain the direct trust value and multiple indirect trust value of an evaluated object; 3) calculating a fuzzy subset membership function for each trust value by utilizing the concepts of membership and linguistic variables of a fuzzy set theory, performing fuzzy classification on the various the trust values to form the basic confidence function of the D-S evidence theory; 4) calculating the evidence difference of thedirect trust value and the indirect trust values of the evaluated node and altering the weights of the indirect trust values; and 5) adopting the Dempster synthesis rule to obtain the comprehensive trust value of the evaluated node and a final basic confidence designated value according to the altered trust weights. The invention solves the problem of difficult identification of malicious nodes in a network and ensures the safety of network data transmission.

Description

Sensor network trust evaluating method based on nodes ' behavior and D-S evidence theory

(1) technical field

The present invention relates to the appraisal procedure of trusting each other between a kind of neighbors based on nodes ' behavior strategy and D-S evidence theory, relate in particular to the sensor network trust evaluating method based on nodes ' behavior and D-S evidence theory.It belongs to safe trust evaluation technical field in the wireless sensor network.

(2) background technology

Wireless sensor network (wireless sensor networks, abbreviation WSNs) integrates sensor technology, MEMS (micro electro mechanical system) (MEMS) technology, embedded computing technique, wireless communication technology and distributed information processing, mutual with the external world by transducer, the functions such as data acquisition, processing, communication and management can be applicable to Industry Control and monitoring, home automation and consumer electronics, safety and military affairs, goods and materials tracking and the many-sides such as supply chain management, reading intelligent agriculture, environment sensing and health monitoring.

Along with WSNs uses day by day complexity, demand for security is more and more higher, is mainly manifested in: use wireless channel to make WSNs be easy to be subject to the attack of variety of ways such as passive wiretapping, initiatively invasion, information obstruction, information personation; Eavesdropping may make the enemy obtain security information; Active attack may make the enemy delete information, inserting error information, modification information or pretend to be a certain node, thereby has destroyed availability, integrality, confirmability and the non repudiation of network.Safeguard construction based on key can only solve the safety problem that external network attack causes, and mentality of designing and method all can't solve the network internal safety problem that internal node is captured or was lost efficacy and cause; Trust management can be identified malicious node, selfish node, and misdata etc. effectively solve the safety problem that causes after the network internal authorization node is captured.Because WSNs has the characteristics such as node resource is limited, network application is relatively single, the delegated strategy of WSNs is comparatively simple, need not to take the trust management of authorized certificate mode; But, trust valuation mechanism must satisfy the requirements such as self-organizing, robustness, energy efficient, Distributed Calculation, and embody node trust " difficulty, lose easily " characteristic, according to the comprehensive trust value of nodes ' behavior subjectivity and indeterminacy of calculation node.

At present the method for trust management mainly can be divided into two large classes: based on the distributed authorization system of trust chain with based on the network trust evaluating system of behavior.1. the distributed authorization based on trust chain allows to authorize individual the collection to be authorized to individual all relevant informations (voucher), checks consistency according to local policy and authorization requests by tactful inference engine, determines whether authorize; Namely authorize the individual qualified right that oneself is had to license to the object of oneself trusting.These class methods exist authorizes the excessive shortcoming of individual right, in case authorize individuality to be put in the bag, whole network just has the danger of collapse, therefore is unsuitable for being applied in the wireless sensor network.2. based on individuality in the network trust evaluating system of nodes ' behavior that all relevant informations are good, comprise behavior observation to evaluated individuality, with the mutual discipline of evaluated individuality and other individual suggestions etc., utilize suitable computation model to obtain the other side's trust value; This type of uses that resource is relatively less, node is equal, without characteristics such as centers, be applicable to wireless sensor network.

In the research of wireless sensor network trust evaluating method, the Trust Valuation Model based on Classical Probability Spaces that at present typical research work has this people such as (Crosby) of clo to propose, model utilizes the simple statistics mode to realize trust value calculating, and trust is incorporated in the election process of bunch head, adopt redundancy strategy and challenge response means, guarantee that the leader cluster node that elects is credible; Add the people such as Rewa (Ganeriwal) and proposed Trust Valuation Model based on bayesian theory, utilize the analysis of uncertainty of bayes method realization to trusting, but owing to lacking the wireless sensor network priori, therefore, its prior distribution knowledge is obtained by the supposition of subjectivity; For the subjective ambiguity of trusting, the people such as Tang Wen have proposed the Trust Valuation Model based on fuzzy logic, linguistic variable, fuzzy logic are incorporated in supervisor's trust management, use fuzzy IF-THEN rule that general knowledge and experience that the mankind trust reasoning are carried out modeling, realize the fuzzy Judgment to the node trust value; Cray this-people such as Butterworth base (Krasniewski) have proposed the TIBFIT agreement, utilize the base station that election of cluster head is adopted centralized trust management, leader cluster node is divided near the node the institute reporting event position to send report and send reports two groups, calculate respectively its trust value, think that the node in one group of the high trust value is credible.If find new bunch insincere, then re-elect a bunch head, can effectively avoid malicious node to serve as a bunch head.But centralized trust management pattern has increased the network service load, and passive trust decisions has reduced the convergence rate of election of cluster head; PTM is UBISEC (general fit calculation of safety) the research sub-project that European IST FP6 supports.It has defined based on dynamic trust model between the territory of general environment, mainly adopts the method for evidence theory to carry out modeling; The method is strictly punished malicious act, embodies and trusts the characteristic that is easy to that obtains difficulty, mistake, and have preferably computational convergence and extensibility; The people such as Hai Nieer (Hsieh) adopt clustering architecture to realize the safety of wireless sensor network; Comprise two large modules: 1. when cluster and the node of newly arriving add bunch, adopt the dynamic key authorization method, stop outside malicious node to add; 2. in running, a bunch interior nodes is monitored mutually, according to the difference of node role (bunch head, ordinary node), formulates different trust computational methods.

(3) summary of the invention

1, purpose: the objective of the invention is: the sensor network trust evaluating method based on nodes ' behavior and D-S evidence theory is provided, its embodies the characteristic of trust value " obtain difficulty, lose easily ", solve the impalpable problem of malicious node in the network, stop the information leakage problem that internal node is captured to cause, guaranteed the fail safe of network data transmission.

The present invention mainly trusts the security evaluation technology for node in the wireless sensor network and proposes, realistic sensor network energy, communication capacity, computing capability, the limited characteristic of storage capacity, but the malicious act of node in the efficient identification self-organizing network improves the adaptability of algorithm in actual environment.

2, technical scheme: for the location implementing method that the clear the present invention of elaboration proposes, define some relevant definition and concepts, the hypothesis evaluation node i is carried out trust evaluation to evaluated node j: wherein node i is called main body, and node j is called object.

Definition 1: packet receiving rate factor R PF _{I, j}(t)---in time period t and the t-1, the situation of change of node j packet receiving rate;

Definition 2: the success rate of giving out a contract for a project factor S PF _{I, j}(t)---in the time period t, node j sends the packet success rate;

Definition 3: forward rate factor TPF _{I, j}(t)---in the time period t, node j sends bag number and the relation of accepting the bag number;

Definition 4: consistency factor CPF _{I, j}(t)---in the time period t, node j sends the difference degree of packet and neighbors;

Definition 5: time rate factor TFF (t)---in the time period t, the time correlation sex factor of node j context;

Definition 6: availability factor factor HPF _{I, j}(t)---in the time period t, node i sends detection packet and receives feedback relationship;

Definition 7: level of security SG---different according to the network application occasion, formulate different coefficient of safetys;

The present invention is based on nodes ' behavior in conjunction with method for evaluating trust between the neighbors of D-S evidence theory, the method concrete steps are as follows:

Step 1: each trust-factor of node is generated strategy in the wireless sensor network, and concrete grammar is:

For the sensor node characteristic, the various Components of defined node degree of belief are without loss of generality, and suppose that node i (assessment node) is to node j (evaluated node) trust evaluation.

