CN107750053A - Based on multifactor wireless sensor network dynamic trust evaluation system and method - Google Patents

Abstract

The invention discloses a multi-factor based wireless sensor network dynamic trust evaluation system and method, which is characterized in that the system includes a trust factor acquisition module (11), a trust measurement module (12), a trust calculation module (13) and a trust Evaluation module (14); step 1, obtain the trust factors that may affect the trust value of nodes, step 2, define the measurement methods of each trust factor, step 3, realize each trust calculation, step 4, compare the target nodes obtained by the trust calculation process The relationship between the comprehensive trust value and the trust threshold can be used to judge whether the target node is trustworthy. Compared with the prior art, the present invention comprehensively considers the behavior of nodes when they transmit sensing data, so as to monitor the behavior of nodes more comprehensively and ensure timely and effective processing of untrustworthy nodes in the network; in addition, direct trust and The indirect trust comprehensive method measures the trust value of nodes and solves the positive effect of the deviation of evaluation results between nodes.

Description

Multi-factor based dynamic trust evaluation system and method for wireless sensor networks

technical field

The invention belongs to the wireless communication network technology in the security field of the Internet of Things, and in particular relates to a multi-factor-based wireless sensor network dynamic trust evaluation system and method.

Background technique

With the popularization of Internet of Things applications, wireless sensor networks, as the most widely used underlying sensing devices of the Internet of Things, have been successfully applied to important fields such as environmental monitoring, space monitoring, space stations, military defense, smart homes, smart cities, and medical monitoring. At the same time, due to the limitation of the sensor node itself and the variability of the application environment, the security problems of wireless sensor networks are becoming more and more serious.

The wireless sensor network is a dynamic network: network nodes can move anywhere; nodes may fail or even fail due to battery energy exhaustion or other reasons, thus exiting the network operation; new nodes are constantly added, etc. The wireless sensor network is also an open network. The sensor nodes are often deployed randomly in an unsupervised and harsh environment, and the nodes are in direct contact with the environment. On the other hand, due to the limited resources of nodes, nodes often need to transmit data through local cooperation, and communicate using wireless communication methods that are vulnerable to attacks. Due to the limitation of the sensor node's own conditions and the variability of the application environment, the wireless sensor network is often vulnerable to various attacks. Including selective attacks, flood attacks, Wormhole attacks, Sybil attacks, Sinkhole attacks, false routing, ACK counterfeit attacks, DOS attacks, conflict-causing attacks, unfair resource possession attacks, etc. According to the attack mode, attacks can be divided into external attacks and internal attacks. An external attack refers to an attack from outside the system. The identity of the initiator of this attack has not been legally verified. The attacker can only separate the network or introduce a large amount of data flow to interfere with the normal communication of the network. Compared with external attacks, internal attacks are carried out by capturing the internal nodes of the network or cracking the key structure of the network. This attack method is more difficult to detect and poses a greater threat to the normal communication of the network.

The traditional encryption algorithm-based security mechanism can only solve the external attack whose identity has not been authenticated, but cannot solve the internal attack problem caused by the capture of the node. The study found that, as an effective supplement to the traditional encryption mechanism, the trust evaluation mechanism can well solve the internal attack problem.

The research on trust evaluation of wireless sensor networks mainly focuses on modeling and calculating the factors that affect the trust value of nodes, and controls and manages them according to the trust value of nodes, and strives to establish a trusted network composed of trusted nodes, so as to ensure that the network can Long-term safe and reliable operation.

Contents of the invention

In order to reduce the influence of malicious nodes on the network and improve network security, the present invention proposes a multi-factor-based wireless sensor network dynamic trust evaluation system and method, and designs and implements a wireless sensor network dynamic trust evaluation platform. The dynamic trust evaluation platform for wireless sensor networks realizes the formal analysis of wireless sensor networks and the security detection of wireless sensor networks common to multiple protocols.

A multi-factor based wireless sensor network dynamic trust evaluation system of the present invention, the system includes a trust factor acquisition module 11, a trust measurement module 12, a trust calculation module 13 and a trust evaluation module 14, wherein

The trust factor acquisition module 11 acquires the trust factors involved in the trust evaluation process, including direct trust factors, indirect trust factors and comprehensive trust factors;

The trust measurement module 12 defines the measurement methods of each trust factor including direct trust factors, indirect trust factors and comprehensive trust factors;

The trust calculation module 13 realizes the calculation of each trust value according to the direct trust value, indirect trust value and comprehensive trust value measurement methods defined by the above-mentioned trust measurement module;

The trust evaluation module 14 compares the relationship between the comprehensive trust value of the target node obtained in the trust calculation process and the trust threshold, and judges whether the target node is credible based on this.

A kind of wireless sensor network dynamic trust evaluation method based on multiple factors of the present invention, this method comprises the following steps:

Step 1. Obtain the trust factors that may affect the trust value of nodes, that is, the current direct trust factors and historical direct trust factors included in the direct trust factors. The current direct trust factors are composed of the communication trust factors and sensing trust factors in the current cycle, and the two Both of them are obtained through the direct interaction process between the monitoring node working in promiscuous mode and other nodes; the indirect trust factor comes from the recommendation of other adjacent nodes; the direct trust factor and indirect trust factor are combined into a comprehensive trust factor, and the comprehensive The trust value is the ultimate basis for evaluating the credibility of nodes;

Step 2. Define the measurement methods of each trust factor including direct trust factor, indirect trust factor and comprehensive trust factor;

Step 3. Realize various trust calculations, including:

Direct trust calculation, analyzing and defining communication trust factor w ₁ , sensing trust value factor w ₂ , time decay factor f(tt ₀ ), and historical impact factor θ used in the calculation process;

Indirect trust calculation, analysis and definition of recommendation trust classification algorithm; use this algorithm to divide the recommendation opinions of adjacent nodes into two types: definite recommendation and uncertain recommendation, and define subjective evaluation algorithm and deviation test algorithm for trust calculation Process weight assignment;

Comprehensive trust calculation, which defines a dynamic weight distribution algorithm based on the number of interactions, and evaluates the credibility of nodes through the comprehensive trust value obtained by synthesizing direct trust value and indirect trust value;

Step 4, comparing the relationship between the comprehensive trust value of the target node obtained in the trust calculation process and the trust threshold to determine whether the target node is credible;

If the trust value of the node is greater than or equal to the trust threshold, it is judged that the node is credible and can continue to interact with it; if the trust value of the node is less than the trust threshold, the node is untrustworthy, no longer interacts with it, and will be deleted Related routing information, identification of untrusted nodes and other follow-up processing.

Compared with the prior art, the present invention is different from the previous wireless sensor network dynamic trust evaluation method, which only focuses on the behavior of nodes in the route discovery process; comprehensively considers the behavior of nodes when transmitting sensing data, so as to monitor nodes more comprehensively Behaviors to ensure timely and effective processing of untrustworthy nodes in the network; In addition, the present invention uses a comprehensive method of direct trust and indirect trust to measure the trust value of nodes, which solves the problem of lack of interaction records and insufficient evidence between nodes. Positive effects that lead to deviation problems in evaluation results.

Description of drawings

Fig. 1 is the frame diagram of the wireless sensor network dynamic trust evaluation system based on multiple factors of the present invention;

Figure 2 is a diagram of how to obtain trust factors;

Fig. 3 is a flow chart of indirect trust value calculation;

Figure 4 is a recommended trust relationship diagram;

Figure 5 is a flow chart of dynamic trust evaluation;

Figure 6 is an example diagram of node deployment;

Figure 7 is a comparison diagram of experimental results; (a), 10% malicious node network throughput; (b), 10% malicious node network normal delivery rate; (c), 20% malicious node network throughput; (d), 20% %Malicious node network normal delivery rate; (e), 30% malicious node network throughput; (f), 30% malicious node network normal delivery rate; (g), original network overhead; (h), trust mechanism-based Network overhead.

Detailed ways

The present invention will be described in further detail below in conjunction with the accompanying drawings.

As shown in Figure 1, the multi-factor based WSN dynamic trust evaluation system mainly includes four parts:

1. Trust factor acquisition module 11: mainly used to acquire trust factors involved in the trust evaluation process, including direct trust factors (current direct trust factors, historical direct trust factors), indirect trust factors and comprehensive trust factors. As shown in FIG. 2 , it is a schematic diagram of a manner of obtaining trust factors. Direct trust factors are further divided into communication trust factors and sensing trust factors, which are obtained by monitoring the direct interaction process of nodes working in promiscuous mode. The current direct trust factor is composed of the communication trust factor and the sensing trust factor of the current period, and both of them are obtained through the direct interaction process between the nodes working in promiscuous mode and other nodes. The historical direct trust factor is the direct trust factor of the historical cycle. The indirect trust factor comes from the recommendation of other neighboring nodes. The direct trust factor and the indirect trust factor constitute the comprehensive trust factor, and the comprehensive trust value is the ultimate basis for evaluating the credibility of nodes. This module is used to analyze and summarize the factors that may affect the node trust value and how to obtain it. The interaction mentioned above means that when a node sends a data packet to the next hop node, it judges whether the next hop node is correct by comparing the data packet cached locally by the node with the data packet forwarded by the monitored next hop node. The packet was forwarded. The basis of monitoring implementation is to set the node to work in promiscuous mode. If it is forwarded correctly, it is regarded as a successful interaction, otherwise it is a failed interaction. Trust factors come from the historical interaction behavior records between the main entity (referring to the wireless sensor network source node) and the target entity (referring to the wireless sensor network destination node), the current behavior characteristics of the target entity, and the recommendations of other entities.