(1) packet receiving rate factor R PF _{I, j}(t): node j whenever receives a packet, needs to send the ACK feedback information and determines; Node i listens to the quantity of the ack msg bag that j sends, thereby obtains the packet receiving rate information of j; According to the node j packet receiving rate situation of change in time period t and the t-1, whether node j exists the personation replay attacks as can be known.If the variation of j packet receiving rate remains in (ξ, ξ), think that the j node is working properly; Its computing formula is:

${\mathrm{RPF}}_{i,j}\left(t\right)=\frac{{\mathrm{RP}}_{\mathrm{ij}}\left(t\right)-{\mathrm{RP}}_{\mathrm{ij}}(t-1)}{({\mathrm{RP}}_{\mathrm{ij}}\left(t\right)+{\mathrm{RP}}_{\mathrm{ij}}(t-1))/2}---\left(1\right)$

Wherein, RP _{I, j}(t) be packet receiving quantity;

(2) send success rate factor S PF _{I, j}(t): each packet that node sends stabs if having time.Suppose that j sends packet to node k, k is outside the communication range of i, and this moment, i can't monitor the transmission success rate of node j.Therefore, node i adopts and monitors the transmission success rate that j transmission identical data packet number of times is determined j.Different according to node time stamp, even packet content is identical, effective distinguishes data bag is still determined the transmission times of certain packet; Its computing formula is:

${\mathrm{SPF}}_{i,j}\left(t\right)=\frac{{\mathrm{SP}}_{\mathrm{ij}}\left(t\right)-{\mathrm{SF}}_{\mathrm{ij}}\left(t\right)}{{\mathrm{SP}}_{\mathrm{ij}}\left(t\right)+{\mathrm{SF}}_{\mathrm{ij}}\left(t\right)}---\left(2\right)$

Wherein, SF _{I, j}(t) be the quantity of giving out a contract for a project;

(3) forward rate factor TPF _{I, j}(t): because WSNs self, most nodes can't with base station (BS) direct communication, therefore need next-hop node to carry out data retransmission.Suppose that node k sends packet to node j, k is outside the communication range of i, and this moment, i can't monitor the packet receiving quantity of node j, so node i adopts and monitor j and send the ACK feedback information and determine j packet receiving quantity, and the monitoring node j situation of giving out a contract for a project is determined its quantity of giving out a contract for a project.According to its rate of change situation, effectively avoid cesspool to attack and Sybil attack, and effectively whether recognition node is selfish node; Its computing formula is:

${\mathrm{TPF}}_{i,j}\left(t\right)=\frac{{\mathrm{FP}}_{\mathrm{ij}}\left(t\right)-{\mathrm{FP}}_{\mathrm{ij}}(t-1)}{({\mathrm{FP}}_{\mathrm{ij}}\left(t\right)+{\mathrm{FP}}_{\mathrm{ij}}(t-1))/2}---\left(3\right)$

Wherein, FP _{I, j}(t) for transmitting bag quantity;

(4) consistency factor CPF _{I, j}(t): because the node data bag has spatial coherence, packet is more similar between the localized network neighbor node, therefore introduces the consistency factor, avoids malicious node that packet is distorted.Node i is obtained the forwarding data bag of j at random, compares with its data bag, if i, the j variance rate remains in (ξ, ξ), thinks that the j node is working properly; Its computing formula is:

${\mathrm{CPF}}_{i,j}\left(t\right)=\frac{{\mathrm{NEP}}_{\mathrm{ij}}\left(t\right)}{{\mathrm{EP}}_{\mathrm{ij}}\left(t\right)+{\mathrm{NEP}}_{\mathrm{ij}}\left(t\right)}---\left(4\right)$

Wherein, EP _{I, j}(t) be consistent bag quantity, NEP _{I, j}(t) be inconsistent bag quantity;

(5) time granularity factor TFF (t): because trust value has time, content context relation, the node trust value is to change at a upper time segment base plinth, therefore must the joining day factor be analyzed by trust value.The time granularity size is analyzed as the case may be; If granularity is excessive, then comprehensive trust value is subjected to historical influence excessive, and assessment may make a mistake to node; If undersized, then trust value relies on excessive to the single time period; Therefore, in the situation that level of security is higher, TFF (t)=0.75, in the lower situation of level of security, trust-factor is TFF (t)=0.25.Generally speaking, TFF (t)=0.5

(6) availability factor factor HPF _{I, j}(t): because there is unreachable situation in the factors such as channel, environment between neighbors; Only have between neighbors and can communicate and monitor, above various trust-factor are just meaningful, therefore introduce the availability factor.In the time period t, node i sends at random m HELLO bag and surveys, if receive that j to the feedback information ACK-HELLO packet of HELLO, then thinks can reach between node; Its computing formula is:

${\mathrm{HPF}}_{i,j}\left(t\right)=\frac{{\mathrm{NACK}}_{H_{\mathrm{ij}}}\left(t\right)}{{\mathrm{ACK}}_{H_{\mathrm{ij}}}\left(t\right)+\mathrm{NAC}{K}_{H_{\mathrm{ij}}}\left(t\right)}---\left(5\right)$

Wherein, ACK _{I, j}(t) for answering bag quantity, NACK _{I, j}(t) for not answering bag quantity;

(7) level of security SG:WSNs is in different applied environments and scene, and required level of security is different; The network security requirement difference that for example is applied to battlefield and environmental monitoring is huge.Therefore, SG adopts exponential manner to calculate; When safety requirements is higher, SG=3; When low, SG=1; When general, SG=2.

Step 2: according to the network application scene, the trust-factor weight is set, simultaneously computing node behavior coefficient μ obtains direct trust value and a plurality of indirect trust values of evaluated object, and concrete grammar is:

(1) level of security is relevant with network application environment, when safety requirements is higher, and level of security SG=3; When low, SG=1; When general, SG=2.In like manner, the time granularity size when safety requirements is higher, time rate factor TFF (t)=0.75; When low, TFF (t)=0.25; When general, TFF (t)=0.5.Other trust-factor generally speaking coefficient equates, and summation is 1.

(2) in order to guarantee the reasonability of trust evaluation, need to formulate the behavior coefficient μ in the different time sections.For example: in time period t-1, i and j mutual 10000 times altogether, wherein 5000 satisfaction; In the time period t, i and j mutual 1000 times altogether, wherein 500 satisfaction; Satisfied behavior ratio is 1/2, and obviously the trust value of time period t-1 is more credible; Therefore, the behavior coefficient formula is:

$\mathrm{\μ}=\frac{S_{\mathrm{ij}}\left(t\right)/(F_{\mathrm{ij}}\left(t\right)+S_{\mathrm{ij}}\left(t\right))}{S_{\mathrm{ij}}(t-1)/(F_{\mathrm{ij}}(t-1)+S_{\mathrm{ij}}(t-1))}---\left(6\right)$

Wherein, S _Ij(t) be number of success, F _Ij(t) be the frequency of failure; S _Ij, F _IjBe respectively average successful, the miss data bag of each time period in the network operation process.

(3) according to each trust-factor in the step 1, in conjunction with node resource and network operation actual environment, the COMPREHENSIVE CALCULATING node i is to the direct trust value DTE of node j _{I, j}(t) be:

DTE _i，j(t)＝μ ^SG*TFF(t)*(1-w ₁*RPF _i，j(t)-w ₂*SPF _i，j(t)-w ₃*TPF _i，j(t)-w ₄*CPF _i，j(t)-w ₅*HPF _i，j(t))

+(1-TFF(t))*DTE _i，j(t-1) (7)

Wherein, w ₁, w ₂, w ₃, w ₄, w ₅Be each trust-factor coefficient, adjustable according to concrete application; RPF _{I, j}(t) be the packet receiving rate factor, SPF _{I, j}(t) be the success rate factor of giving out a contract for a project, TPF _{I, j}(t) be the forward rate factor, CPF _{I, j}(t) be the consistency factor, HPF _{I, j}(t) be the availability factor factor; And w ₁+ w ₂+ w ₃+ w ₄+ w ₅=1 (8)

Assessment main body i collects other nodes to the indirect recommendation of trust value conduct of evaluated object j.In order to reduce network energy consumption and the communication load of WSNs, and prevent from trusting circular recursion, recommendation is only limited to the common neighbors of i, j to the direct trust value of j.Suppose that k is i, the common neighbours of j transmit the decline principle according to trusting, and recommendation trust recommends neighborhood to be illustrated in figure 3 as: ITE as shown in Figure 2 _{I, j}(t)=DTE _{I, k}(t) * DTE _{K, j}(t) (9)

Step 3: utilize the degree of membership of fuzzy set theory and the concept of linguistic variable, calculate fuzzy subset's membership function of various trust values, various trust values are carried out fuzzy classification, form the basic confidence level function of D-S evidence theory; Fuzzy classification as shown in Figure 2.Concrete grammar is:

The fuzzy classification of (1) node being trusted: at first node is divided into " insincere " " uncertain " and " definitely credible " three level of trusts.Secondly, corresponding these three level of trusts, domain [0 at node trust value T, 1] upper fuzzy subset T1, T2 and the T3 of dividing, set up membership function μ 1 (t), μ 2 (t) and μ 3 (t), wherein, membership function μ 1 (t), μ 2 (t) and μ 3 (t) need satisfy μ 1 (t)+μ 2 (t)+μ 3 (t)=1.