2. The trust measurement module 12 is used to define the measurement mode of each trust factor:

Definition 1. The current direct trust value CDT _i,j (t) is defined as:

Among them, RSCF _i,j (t) and DSCF _i,j (t) represent the communication trust value and sensing trust value of the master node i with respect to the target node j in the current period t respectively, w ₁ and w ₂ represent the communication trust factor and the sensing trust value factor.

Definition 2. The historical direct trust value HDT _i,j (t) is defined as:

HDT _i,j (t)=f(tt ₀ )*CDT _i,j (t ₀ ) (2)

Among them, t CDT _i,j (t ₀ ) represents the direct trust value of the master node i on the target node j in the historical period t ₀ , and the time decay factor f(tt ₀ ) represents the time interval between the current period t and the historical period t ₀ linear function of .

Definition 3. The direct trust value DT _i,j (t) is defined as:

DT _i,j (t)＝θ*HDT _i,j (θ)+(1-θ)*CDT _i,j (t) (3)

Among them, HDT _i,j (t) and CDT _i,j (t) represent the historical direct trust value and the current direct trust value of the master node i on the target node j in the current period t, respectively, and θ represents the historical impact factor;

Definition 4. The indirect trust value IT _i,j (t) is defined as:

Among them, DT _i,k (t) represents the direct trust value of master node i on neighbor node k in the current period t, the number of interactions between master node i and neighbor node k is m, and T _k,j (t) represents the current period t The comprehensive trust value of internal target node j with respect to neighbor node k is T _k,j (t), and the recommended trust synthesis weight of neighbor node k is w' _k (t);

Definition 5. The comprehensive trust value T _i,j (t) is defined as:

T _i,j (t)=β′*DT _i,j (t)+(1-β′)*IT _i,j (t) (5)

DT _i,j (t), IT _i,j (t) and T _i,j (t) respectively represent the direct trust value, indirect trust value and comprehensive trust value of master node i on target node j in period t; β′ Indicates the weight of the direct trust value DT _i,j (t) in the synthesis process of the comprehensive trust value T _i,j (t).

3. The trust calculation module 13 is used to realize the calculation of each trust value according to the direct trust value, indirect trust value and comprehensive trust value measurement methods defined by the above-mentioned trust measurement module:

(1) Direct Trust Computing

For direct trust calculation, the communication trust factor w ₁ , sensor trust value factor w ₂ , time decay factor f(tt ₀ ), and historical impact factor θ are analyzed and defined in the calculation process, according to the formula (1)- (3) Calculate the direct trust value.

The direct trust value is obtained by calculating the probability of successful interaction between nodes in the process of establishing the path and transmitting the sensory data, and performing a weighted average on it.

In period t, t represents the probability of successful interaction between node i and node j in the process of establishing the path. Let the number of successful interactions be rsc _i,j (t), and the number of failed interactions be rfc _i,j (t), then

In period t, DSCF _i,j (T) represents the probability of successful interaction between node i and node j in the process of transmitting sensor data, let the number of successful interactions be dsc _i,j (t), and the number of failed interactions be dfc _{i ,j} (t), then

The values of the communication trust factor w ₁ and the sensing trust factor w ₂ are determined by the possible impact range of the node’s malicious behavior on the network during the node interaction process. A malicious behavior of a node when establishing a path may affect the normal construction of multiple paths, and a malicious behavior when transmitting sensing data will only affect the normal transmission of the data packet forwarded this time, so it is usually defined that 0≤w ₂ ≤w ₁ ≤ 1.

The direct trust value is a combination of the direct trust value in the historical cycle and the direct trust value in the current cycle with a certain historical impact factor. Among them, the historical direct trust value is obtained after the direct trust value decays in the historical period. In order to avoid the situation that originally trusted nodes are mistaken for untrusted nodes because they have not interacted for a long time, resulting in waste of resources, this paper defines the attenuation as bounded. Assume, CDT _i,j (t ₀ ) represents the historical direct trust value of node i on node j in period t ₀ , the time decay factor f(tt ₀ ), the trust threshold Th for distinguishing credible and untrustworthy, CDT _i,j (t ₀ ) becomes HDT _i,j (t) after attenuation, expressed as:

In addition, considering the "volatile and hard-to-get" of trust, the historical impact factor θ defined in formula (22):

Wherein 0<θ ₂ <0.5<θ ₁ <1.

The values of θ ₁ , θ ₂ , Th, and f above are all determined by the actual application environment.