(2) according to fuzzy membership function, node i is expressed as vector form to the trusting degree of node j, and direct, indirect, comprehensive trust vector is respectively:

${\mathrm{DTE}}_{\mathrm{ij}}\left(t\right)=(m_{\mathrm{ij}}^D\left(\right\{T\left\}\right),m_{i,j}^D\left(\right\{T\left\}\right),m_{i,j}^D\left(\right\{T,-T\left\}\right))$

${\mathrm{ITE}}_{i,j}^k\left(t\right)=(m_{i,j}^k\left(\right\{T\left\}\right),m_{i,j}^k\left(\right\{-T\left\}\right),m_{i,j}^k\left(\right\{T,-T\left\}\right))---\left(10\right)$

AT _i，j(t)＝(m _i，j({T})，m _i，j({-T})，m _i，j({T，-T}))

The following formula symbol description is as follows: DTE _{I, j}(t) be direct trust vector, ITE _{I, j}K (t) is indirect trust vector, AT _{I, j}(t) be synthetic trust vector; M ({ T}), m (T}), m (T ,-T}) correspondingly respectively determine credible component, insincere component, uncertain component.

The membership function that utilizes node to trust classification calculates object of assessment j to " insincere ", and the degree of membership of " uncertain " and " definitely credible " three level of trusts is u1, u2 and u3:

$\left\{\begin{array}{c}{u}_1={\mathrm{\μ}}_1\left({\mathrm{DTE}}_{i,j}\left(t\right)\right)\\ u_2={\mathrm{\μ}}_2\left({\mathrm{DTE}}_{i,j}\left(t\right)\right)\\ u_3={\mathrm{\μ}}_3\left({\mathrm{DTE}}_{i,j}\left(t\right)\right)\end{array}\right.$ $\left\{\begin{array}{c}{u}_1={\mathrm{\μ}}_1\left({\mathrm{ITE}}_{i,j}\left(t\right)\right)\\ u_2={\mathrm{\μ}}_2\left({\mathrm{ITE}}_{i,j}\left(t\right)\right)\\ u_3={\mathrm{\μ}}_3\left({\mathrm{ITE}}_{i,j}\left(t\right)\right)\end{array}\right.---\left(11\right)$

Node is trusted the membership function of classification and regard proposition { T}{-T}{T as, the basic confidence level function of-T}, u1 then, u2 and u3 have represented respectively the trust evidence object of assessment j have been " insincere ", the degree of support of " uncertain " and " definitely credible ", this moment is trust vector DTE directly _{I, j}(t) the component m in _Ij ^D(T}), m _Ij ^D(T ,-T}) and m _Ij ^D({ T}) equal respectively u1, u2, u3; In like manner, i is to the indirect trust vector IT of j _IjComponent m _Ij ^k(T}), m _Ij ^k(T ,-T}) and m _Ij ^k({ T}) also correspond respectively to u1, u2, u3.

Step 4: calculate the evidence difference between each indirect trust values of evaluated node and the direct trust value, revise the weight of indirect trust values; Concrete grammar is:

According to the human behavior model and trust subjectivity as can be known, direct trust value should be not identical with the indirect trust values weight, so need DTE in the D-S building-up process _j(t) and ITE _j(t) carry out the weight correction.Suppose that the direct trust value weight is 1, according to the similarity of indirect trust values and direct trust value, the weight of corresponding adjustment circumstantial evidence obtains main body i to the comprehensive trust value AT of object j _{I, j}(t); Suppose that i receives the indirect trust values of n assessment j, indirectly trust with the distance of directly trusting to be calculated as:

${\mathrm{ITE}}_{i,j}^{k_1}\left(t\right)=(m_{i,j}^{k_1}\left(\right\{T\left\}\right),m_{i,j}^{k_1}\left(\right\{-T\left\}\right),m_{i,j}^{k_1}\left(\right\{T,-T\left\}\right))$

${\mathrm{ITE}}_{i,j}^{k_2}\left(t\right)=(m_{i,j}^{k_2}\left(\right\{T\left\}\right),m_{i,j}^{k_2}\left(\right\{-T\left\}\right),m_{i,j}^{k_2}\left(\right\{T,-T\left\}\right))---\left(12\right)$

$......$

${\mathrm{ITE}}_{i,j}^{k_n}\left(t\right)=(m_{i,j}^{k_n}\left(\right\{T\left\}\right),m_{i,j}^{k_n}\left(\right\{-T\left\}\right),m_{i,j}^{k_n}\left(\right\{T,-T\left\}\right))$

According to the distance of each circumstantial evidence and positive evidence, calculate the difference D of any circumstantial evidence and positive evidence _{Kn, i}(t):

$D_{k_n,j}\left(t\right)=\sqrt{\frac{1}{2}{(\stackrel{\&RightArrow;}{m_{i,j}^{k_n}}-\stackrel{\&RightArrow;}{m_{i,j}^D})}^{T}D(\stackrel{\&RightArrow;}{m_{i,j}^{k_n}}-\stackrel{\&RightArrow;}{m_{i,j}^D})}=\sqrt{\frac{1}{2}({\left|\right|m_{i,j}^{k_n}\left|\right|}^2+{\left|\right|m_{i,j}^D\left|\right|}^2-2<\stackrel{\&RightArrow;}{m_{i,j}^{k_n}},\stackrel{\&RightArrow;}{m_{i,j}^D}>)}---\left(13\right)$

The following formula symbol description is as follows: DTE _{I, j}(t) be direct trust vector, ITE _{I, j} ^k(t) be indirect trust vector, AT _{I, j}(t) be synthetic trust vector; M ({ T}), m (T}), m (T ,-T}) correspondingly respectively determine credible component, insincere component, uncertain component.

As evidence difference D _{Kn, i}(t)＞and during ξ, adjust the coefficient of similarity between evidence, coefficient is:

$S_{k_n,i}\left(t\right)=1-D_{k_n,i}\left(t\right)---\left(14\right)$

Therefore, the correction weight of each evidence is:

$\left\{\begin{array}{c}{\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{T\left\}\right)=S_{k_n,i}\left(t\right)*m_{i,j}^{k_n}\left(\right\{T\left\}\right)\\ \mathrm{\&Delta;}{m}_{i,j}^{k_n}\left(\right\{-T\left\}\right){=S}_{k_n,i}\\ {\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{T,-T\left\}\right)=1-m_{i,j}^{k_n}\left(\right\{T\left\}\right)-m_{i,j}^{k_n}\left(\right\{-T\left\}\right)\end{array}\right.---\left(15\right)$

Circumstantial evidence is adjusted into:

$\mathrm{\&Delta;}{\mathrm{ITE}}_{i,j}^{k_n}\left(t\right)=({\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{T\left\}\right),{\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{-T\left\}\right),{\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{T,-T\left\}\right))---\left(16\right)$

Step 5: according to trust weight after revising, adopt the Dempster composition rule, obtain the comprehensive trust value of evaluated node; And obtain final basic confidence level assigned value.Concrete grammar is:

Dempster composition rule: suppose bel ₁And bel ₂Be two belief functions on the same identification framework Ω, the basic reliability function of its correspondence is respectively m ₁And m ₂, suppose

Bel then ₁And bel ₂Quadrature and bel=bel ₁θ bel ₂A ∈ Ω wherein, the basic reliability function m of bel is as follows:

Can get main body i according to formula (16) to the comprehensive trust value of object j is:

$m_{i,j}\left(A\right)=m_{i,j}^D\left(A\right)\&CirclePlus;\underset{n=0}{\overset{l}{\mathrm{\&Sigma;}}}{\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(A\right)---\left(19\right)$

At last, obtain final basic confidence level assigned value by formula (19), if decision-making is satisfied

$\left\{\begin{array}{c}{m}_{i,j}\left(\right\{T\left\}\right)-m_{i,j}\left(\right\{-T\left\}\right)>\mathrm{\&epsiv;}\\ m_{i,j}\left(\right\{T,-T\left\}\right)<\mathrm{\&theta;}\\ m_{i,j}\left(\right\{T\left\}\right)>m_{i,j}\left(T\right\{T,-T\left\}\right)\end{array}\right.---\left(20\right)$

Then main body i thinks object j " credible ", and mark j is trusted node in the trust table of node i; In like manner, can be labeled as " insincere " or " uncertain ".

3, advantage and effect:

(1) utilizes the behavior of observing each other the other side between neighbor node, formulate various trust-factor, in conjunction with network security rank and context temporal correlation, obtain trust value each other between adjacent node; Better embody time, spatial coherence between nodes ' behavior than conventional method.

(2) be under the jurisdiction of the degree of fuzzy set based on strict trust value, demonstrate fully subjectivity and the uncertainty of trusting between node, have higher adaptive ability than conventional method.

(3) improve D-S evidence theory rule of combination, according to each trust vector weight of difference correction between evidence, demonstrate fully direct trust and the percentage contribution of indirectly trusting comprehensive trust value, have higher accuracy than conventional method.

(4) for different network environments (military affairs, the disaster relief, civilian etc.), formulate different coefficient of safetys, have stronger versatility.

(4) description of drawings

Fig. 1 is the flow chart that the present invention is based on the sensor network trust evaluating method of nodes ' behavior and D-S evidence theory;

Fig. 2 is fuzzy set membership function schematic diagram among the present invention;

Fig. 3 assesses recommendation trust relation principle figure between node among the present invention;

Fig. 4 is the concrete example integral experiment schematic diagram of using among the present invention;

Fig. 5 is the concrete example node schematic diagram of using among the present invention;

Fig. 6 is the concrete example gateway schematic diagram of using among the present invention;

Fig. 7 is the concrete example monitor supervision platform program flow diagram of using among the present invention;

(5) embodiment

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

Sensor network trust evaluating method based on nodes ' behavior and D-S evidence theory provided by the invention is applicable to have the sensor network system of self-organizing feature; For fear of trusting the Infinite Cyclic state that enters that calculates between node, assessment node (main body) only calculates the trust value of evaluated node (object) in the double bounce; At first, adopt the statistic of unit interval t interior nodes behavior, determine various trust-factor, obtain respectively the direct and indirect trust values of evaluated node; Then, according to directly with the degree of membership relation of indirect trust values to fuzzy set, the basic input vector of formation evidence theory; At last, according to the evidence difference of each indirect trust values and direct trust value, revise the Damper rule of combination, synthesize the comprehensive trust value of evaluated node, reach the purpose of malicious node in the recognition network.

The present invention is based on the sensor network trust evaluating method of nodes ' behavior and D-S evidence theory, and the assessment node is realized by following steps to the trust evaluation flow process of evaluated node as shown in Figure 1:

Step 1: each trust-factor of node is generated strategy in the wireless sensor network, and concrete grammar is:

For the sensor node characteristic, the various Components of defined node degree of belief are without loss of generality, and suppose that node i (assessment node) is to node j (evaluated node) trust evaluation.

(1) packet receiving rate factor R PF _{I, j}(t): node j whenever receives a packet, needs to send the ACK feedback information and determines; Node i listens to the quantity of the ack msg bag that j sends, thereby obtains the packet receiving rate information of j; According to the node j packet receiving rate situation of change in time period t and the t-1, whether node j exists the personation replay attacks as can be known.If the variation of j packet receiving rate remains in (ξ, ξ), think that the j node is working properly; Its computing formula is:

${\mathrm{RPF}}_{i,j}\left(t\right)=\frac{{\mathrm{RP}}_{\mathrm{ij}}\left(t\right)-{\mathrm{RP}}_{\mathrm{ij}}(t-1)}{({\mathrm{RP}}_{\mathrm{ij}}\left(t\right)+{\mathrm{RP}}_{\mathrm{ij}}(t-1))/2}---\left(1\right)$

Wherein, RP _{I, j}(t) be packet receiving quantity;

(2) send success rate factor S PF _{I, j}(t): each packet that node sends stabs if having time.Suppose that j sends packet to node k, k is outside the communication range of i, and this moment, i can't monitor the transmission success rate of node j.Therefore, node i adopts and monitors the transmission success rate that j transmission identical data packet number of times is determined j.Different according to node time stamp, even packet content is identical, effective distinguishes data bag is still determined the transmission times of certain packet; Its computing formula is

${\mathrm{SPF}}_{i,j}\left(t\right)=\frac{{\mathrm{SP}}_{\mathrm{ij}}\left(t\right)-{\mathrm{SF}}_{\mathrm{ij}}\left(t\right)}{{\mathrm{SP}}_{\mathrm{ij}}\left(t\right)+{\mathrm{SF}}_{\mathrm{ij}}\left(t\right)}---\left(2\right)$

Wherein, SF _{I, j}(t) be the quantity of giving out a contract for a project;

(3) forward rate factor TPF _{I, j}(t): because WSNs self, most nodes can't with base station (BS) direct communication, therefore need next-hop node to carry out data retransmission.Suppose that node k sends packet to node j, k is outside the communication range of i, and this moment, i can't monitor the packet receiving quantity of node j, so node i adopts and monitor j and send the ACK feedback information and determine j packet receiving quantity, and the monitoring node j situation of giving out a contract for a project is determined its quantity of giving out a contract for a project.According to its rate of change situation, effectively avoid cesspool to attack and Sybil attack, and effectively whether recognition node is selfish node; Its computing formula is:

${\mathrm{TPF}}_{i,j}\left(t\right)=\frac{{\mathrm{FP}}_{\mathrm{ij}}\left(t\right)-{\mathrm{FP}}_{\mathrm{ij}}(t-1)}{({\mathrm{FP}}_{\mathrm{ij}}\left(t\right)+{\mathrm{FP}}_{\mathrm{ij}}(t-1))/2}---\left(3\right)$

Wherein, FP _{I, j}(t) for transmitting bag quantity;

(4) consistency factor CPF _{I, j}(t): because the node data bag has spatial coherence, packet is more similar between the localized network neighbor node, therefore introduces the consistency factor, avoids malicious node that packet is distorted.Node i is obtained the forwarding data bag of j at random, compares with its data bag, if i, the j variance rate remains in (ξ, ξ), thinks that the j node is working properly; Its computing formula is:

${\mathrm{CPF}}_{i,j}\left(t\right)=\frac{{\mathrm{NEP}}_{\mathrm{ij}}\left(t\right)}{{\mathrm{EP}}_{\mathrm{ij}}\left(t\right)+{\mathrm{NEP}}_{\mathrm{ij}}\left(t\right)}---\left(4\right)$

Wherein, EP _{I, j}(t) be consistent bag quantity, NEP _{I, j}(t) be inconsistent bag quantity;

(5) time granularity factor TFF (t): because trust value has time, content context relation, the node trust value is to change at a upper time segment base plinth, therefore must the joining day factor be analyzed by trust value.The time granularity size is analyzed as the case may be; If granularity is excessive, then comprehensive trust value is subjected to historical influence excessive, and assessment may make a mistake to node; If undersized, then trust value relies on excessive to the single time period; Therefore, in the situation that level of security is higher, TFF (t)=0.75, in the lower situation of level of security, trust-factor is TFF (t)=0.25.Generally speaking, TFF (t)=0.5