(2) Indirect Trust Computing

Indirect trust calculation, analysis and definition of recommendation trust classification algorithm; in order to solve the problem of coarse granularity of recommendation trust classification, use this algorithm to divide the recommendation trust of adjacent nodes into two categories: definite recommendation and uncertain recommendation, and define subjective evaluation respectively Algorithm and Deviation Test Algorithm is used in the weight distribution of the trust calculation process, and the indirect trust value is calculated according to formula (4).

The process of indirect trust acquisition starts with the master node periodically broadcasting a recommendation trust request about a target node to the adjacent nodes in a single hop. Set the credible threshold Th, and determine the recommendation _n'min , step 301; node i broadcasts the recommendation request packet RequestTrust about the target node j, step 302; neighbor node k receives the recommendation request packet RequestTrust from node i, step 303; Search the trust table Trust maintained by itself, step 304; judge whether there is a comprehensive trust value of node j in the trust table Trust, step 305; if not, discard the recommendation request packet RequestTrust, and the process ends, step 306; Node k sends a recommended reply ReplyTrust to the node, step 307; node j receives the recommended reply ReplyTrust from neighbor node k, step 308; finds the trust table Trust maintained by itself, step 309; judges that there is a direct trust value of neighbor node k and the trust value is greater than Th, step 310; if it does not exist, then discard ReplyTrust, step 311; if it exists, further judge whether the number of interactions with node k is greater than _n'min , step 312; if greater, then neighbor node k is to determine the recommended trust class, Step 313; Execute the subjective evaluation algorithm to determine the composite weight, step 314; determine the indirect trust value of the recommended class synthesis, step 315; if not greater, then the neighbor node k is an uncertain recommended trust class, step 316; execute the deviation test algorithm Determine the composition weight, step 317 ; determine the indirect trust value of the recommended category synthesis, step 318 ; obtain the final indirect trust value by weighting, step 319 . The above calculation process is briefly described as follows: after the adjacent node receives the recommendation request packet sent by the master node, it searches the trust table maintained by itself to determine whether there is a trust value of the target node. If it exists, send a recommendation reply packet to the master node; otherwise, it will not be processed. After receiving the recommendation reply data packet within the allowable delay, the master node looks up the cached recommendation request data packet to determine whether the recommendation request has been sent. If it exists, look up the trust table, continue to judge whether the trust value of the recommended node is greater than the recommended trust threshold, if it is greater, accept the recommendation, otherwise not accept it.

In order to solve the problems caused by trust chains, such as complex calculation process, slow convergence speed, and inability to adapt to large-scale networks, in this paper, the master node i only refers to the common neighbor nodes k ₁ , k ₂ , of the master node i and the target node j. .., k _n-1 , k _n recommendations to calculate the indirect trust value of target node j.

As shown in Figure 4, it is a recommended trust relationship graph between node i and node j. Among them, the one-way solid line indicates the direct trust relationship, the two-way solid line indicates the comprehensive trust relationship, and the dotted line indicates the recommendation trust relationship. In order to solve the problem of malicious recommendation and collusion attack, it is defined that node i judges whether to accept the recommendation of neighbor node k according to its own trust evaluation result of neighbor node k, that is, the direct trust value. If the direct trust value of the neighbor node is less than the trust threshold, the neighbor node is an untrustworthy node, and node i does not accept the recommended trust value provided by it. For an acceptable recommended trust value, in order to avoid the secondary impact on the comprehensive trust evaluation result of node j due to the small number of interactions and insufficient evidence between node i and adjacent node k, this paper comprehensively considers two factors: the number of interactions and the direct trust value Determine the recommendation trust composition weight.

Assuming that the total number of interactions between node i and neighbor node k is m, and the direct trust value of node i with respect to node k is DT _i,k (t), then it can be known from formula (3-4) that the composite weight w′ _k is:

w′ _k (t)=f″(D _Ti,k (t),m) (10)

According to the number of interactions between nodes, the recommendation trust is divided into two categories: definite recommendation and uncertain recommendation, and the subjective evaluation algorithm and deviation test algorithm are respectively defined for the weight distribution in the process of indirect trust synthesis.

As stated in the extended Hoeffding's inequality: "Within the acceptable error range t, if and only if the number of node interactions is at least n _min , the credibility of the trust value obtained by monitoring the direct interaction process satisfies at least α" . Assuming that the minimum number of interactions within an acceptable error range t' and reliability α' is _n'min , _n'min is the threshold for determining the recommendation, and based on this, the definite recommendation class and the uncertain recommendation class are divided.