(6) availability factor factor HPF _{I, j}(t): because there is unreachable situation in the factors such as channel, environment between neighbors; Only have between neighbors and can communicate and monitor, above various trust-factor are just meaningful, therefore introduce the availability factor.In the time period t, node i sends at random m HELLO bag and surveys, if receive that j to the feedback information ACK-HELLO packet of HELLO, then thinks can reach between node; Its computing formula is:

${\mathrm{HPF}}_{i,j}\left(t\right)=\frac{{\mathrm{NACK}}_{H_{\mathrm{ij}}}\left(t\right)}{{\mathrm{ACK}}_{H_{\mathrm{ij}}}\left(t\right)+\mathrm{NAC}{K}_{H_{\mathrm{ij}}}\left(t\right)}---\left(5\right)$

Wherein, ACK _{I, j}(t) for answering bag quantity, NACK _{I, j}(t) for not answering bag quantity;

(7) level of security SG:WSNs is in different applied environments and scene, and required level of security is different; The network security requirement difference that for example is applied to battlefield and environmental monitoring is huge.Therefore, SG adopts exponential manner to calculate; When safety requirements is higher, SG=3; When low, SG=1; When general, SG=2.

Step 2: according to the network application scene, the trust-factor weight is set, simultaneously computing node behavior coefficient μ obtains direct trust value and a plurality of indirect trust values of evaluated object, and concrete grammar is:

(1) level of security is relevant with network application environment, when safety requirements is higher, and level of security SG=3; When low, SG=1; When general, SG=2.In like manner, the time granularity size when safety requirements is higher, time rate factor TFF (t)=0.75; When low, TFF (t)=0.25; When general, TFF (t)=0.5.Other trust-factor generally speaking coefficient equates, and summation is 1.

(2) in order to guarantee the reasonability of trust evaluation, need to formulate the behavior coefficient μ in the different time sections.For example: in time period t-1, i and j mutual 10000 times altogether, wherein 5000 satisfaction; In the time period t, i and j mutual 1000 times altogether, wherein 500 satisfaction; Satisfied behavior ratio is 1/2, and obviously the trust value of time period t-1 is more credible; Therefore, the behavior coefficient formula is:

$\mathrm{\&mu;}=\frac{S_{\mathrm{ij}}\left(t\right)/(F_{\mathrm{ij}}\left(t\right)+S_{\mathrm{ij}}\left(t\right))}{S_{\mathrm{ij}}(t-1)/(F_{\mathrm{ij}}(t-1)+S_{\mathrm{ij}}(t-1))}---\left(6\right)$

Wherein, S _Ij(t) be number of success, F _Ij(t) be the frequency of failure; S _Ij, F _IjBe respectively average successful, the miss data bag of each time period in the network operation process.

(3) according to each trust-factor in the step 1, in conjunction with node resource and network operation actual environment, the COMPREHENSIVE CALCULATING node i is to the direct trust value DTE of node j _{I, j}(t) be:

DTE _i，j(t)＝μ ^SG*TFF(t)*(1-w ₁*RPF _i，j(t)-w ₂*SPF _i，j(t)-w ₃*TPF _i，j(t)-w ₄*CPF _i，j(t)-w ₅*HPF _i，j(t))

+(1-TFF(t))*DTE _i，j(t-1) (7)

Wherein, w ₁, w ₂, w ₃, w ₄, w ₅Be each trust-factor coefficient, adjustable according to concrete application; RPF _{I, j}(t) be the packet receiving rate factor, SPF _{I, j}(t) be the success rate factor of giving out a contract for a project, TPF _{I, j}(t) be the forward rate factor, CPF _{I, j}(t) be the consistency factor, HPF _{I, j}(t) be the availability factor factor; And w ₁+ w ₂+ w ₃+ w ₄+ w ₅=1 (8)

Assessment main body i collects other nodes to the indirect recommendation of trust value conduct of evaluated object j.In order to reduce network energy consumption and the communication load of WSNs, and prevent from trusting circular recursion, recommendation is only limited to the common neighbors of i, j to the direct trust value of j.Suppose that k is i, the common neighbours of j transmit the decline principle according to trusting, and recommendation trust recommends neighborhood to be illustrated in figure 3 as: ITE as shown in Figure 2 _{I, j}(t)=DTE _{I, k}(t) * DTE _{K, j}(t) (9)

Step 3: utilize the degree of membership of fuzzy set theory and the concept of linguistic variable, calculate fuzzy subset's membership function of various trust values, various trust values are carried out fuzzy classification, form the basic confidence level function of D-S evidence theory; Fuzzy classification as shown in Figure 2.Concrete grammar is:

The fuzzy classification of (1) node being trusted: at first node is divided into " insincere " " uncertain " and " definitely credible " three level of trusts.Secondly, corresponding these three level of trusts, domain [0 at node trust value T, 1] upper fuzzy subset T1, T2 and the T3 of dividing, set up membership function μ 1 (t), μ 2 (t) and μ 3 (t), wherein, membership function μ 1 (t), μ 2 (t) and μ 3 (T) need satisfy μ 1 (t)+μ 2 (t)+μ 3 (t)=1.

(2) according to fuzzy membership function, node i is expressed as vector form to the trusting degree of node j, and direct, indirect, comprehensive trust vector is respectively:

${\mathrm{DTE}}_{\mathrm{ij}}\left(t\right)=(m_{\mathrm{ij}}^D\left(\right\{T\left\}\right),m_{i,j}^D\left(\right\{T\left\}\right),m_{i,j}^D\left(\right\{T,-T\left\}\right))$

${\mathrm{ITE}}_{i,j}^k\left(t\right)=(m_{i,j}^k\left(\right\{T\left\}\right),m_{i,j}^k\left(\right\{-T\left\}\right),m_{i,j}^k\left(\right\{T,-T\left\}\right))---\left(10\right)$

AT _i，j(t)＝(m _i，j({T})，m _i，j({-T})，m _i，j({T，-T}))

The following formula symbol description is as follows: DTE _{I, j}(t) be direct trust vector, ITE _{I, j} ^k(t) be indirect trust vector, AT _{I, j}(t) be synthetic trust vector; M ({ T}), m (T}), m (T ,-T}) correspondingly respectively determine credible component, insincere component, uncertain component.

The membership function that utilizes node to trust classification calculates object of assessment j to " insincere ", and the degree of membership of " uncertain " and " definitely credible " three level of trusts is u1, u2 and u3:

$\left\{\begin{array}{c}{u}_1={\mathrm{\&mu;}}_1\left({\mathrm{DTE}}_{i,j}\left(t\right)\right)\\ u_2={\mathrm{\&mu;}}_2\left({\mathrm{DTE}}_{i,j}\left(t\right)\right)\\ u_3={\mathrm{\&mu;}}_3\left({\mathrm{DTE}}_{i,j}\left(t\right)\right)\end{array}\right.$ $\left\{\begin{array}{c}{u}_1={\mathrm{\&mu;}}_1\left({\mathrm{ITE}}_{i,j}\left(t\right)\right)\\ u_2={\mathrm{\&mu;}}_2\left({\mathrm{ITE}}_{i,j}\left(t\right)\right)\\ u_3={\mathrm{\&mu;}}_3\left({\mathrm{ITE}}_{i,j}\left(t\right)\right)\end{array}\right.---\left(11\right)$

Node is trusted the membership function of classification and regard proposition { T}{-T}{T as, the basic confidence level function of-T}, u1 then, u2 and u3 have represented respectively the trust evidence object of assessment j have been " insincere ", the degree of support of " uncertain " and " definitely credible ", this moment is trust vector DTE directly _{I, j}(t) the component m in _Ij ^D(T}), m _Ij ^D(T ,-T}) and m _Ij ^D({ T}) equal respectively u1, u2, u3; In like manner, i is to the indirect trust vector IT of j _IjComponent m _Ij ^k(T}), m _Ij ^k(T ,-T}) and m _Ij ^k({ T}) also correspond respectively to u1, u2, u3.