If _m≥n′min , it is considered that the recommendation trust of neighbor node k belongs to the definite recommendation trust category, and the weight distribution algorithm based on subjective evaluation is adopted to synthesize the definite recommendation trust value (node i is based on the direct trust value DT _{i,k of neighbor node k} ( t) to determine the synthetic weight), otherwise it is an uncertain recommendation trust class, and at the same time adopts a deviation-based weight distribution algorithm to synthesize an uncertain recommended trust value (node i is based on the recommended trust value provided by neighbor node k and other neighbor nodes provide The degree of deviation between the recommended trust values DT _i,k (t) to determine the synthetic weight).

Finally, the weighted average operation is performed on the confirmed recommended trust value and the uncertain recommended trust value to obtain the indirect trust value of node i on node j.

Algorithm 1: Subjective evaluation weight distribution

Input: node i's direct trust value DT _i,1 (t), DT _i,2 (t),...,DT _i,m (t) of neighbor nodes k ₁ ,k ₂ ,...,k _m .

Output: weights of neighbor nodes k ₁ , k ₂ ,...,k _m .

①Calculate the sum of the direct trust values of node i with respect to all neighbor nodes belonging to the determined recommendation class.

Among them, DT _i,k (t) represents the direct trust value of input node i and neighbor node k _k ; k represents the number of neighbor nodes;

② Calculate the weight w′ _k (t) of the neighbor node k;

w'k (t)=DT _{i,k (} t)/S'(t) (12)

Among them, DT _i,k (t) represents the direct trust value of input node i and neighbor node k _k ;

Algorithm 2: Deviation test weight distribution

Input: neighbor nodes k _m+1 ,k _m+2 ,...,k _n 's comprehensive trust value T _m+1,j (t), T _m+2,j (t), .. .,T _n,j (t);

Output: weights of neighbor nodes k _m+1 ,k _m+2 ,...,k _n .

① Calculate the degree of deviation between the recommended trust value provided by neighbor node k and the recommended trust value provided by other nodes.

Among them, T _k,j (t) means that calculated by formula (5) is the comprehensive trust value of master node k with respect to target node j in period t; T _r,j (t) means that calculated by formula (5) The output is the comprehensive trust value of the master node r with respect to the target node j in the period t; n represents the number of neighbor nodes;

②Calculate the total deviation value of the uncertain trust recommendation class.

Among them, s _k (t) represents the degree of deviation between the recommended trust value provided by neighbor node k calculated by formula (13) and the recommended trust value provided by other nodes;

③ Calculate the proportional relationship between the total deviation value and the deviation value s _k (t) of the neighbor node k.

d _k (t) = S (t) / s _k (t) (15)

Among them, s _k (t) represents the degree of deviation between the recommended trust value provided by neighbor node k calculated by formula (13) and the recommended trust value provided by other nodes; S(t) represents the degree of deviation calculated by formula (14) The total deviation value of the uncertain trust recommendation class;

④ Normalize d _k (t) to get the weight w′ _k (t) of node k.

Among them, d _k (t) is the total deviation value calculated by formula (15).

(3) Comprehensive trust calculation method

Comprehensive trust calculation, which defines a dynamic weight distribution algorithm based on the number of interactions, and evaluates the credibility of nodes by synthesizing the comprehensive trust value obtained by direct trust value and indirect trust value; in addition, the trust value needs to be updated periodically in order to observe in time The behavior of the node changes.

When the number of interactions between the master node and the target node is small, the master node may not be able to fully rely on its own knowledge to make an accurate evaluation of the target node due to insufficient evidence. In order to ensure the accuracy of the evaluation results, this paper uses a combination of direct trust and indirect trust to evaluate the trust value of nodes. In addition, the trust value needs to be updated periodically, so as to achieve the purpose of real-time monitoring of node behavior.

Most of the existing trust evaluation methods for wireless sensor networks use the method of static synthesis of direct trust value and indirect trust value to calculate the comprehensive trust value of nodes, which does not conform to the dynamic characteristics of wireless sensor networks. On the determination of the composite weight, based on the Hoeffding's inequality, the dynamic corresponding relationship between the number of interactions and the composite weight is established, which solves the problem that the static weight distribution in the previous wireless sensor network trust evaluation method does not conform to the dynamic change characteristics of the wireless sensor network; Therefore, this paper defines a dynamic weight assignment algorithm based on Hoeffding's inequality: within the allowable range of error t, when the number of node interactions is n, the minimum credibility of the evaluation results obtained is the composite weight of the direct trust value in the comprehensive trust synthesis process .

In addition, in order to monitor the behavior of nodes to the greatest extent, we define that the synthetic weight of the direct trust value is bounded, that is to say, node i should more or less refer to the recommendations of other adjacent nodes. The weight value range of the direct trust value in the process of comprehensive trust synthesis is determined by the actual application environment. It is assumed that the value range is [c,d], where 0<c<d<1.