Step 4: calculate the evidence difference between each indirect trust values of evaluated node and the direct trust value, revise the weight of indirect trust values; Concrete grammar is:

According to the human behavior model and trust subjectivity as can be known, direct trust value should be not identical with the indirect trust values weight, so need DTE in the D-S building-up process _j(t) and ITE _j(t) carry out the weight correction.Suppose that the direct trust value weight is 1, according to the similarity of indirect trust values and direct trust value, the weight of corresponding adjustment circumstantial evidence obtains main body i to the comprehensive trust value AT of object j _{I, j}(t); Suppose that i receives the indirect trust values of n assessment j, indirectly trust with the distance of directly trusting to be calculated as:

${\mathrm{ITE}}_{i,j}^{k_1}\left(t\right)=(m_{i,j}^{k_1}\left(\right\{T\left\}\right),m_{i,j}^{k_1}\left(\right\{-T\left\}\right),m_{i,j}^{k_1}\left(\right\{T,-T\left\}\right))$

${\mathrm{ITE}}_{i,j}^{k_2}\left(t\right)=(m_{i,j}^{k_2}\left(\right\{T\left\}\right),m_{i,j}^{k_2}\left(\right\{-T\left\}\right),m_{i,j}^{k_2}\left(\right\{T,-T\left\}\right))---\left(12\right)$

$......$

${\mathrm{ITE}}_{i,j}^{k_n}\left(t\right)=(m_{i,j}^{k_n}\left(\right\{T\left\}\right),m_{i,j}^{k_n}\left(\right\{-T\left\}\right),m_{i,j}^{k_n}\left(\right\{T,-T\left\}\right))$

According to the distance of each circumstantial evidence and positive evidence, calculate the difference D of any circumstantial evidence and positive evidence _{Kn, i}(t):

$D_{k_n,j}\left(t\right)=\sqrt{\frac{1}{2}{(\stackrel{\&RightArrow;}{m_{i,j}^{k_n}}-\stackrel{\&RightArrow;}{m_{i,j}^D})}^{T}D(\stackrel{\&RightArrow;}{m_{i,j}^{k_n}}-\stackrel{\&RightArrow;}{m_{i,j}^D})}=\sqrt{\frac{1}{2}({\left|\right|m_{i,j}^{k_n}\left|\right|}^2+{\left|\right|m_{i,j}^D\left|\right|}^2-2<\stackrel{\&RightArrow;}{m_{i,j}^{k_n}},\stackrel{\&RightArrow;}{m_{i,j}^D}>)}---\left(13\right)$

The following formula symbol description is as follows: DTE _{I, j}(t) be direct trust vector, ITE _{I, j} ^k(t) be indirect trust vector, AT _{I, j}(t) be synthetic trust vector; M ({ T}), m (T}), m (T ,-T}) correspondingly respectively determine credible component, insincere component, uncertain component.

As evidence difference D _{Kn, i}(t)＞and during ξ, adjust the coefficient of similarity between evidence, coefficient is:

$S_{k_n,i}\left(t\right)=1-D_{k_n,i}\left(t\right)---\left(14\right)$

Therefore, the correction weight of each evidence is:

$\left\{\begin{array}{c}{\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{T\left\}\right)=S_{k_n,i}\left(t\right)*m_{i,j}^{k_n}\left(\right\{T\left\}\right)\\ \mathrm{\&Delta;}{m}_{i,j}^{k_n}\left(\right\{-T\left\}\right){=S}_{k_n,i}\\ {\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{T,-T\left\}\right)=1-m_{i,j}^{k_n}\left(\right\{T\left\}\right)-m_{i,j}^{k_n}\left(\right\{-T\left\}\right)\end{array}\right.---\left(15\right)$

Circumstantial evidence is adjusted into:

$\mathrm{\&Delta;}{\mathrm{ITE}}_{i,j}^{k_n}\left(t\right)=({\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{T\left\}\right),{\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{-T\left\}\right),{\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{T,-T\left\}\right))---\left(16\right)$

Step 5: according to trust weight after revising, adopt the Dempster composition rule, obtain the comprehensive trust value of evaluated node; And obtain final basic confidence level assigned value.Concrete grammar is:

Dempster composition rule: suppose bel ₁And bel ₂Be two belief functions on the same identification framework Ω, the basic reliability function of its correspondence is respectively m ₁And m ₂, suppose Bel then ₁And bel ₂Quadrature and bel=bel ₁θ bel ₂A ∈ Ω wherein, the basic reliability function m of bel is as follows:

Can get main body i according to formula (16) to the comprehensive trust value of object j is:

$m_{i,j}\left(A\right)=m_{i,j}^D\left(A\right)\&CirclePlus;\underset{n=0}{\overset{l}{\mathrm{\&Sigma;}}}{\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(A\right)---\left(19\right)$

At last, obtain final basic confidence level assigned value by formula (19), if decision-making is satisfied

$\left\{\begin{array}{c}{m}_{i,j}\left(\right\{T\left\}\right)-m_{i,j}\left(\right\{-T\left\}\right)>\mathrm{\&epsiv;}\\ m_{i,j}\left(\right\{T,-T\left\}\right)<\mathrm{\&theta;}\\ m_{i,j}\left(\right\{T\left\}\right)>m_{i,j}\left(T\right\{T,-T\left\}\right)\end{array}\right.---\left(20\right)$

Then main body i thinks object j " credible ", and mark j is trusted node in the trust table of node i; In like manner, can be labeled as " insincere " or " uncertain ".

The present invention describes in conjunction with concrete example:

Utilize wireless sensor network testing platform that it has been carried out the Physical Experiment checking, experimental principle as shown in Figure 4: a plurality of real sensor nodes are arranged to live network, and the operation method for evaluating trust that the present invention carries.Each node and its neighbor node in the net carry out trust evaluation mutually; Trust evaluation result and topology information in sensor node is periodically compiled bunch, and to gateway (Sink node) report; Gateway is collected, trust information and the topology information of arrangement network, and reports to the PC terminal by serial ports; The PC terminal utilizes serial ports to receive the data that gateway sends, and shows the Evaluated effect of node trust and the safe condition of program operation by application programming interfaces.Node procedure is write based on TinyOS, and the design of node schematic diagram is Fig. 5; Gateway program is write based on μ cLinux, and the gateway designs schematic diagram is Fig. 6; Monitor supervision platform is write based on VS2008 on PC, and program flow diagram is Fig. 7, and experimental result is as shown in table 1.

Table 1

In sum, the present invention proposes a kind of based on node confidence in conjunction with the method for evaluating trust that improves evidence theory, emphasized ambiguity, subjectivity and the availability of trusting; Analyze the network actual environment, formulated the various trust-factor of node and carry out quantitative, qualitative analysis, determined direct between neighbor node and indirect trust values; Utilize Fuzzy set theory, the value of establishing trust is subordinated to the degree of trusting classification, forms the basic input vector of evidence theory; Improve the Dempster Evidence, according to the evidence difference degree, adjust the coefficient of similarity between evidence, finally the trust value of synthetic evaluated node.The method has stronger versatility, embodies the characteristic of trust value " obtain difficulty, lose easily ", and has 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. sensor network trust evaluating method based on nodes ' behavior and D-S evidence theory, it is characterized in that: it utilizes the behavior of neighbor node, formulates various trust-factor, according to network application scene difference each factor coefficient is set; And in conjunction with improved D-S Evidence Combination, obtain the comprehensive trust value of node, the method concrete steps are as follows:

Step 1: each trust-factor of node is generated strategy in the wireless sensor network, and concrete grammar is:

For the sensor node characteristic, the various Components of defined node degree of belief, establishing node i, namely to assess node be evaluated node trust evaluation to node j;

(1) packet receiving rate factor R PF _{I, j}(t): node j whenever receives a packet, needs to send the ACK feedback information and determines; Node i listens to the quantity of the ack msg bag that j sends, thereby obtains the packet receiving rate information of j; According to the node j packet receiving rate situation of change in time period t and the t-1, whether node j exists the personation replay attacks as can be known; If the variation of j packet receiving rate remains in (ξ, ξ), think that the j node is working properly; Its computing formula is:

${\mathrm{RPF}}_{i,j}\left(t\right)=\frac{{\mathrm{RP}}_{\mathrm{ij}}\left(t\right)-{\mathrm{RP}}_{\mathrm{ij}}(t-1)}{({\mathrm{RP}}_{\mathrm{ij}}\left(t\right)+{\mathrm{RP}}_{\mathrm{ij}}(t-1))/2}---\left(1\right)$

Wherein, RP _Ij(t) be packet receiving quantity;

(2) send success rate factor S PF _{I, j}(t): each packet that node sends stabs if having time, establishes j and sends packet to node k, and k is outside the communication range of i, and this moment, i can't monitor the transmission success rate of node j; Therefore, node i adopts and monitors the transmission success rate that j transmission identical data packet number of times is determined j; Different according to node time stamp, even packet content is identical, effective distinguishes data bag is still determined the transmission times of certain packet; Its computing formula is:

${\mathrm{SPF}}_{i,j}\left(t\right)=\frac{{\mathrm{SP}}_{\mathrm{ij}}\left(t\right)-{\mathrm{SF}}_{\mathrm{ij}}\left(t\right)}{{\mathrm{SP}}_{\mathrm{ij}}\left(t\right)+{\mathrm{SF}}_{\mathrm{ij}}\left(t\right)}---\left(2\right)$

Wherein, SP _Ij(t) quantity, the SF for giving out a contract for a project _Ij(t) the successful quantity that repeats to give out a contract for a project;

(3) forward rate factor TPF _{I, j}(t): because WSNs self, most nodes can't with base station (BS) direct communication, therefore need next-hop node to carry out data retransmission; If node k sends packet to node j, k is outside the communication range of i, and this moment, i can't monitor the packet receiving quantity of node j, so node i adopts and monitor j and send the ACK feedback information and determine j packet receiving quantity, and the monitoring node j situation of giving out a contract for a project is determined its quantity of giving out a contract for a project; According to its rate of change situation, effectively avoid cesspool to attack and Sybil attack, and effectively whether recognition node is selfish node; Its computing formula is:

${\mathrm{TPF}}_{i,j}\left(t\right)=\frac{{\mathrm{FP}}_{\mathrm{ij}}\left(t\right)-{\mathrm{FP}}_{\mathrm{ij}}(t-1)}{({\mathrm{FP}}_{\mathrm{ij}}\left(t\right)+{\mathrm{FP}}_{\mathrm{ij}}(t-1))/2}---\left(3\right)$

Wherein, FP _Ij(t) for transmitting bag quantity;

(4) consistency factor CPF _{I, j}(t): because the node data bag has spatial coherence, packet is more similar between the localized network neighbor node, therefore introduces the consistency factor, avoids malicious node that packet is distorted; Node i is obtained the forwarding data bag of j at random, compares with its data bag, if i, the j variance rate remains in (ξ, ξ), thinks that the j node is working properly; Its computing formula is:

${\mathrm{CPF}}_{i,j}\left(t\right)=\frac{{\mathrm{NEP}}_{\mathrm{ij}}\left(t\right)}{{\mathrm{EP}}_{\mathrm{ij}}\left(t\right)+{\mathrm{NEP}}_{\mathrm{ij}}\left(t\right)}---\left(4\right)$

Wherein, EP _Ij(t) be consistent bag quantity, NEP _Ij(t) be inconsistent bag quantity;

(5) time granularity factor TFF (t): because trust value has time, content context relation, the node trust value is to change at a upper time segment base plinth, therefore must the joining day factor be analyzed by trust value; The time granularity size is analyzed as the case may be; If granularity is excessive, then comprehensive trust value is subjected to historical influence excessive, and assessment may make a mistake to node; If undersized, then trust value relies on excessive to the single time period; Therefore, in the situation that level of security is higher, TFF (t)=0.75, in the lower situation of level of security, trust-factor is TFF (t)=0.25; TFF (t)=0.5;

(6) availability factor factor HPF _{I, j}(t): because there is unreachable situation in the factors such as channel, environment between neighbors; Only have between neighbors and can communicate and monitor, above various trust-factor are just meaningful, therefore introduce the availability factor; In the time period t, node i sends at random m HELLO bag and surveys, if receive that j to the feedback information ACK-HELLO packet of HELLO, then thinks can reach between node; Its computing formula is:

${\mathrm{HPF}}_{i,j}\left(t\right)=\frac{{\mathrm{NACK}}_{H_{\mathrm{ij}}}\left(t\right)}{{\mathrm{ACK}}_{H_{\mathrm{ij}}}\left(t\right)+{\mathrm{NACK}}_{H_{\mathrm{ij}}}\left(t\right)}---\left(5\right)$

Wherein,

Wrap quantity for answering,

For not answering bag quantity;

(7) level of security SG:WSNs is in different applied environments and scene, and required level of security is different; Therefore, SG adopts exponential manner to calculate; When safety requirements is higher, SG=3; When low, SG=1; SG=2;

Step 2: according to the network application scene, the trust-factor weight is set, simultaneously computing node behavior coefficient μ obtains direct trust value and a plurality of indirect trust values of evaluated object, and concrete grammar is:

(1) level of security is relevant with network application environment, when safety requirements is higher, and level of security SG=3; When low, SG=1; SG=2; In like manner, the time granularity size when safety requirements is higher, time rate factor TFF (t)=0.75; When low, TFF (t)=0.25; TFF (t)=0.5; Other trust-factor coefficient equates, and summation is 1;

(2) in order to guarantee the reasonability of trust evaluation, need to formulate the behavior coefficient μ in the different time sections; The behavior coefficient formula is:

$\mathrm{\&mu;}=\frac{S_{\mathrm{ij}}\left(t\right)/(F_{\mathrm{ij}}\left(t\right)+S_{\mathrm{ij}}\left(t\right))}{S_{\mathrm{ij}}(t-1)/(F_{\mathrm{ij}}(t-1)+S_{\mathrm{ij}}(t-1))}---\left(6\right)$

Wherein, S _Ij(t) be number of success, F _Ij(t) be the frequency of failure; S _Ij, F _IjBe respectively average successful, the miss data bag of each time period in the network operation process;

(3) according to each trust-factor in the step 1, in conjunction with node resource and network operation actual environment, the COMPREHENSIVE CALCULATING node i is to the direct trust value DTE of node j _{I, j}(t) be:

DTE _i,j(t)=μ ^SG*TFF(t)*(1-w ₁*RPF _i,j(t)-w ₂*SPF _i,j(t)-w ₃*TPF _i,j(t)-w ₄*CPF _i,j(t)-w ₅*HPF _i,j(t))

+(1-TFF(t))*DTE _i,j(t-1) (7)

Wherein, w ₁, w ₂, w ₃, w ₄, w ₅Be each trust-factor coefficient, adjustable according to concrete application; RPF _{I, j}(t) be the packet receiving rate factor, SPF _{I, j}(t) be the success rate factor of giving out a contract for a project, TPF _{I, j}(t) be the forward rate factor, CPF _{I, j}(t) be the consistency factor, HPF _{I, j}(t) be the availability factor factor; And w ₁+ w ₂+ w ₃+ w ₄+ w ₅=1 (8)

Assessment main body i collects other nodes to the indirect recommendation of trust value conduct of evaluated object j; In order to reduce network energy consumption and the communication load of WSNs, and prevent from trusting circular recursion, recommendation is only limited to the common neighbors of i, j to the direct trust value of j; If k is i, the common neighbours of j transmit decline principle, ITE according to trusting _{I, j}(t)=DTE _{I, k}(t) * DTE _{K, j}(t) (9)

Step 3: utilize the degree of membership of fuzzy set theory and the concept of linguistic variable, calculate fuzzy subset's membership function of various trust values, various trust values are carried out fuzzy classification, form the basic confidence level function of D-S evidence theory; Concrete grammar is:

The fuzzy classification of (1) node being trusted: at first node is divided into " insincere " " uncertain " and " definitely credible " three level of trusts; Secondly, corresponding these three level of trusts, domain [0 at node trust value T, 1] upper fuzzy subset T1, T2 and the T3 of dividing, set up membership function μ 1 (t), μ 2 (t) and μ 3 (t), wherein, membership function μ 1 (t), μ 2 (t) and μ 3 (t) need satisfy μ 1 (t)+μ 2 (t)+μ 3 (t)=1;

(2) according to fuzzy membership function, node i is expressed as vector form to the trusting degree of node j, and direct, indirect, comprehensive trust vector is respectively:

${\mathrm{DTE}}_{i,j}\left(t\right)=(m_{i,j}^D\left(\right\{T\left\}\right),m_{i,j}^D\left(\right\{-T\left\}\right),m_{i,j}^D\left(\right\{T,-T\left\}\right))$

${\mathrm{ITE}}_{i,j}^k\left(t\right)=(m_{i,j}^k\left(\right\{T\left\}\right),m_{i,j}^k\left(\right\{-T\left\}\right),m_{i,j}^k\left(\right\{T,-T\left\}\right))---\left(10\right)$

AT _i,j(t)=(m _i,j({T}),m _i,j({-T}),m _i,j({T,-T}))

The following formula symbol description is as follows: DTE _{I, j}(t) be direct trust vector,

Be indirect trust vector, AT _{I, j}(t) be synthetic trust vector; M ({ T}), m (T}), m (T ,-T}) correspondingly respectively determine credible component, insincere component, uncertain component;

The membership function that utilizes node to trust classification calculates object of assessment j to " insincere ", and the degree of membership of " uncertain " and " definitely credible " three level of trusts is u1, u2 and u3:

$\left\{\begin{array}{c}{u}_1^D={\mathrm{\&mu;}}_1\left({\mathrm{DTE}}_{i,j}\left(t\right)\right)\\ u_2^D={\mathrm{\&mu;}}_2\left({\mathrm{DTE}}_{i,j}\left(t\right)\right)\\ u_3^D={\mathrm{\&mu;}}_3\left({\mathrm{DTE}}_{i,j}\left(t\right)\right)\end{array}\right.$ $\left\{\begin{array}{c}{u}_1^I={\mathrm{\&mu;}}_1\left({\mathrm{ITE}}_{i,j}\left(t\right)\right)\\ u_2^I={\mathrm{\&mu;}}_2\left({\mathrm{ITE}}_{i,j}\left(t\right)\right)\\ u_3^I={\mathrm{\&mu;}}_3\left({\mathrm{ITE}}_{i,j}\left(t\right)\right)\end{array}\right.---\left(11\right)$

Node is trusted the membership function of classification and regard proposition { T}{-T}{T, the basic confidence level function of-T}, directly trust vector DTE as _{I, j}(t) component in

With Equal respectively u1, u2, u3; In like manner, i is to the indirect trust vector IT of j _IjComponent

With

Also correspond respectively to u1, u2, u3;

Step 4: calculate the evidence difference between each indirect trust values of evaluated node and the direct trust value, revise the weight of indirect trust values; Concrete grammar is:

According to the human behavior model and trust subjectivity as can be known, direct trust value should be not identical with the indirect trust values weight, so need DTE in the D-S building-up process _j(t) and ITE _j(t) carry out the weight correction; If the direct trust value weight is 1, according to the similarity of indirect trust values and direct trust value, the weight of corresponding adjustment circumstantial evidence obtains main body i to the comprehensive trust value AT of object j _{I, j}(t); Suppose that i receives the indirect trust values of n assessment j, indirectly trust with the vector of directly trusting to be respectively:

${\mathrm{ITE}}_{i,j}^{k_1}\left(t\right)=(m_{i,j}^{k_1}\left(\right\{T\left\}\right),m_{i,j}^{k_1}\left(\right\{-T\left\}\right),m_{i,j}^{k_1}\left(\right\{T,-T\left\}\right))$

${\mathrm{ITE}}_{i,j}^{k_2}\left(t\right)=(m_{i,j}^{k_2}\left(\right\{T\left\}\right),m_{i,j}^{k_2}\left(\right\{-T\left\}\right),m_{i,j}^{k_2}\left(\right\{T,-T\left\}\right))---\left(12\right)$

……

${\mathrm{ITE}}_{i,j}^{k_n}\left(t\right)=(m_{i,j}^{k_n}\left(\right\{T\left\}\right),m_{i,j}^{k_n}\left(\right\{-T\left\}\right),m_{i,j}^{k_n}\left(\right\{T,-T\left\}\right))$

According to the distance of each circumstantial evidence and positive evidence, calculate the difference D of any circumstantial evidence and positive evidence _{Kn, i}(t):

$D_{k_n,t}\left(t\right)=\sqrt{\frac{1}{2}{(\stackrel{\&RightArrow;}{m_{i,j}^{k_n}}-\stackrel{\&RightArrow;}{m_{i,j}^D})}^{T}D(\stackrel{\&RightArrow;}{m_{i,j}^{k_n}}-\stackrel{\&RightArrow;}{m_{i,j}^D})}=\sqrt{\frac{1}{2}({\left|\right|m_{i,j}^{k_n}\left|\right|}^2+{\left|\right|m_{i,j}^D\left|\right|}^2-2\&lang;\stackrel{\&RightArrow;}{m_{i,j}^{k_n}},\stackrel{\&RightArrow;}{m_{i,j}^D}\&rang;)}---\left(13\right)$

The following formula symbol description is as follows: DTE _{I, j}(t) be direct trust vector,

Be indirect trust vector, AT _{I, j}(t) be synthetic trust vector; M ({ T}), m (T}), m (T ,-T}) correspondingly respectively determine credible component, insincere component, uncertain component;

As evidence difference D _{Kn, i}(t)〉during ξ, adjust the coefficient of similarity between evidence, coefficient is:

$S_{k_n,i}\left(t\right)=1-D_{k_n,i}\left(t\right)---\left(14\right)$

Therefore, the correction weight of each evidence is:

$\left\{\begin{array}{c}{\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{T\left\}\right)=S_{k_n,i}\left(t\right)*m_{i,j}^{k_n}\left(\right\{T\left\}\right)\\ {\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{-T\left\}\right)=S_{k_n,i}\left(t\right)*m_{i,j}^{k_n}\left(\right\{-T\left\}\right)\\ {\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{T,-T\left\}\right)=1-m_{i,j}^{k_n}\left(\right\{T\left\}\right)-m_{i,j}^{k_n}\left(\right\{-T\left\}\right)\end{array}\right.---\left(15\right)$

Circumstantial evidence is adjusted into:

${\mathrm{\&Delta;ITE}}_{i,j}^{k_n}\left(t\right)=({\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{T\left\}\right),{\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{-T\left\}\right),{\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(\right\{T,-T\left\}\right))---\left(16\right)$

Step 5: according to trust weight after revising, adopt the Dempster composition rule, obtain the comprehensive trust value of evaluated node; And obtain final basic confidence level assigned value; Concrete grammar is:

Dempster composition rule: suppose bel ₁And bel ₂Be two belief functions on the same identification framework Ω, the basic reliability function of its correspondence is respectively m ₁And m ₂, suppose Bel then ₁And bel ₂Quadrature and bel=bel ₁θ bel ₂A ∈ Ω wherein, the basic reliability function m of bel is as follows:

Can get main body i according to formula (16) to the comprehensive trust value of object j is:

$m_{i,j}\left(A\right)=m_{i,j}^D\left(A\right)\&CirclePlus;\underset{n=0}{\overset{l}{\mathrm{\&Sigma;}}}{\mathrm{\&Delta;m}}_{i,j}^{k_n}\left(A\right)---\left(19\right)$

At last, obtain final basic confidence level assigned value by formula (19), if decision-making is satisfied

$\left\{\begin{array}{c}{m}_{i,j}\left(\right\{T\left\}\right)-m_{i,j}\left(\right\{-T\left\}\right)>\mathrm{\&epsiv;}\\ m_{i,j}\left(\right\{T,-T\left\}\right)<\mathrm{\&theta;}\\ m_{i,j}\left(\right\{T\left\}\right)>m_{i,j}\left(\right\{T,-T\left\}\right)\end{array}\right.---\left(20\right)$

Then main body i thinks object j " credible ", and mark j is trusted node in the trust table of node i; In like manner, can be labeled as " insincere " or " uncertain ".

Images