Algorithm 3: Comprehensive trust weight distribution, that is, dynamic weight distribution algorithm:

Input: the acceptable error range e of the current environment, and the number of interactions m.

Output: synthetic weight β of direct trust value.

①The weight β, the error range e and the number of interactions m satisfy the equation.

2exp(-2t ² m)＝1-β (17)

Among them, the weight β represents the composite weight of the direct trust value; m represents the number of interactions;

② Transformation equation.

β＝1-2exp(-2t ² m) (18)

Among them, the weight β represents the composite weight of the direct trust value; m represents the number of interactions;

③ Bounded treatment of β.

Among them, the weight range of the direct trust value in the comprehensive trust synthesis process is determined by the actual application environment, and the value range is assumed to be [c,d], where 0<c<d<1.

Prior to this, node i has obtained the direct trust value and indirect trust value of node j through direct trust calculation and indirect trust calculation respectively, and combined with the dynamic weight distribution algorithm defined in this section, it can be obtained that in period t, node i is about The comprehensive trust value of node j.

The definition and extension of Hoeffding's inequality are described as follows:

In probability theory, Hoeffding's inequality is used to define an upper bound on the probability of the absolute difference between the mean of a set of random variables and its expected value.

Assume a series of independent random variables X ₁ , X ₂ ,...,X _n-1 ,X _n , assuming that all X _i are almost bounded variables, that is, satisfy

P(X _i ∈ [a,b])=1,1≤i≤n (20)

Define the empirical expectation (estimated value) of these n random variables as:

for The expectation (true value) of and satisfies the following inequality:

As defined in this paper, the random variable Xi represents the result of the _ith interaction between nodes. If the interaction is successful, Xi ₌ 1, otherwise Xi ₌ 0.

Therefore, this article satisfies P(r*X _i ∈(a,b))=1, and a=0, b=1. which is:

Among them, r is the adjustment factor. In this paper, the adjustment factor is related to communication trust factor, sensing trust factor, time decay factor, and historical impact factor.

Under the premise of a given number n of user transaction evaluation results, the estimation error of private reputation is greater than the upper bound of the probability of a given threshold t Satisfy:

That is:

That is:

In this paper, formula (26) is described as: within the acceptable error t range, if and only if the node interaction times are at least n _min , the credibility of the trust value obtained by monitoring the direct interaction process satisfies at least α.

4. The trust evaluation module 14 is used to compare the relationship between the comprehensive trust value of the target node obtained in the trust calculation process and the trust threshold to determine whether the target node is trustworthy.

If the trust value of the node is greater than or equal to the trust threshold, it is judged that the node is credible and can continue to interact with it; if the trust value of the node is less than the trust threshold, the node is untrustworthy, no longer interacts with it, and will be deleted Related routing information, identification of untrusted nodes and other follow-up processing.

Figure 6 is an example diagram of node deployment in this example. Taking the process of node 27 calculating the comprehensive trust value of node 9 in the sixth calculation cycle as an example, the dynamic trust evaluation method in this paper is described.

In the existing trust evaluation methods, the initial trust value of the setting node is divided into three situations: the maximum value, the middle value and the minimum value. In order to prevent resource waste, we believe that a node should be considered trustworthy if no malicious behavior is detected by the node. Therefore, in the experiment, we define the initial trust value of the node as 1, and set the trust threshold as 0.5.

Setting of other parameters: communication trust factor and sensing trust factor are respectively w ₁ =1, w ₂ =0.5, historical impact factors θ ₁ =0.6, θ ₂ =0.4, time decay factor f(tt ₀ )=0.9*( tt ₀ ), the acceptable error range e=0.1. Regarding the recommended classification, the minimum acceptable reliability is α=Th=0.5, that is to say, the minimum number of interactions n' _m =70 for determining the recommended class. In the comprehensive trust synthesis process, the value range of the weight of the direct trust value is defined as c=0.2, d=0.8. The specific data are shown in Table 1.

Table 1. Calculation data and results

It can be seen from the above table that in the sixth cycle, the historical direct trust value of node 9 is 0.856744, and the current direct trust value is 0.649879. From formula (22), the direct trust value is 0.732625. From the number of interactions 70 determined by the determined recommendation classification algorithm, it can be seen that the node 40, the node 14, the node 35, the node 16, the node 2, the node 39, and the node 22 belong to the determined recommendation category. Node 26, node 43, node 1, node 10, and node 33 belong to the uncertain recommendation category. In addition, since the trust values of node 12 and node 5 are less than the trust threshold 0.5, node 27 does not refer to the recommendations of these two nodes.

It can be known from the comprehensive trust value 0.582219 greater than the trust threshold 0.5: in the sixth cycle, node 9 is credible to node 27.

Analysis of results

As shown in Figure 7, (a) and (b) indicate that when malicious nodes account for 10% of all nodes, the wireless sensor network with added trust mechanism is basically the same as the original wireless sensor network in terms of throughput and correct delivery rate. This is because in the experiment, the network nodes are randomly deployed, so it may be that when the number of malicious nodes is small, the influence of malicious nodes in the entire network is small, and the effect of the trust mechanism is not obvious.

As shown in (c) and (d) in Figure 7, as the number of malicious nodes increases, the effect of the trust mechanism is also more obvious. When malicious nodes account for 20% of all nodes, compared to the original WSN, the throughput and correct delivery rate of the trust-based WSN increase by an average of 15.4% and 27.6%, respectively.

As shown in (e) and (f) in Figure 7, when malicious nodes account for 30% of all nodes, the wireless sensor network based on the trust mechanism is significantly better than the wireless sensor network running the original AODV protocol, in terms of throughput and correct delivery rate The average growth rate was 30.6% and 54.8% respectively.

As shown in (g) and (h) in Figure 7, the network overhead of the wireless sensor network based on the trust mechanism is higher than that of the original wireless sensor network, because the wireless sensor network with the added trust mechanism may have problems due to the processing of untrusted nodes However, the new route discovery process will increase some network overhead, and the trust calculation process, especially the recommended trust acquisition process will require some additional network overhead. However, it can be found that when the number of malicious nodes continues to increase, the original wireless sensor network overhead increases rapidly, while the trust mechanism-based wireless sensor network overhead remains basically unchanged.

Claims (7)

1. a wireless sensor network dynamic trust evaluation system based on multiple factors, is characterized in that, the system includes trust factor acquisition module (11), trust measurement module (12), trust calculation module (13) and trust evaluation module (14 ),in The trust factor acquisition module (11) acquires the trust factors involved in the trust evaluation process, including direct trust factors, indirect trust factors and comprehensive trust factors; The trust measurement module (12) defines the measurement methods of each trust factor including direct trust factors, indirect trust factors and comprehensive trust factors; The trust calculation module (13) realizes the calculation of each trust value according to the direct trust value, indirect trust value and comprehensive trust value measurement methods defined by the above-mentioned trust measurement module; The trust evaluation module (14) compares the relationship between the comprehensive trust value of the target node obtained in the trust calculation process and the trust threshold, and judges whether the target node is credible based on this. 2. A method for evaluating dynamic trust in wireless sensor networks based on multiple factors, characterized in that the method comprises the following steps: Step 1. Obtain the trust factors that may affect the trust value of nodes, that is, the current direct trust factors and historical direct trust factors included in the direct trust factors. The current direct trust factors are composed of the communication trust factors and sensing trust factors in the current cycle, and the two Both of them are obtained through the direct interaction process between the monitoring node working in promiscuous mode and other nodes; the indirect trust factor comes from the recommendation of other adjacent nodes; the direct trust factor and indirect trust factor are combined into a comprehensive trust factor, and the comprehensive The trust value is the ultimate basis for evaluating the credibility of nodes; Step 2. Define the measurement methods of each trust factor including direct trust factor, indirect trust factor and comprehensive trust factor; Step 3. Realize various trust calculations, including: Direct trust calculation, analyzing and defining communication trust factor w ₁ , sensing trust value factor w ₂ , time decay factor f(tt ₀ ), and historical impact factor θ used in the calculation process; Indirect trust calculation, analysis and definition of recommendation trust classification algorithm; use this algorithm to divide the recommendation opinions of adjacent nodes into two types: definite recommendation and uncertain recommendation, and define subjective evaluation algorithm and deviation test algorithm for trust calculation Process weight assignment; Comprehensive trust calculation, which defines a dynamic weight distribution algorithm based on the number of interactions, and evaluates the credibility of nodes through the comprehensive trust value obtained by synthesizing direct trust value and indirect trust value; Step 4, comparing the relationship between the comprehensive trust value of the target node obtained in the trust calculation process and the trust threshold to determine whether the target node is credible; If the trust value of the node is greater than or equal to the trust threshold, it is judged that the node is credible and can continue to interact with it; if the trust value of the node is less than the trust threshold, the node is untrustworthy, no longer interacts with it, and will be deleted Related routing information, identification of untrusted nodes and other follow-up processing. 3. a kind of wireless sensor network dynamic trust evaluation method based on multiple factors as claimed in claim 1, is characterized in that, the definition of described step 2 comprises each trust factor including direct trust factor, indirect trust factor and comprehensive trust factor. factors, including the following steps: Definition 1. The current direct trust value CDT _i,j (t) is defined as: Among them, RSCF _i,j (t) and DSCF _i,j (t) represent the communication trust value and sensing trust value of the master node i with respect to the target node j in the current period t respectively, w ₁ and w ₂ represent the communication trust factor and sensing trust value factor; Definition 2. The historical direct trust value HDT _i,j (t) is defined as: HDT _i,j (t)=f(tt ₀ )*CDT _i,j (t ₀ ) (2) Among them, t CDT _i,j (t ₀ ) represents the direct trust value of the master node i on the target node j in the historical period t ₀ , and the time decay factor f(tt ₀ ) represents the time interval between the current period t and the historical period t ₀ the linear function of; Definition 3. The direct trust value DT _i,j (t) is defined as: DT _i,j (t)＝θ*HDT _i,j (θ)+(1-θ)*CDT _i,j (t) (3) Among them, HDT _i,j (t) and CDT _i,j (t) represent the historical direct trust value and the current direct trust value of the master node i on the target node j in the current period t, respectively, and θ represents the historical impact factor; Definition 4. The indirect trust value IT _i,j (t) is defined as: Among them, DT _i,k (t) represents the direct trust value of master node i on neighbor node k in the current period t, the number of interactions between master node i and neighbor node k is m, and T _k,j (t) represents the current period t The comprehensive trust value of internal target node j with respect to neighbor node k is T _k,j (t), and the recommended trust synthesis weight of neighbor node k is w' _k (t); Definition 5. The comprehensive trust value T _i,j (t) is defined as: T _i,j (t)=β′*DT _i,j (t)+(1-β′)*IT _i,j (t) (5) DT _i,j (t), IT _i,j (t) and T _i,j (t) respectively represent the direct trust value, indirect trust value and comprehensive trust value of master node i on target node j in period t; β′ Indicates the weight of the direct trust value DT _i,j (t) in the synthesis process of the comprehensive trust value T _i,j (t). 4. A kind of wireless sensor network dynamic trust evaluation method based on multi-factors as claimed in claim 1, is characterized in that, each trust calculation of the realization of described step 3, wherein the specific step of direct trust value calculation also includes communication trust factor The values of w ₁ and sensing trust factor w ₂ are determined by the influence range of the node’s malicious behavior on the network during the node interaction process; The analysis process of the time decay factor f(tt ₀ ) is as follows: the historical direct trust value CDT _i,j (t ₀ ) of node i with respect to node j in period t ₀ becomes HDT _i,j (t) after attenuation, expressing for: Among them, f(tt ₀ ) represents the time decay factor, Th represents the trust threshold for distinguishing credible and untrustworthy, The historical impact factor θ is defined as: Among them, 0<θ ₂ <0.5<θ ₁ <1; the values of θ ₁ , θ ₂ , Th, and f are determined by the actual application environment. 5. a kind of wireless sensor network dynamic trust evaluation method based on multiple factors as claimed in claim 1, is characterized in that, the subjective evaluation algorithm of described step 3, this algorithm specifically comprises the following processing: The input is the direct trust value DT _i,1 (t),DT _i,2 (t),...,DT _i,m (t) of node i with respect to neighbor nodes k ₁ ,k ₂ ,...,k _m ; The output is the weight of neighbor nodes k ₁ , k ₂ ,...,k _m ; Calculate the sum of the direct trust values of node i with respect to all neighbor nodes belonging to the determined recommendation class; Among them, DT _i,k (t) represents the direct trust value of input node i and neighbor node k _k ; k represents the number of neighbor nodes; Calculate the weight w′ _k (t) of the neighbor node k; Among them, DT _i,k (t) represents the direct trust value of input node i and neighbor node k _k . 6. a kind of wireless sensor network dynamic trust evaluation method based on multiple factors as claimed in claim 1, is characterized in that, the degree of deviation test algorithm of described step 3, this algorithm specifically comprises the following processing: The input is the comprehensive trust value T _m+1,j (t), T _m+2,j (t) of the neighbor node k _m+1 ,k _m+2 ,...,k _n about the target node j .,T _n,j (t); The output is the weight of neighbor nodes k _m+1 , k _m+2 ,...,k _n ; Calculate the degree of deviation between the recommended trust value provided by neighbor node k and the recommended trust value provided by other nodes; Among them, T _k,j (t) means that calculated by the formula (5) is the relationship between the main node k and the target node j in the period t. Comprehensive trust value; T _r,j (t) means the comprehensive trust value calculated by the formula (5) is the comprehensive trust value of the master node r with respect to the target node j in the period t; n means the number of neighbor nodes; Calculate the total deviation value of the uncertain trust recommendation class; Among them, s _k (t) represents the degree of deviation between the recommended trust value provided by neighbor node k calculated by formula (13) and the recommended trust value provided by other nodes; Calculate the proportional relationship between the total deviation value and the deviation value s _k (t) of the neighbor node k; d _k (t) = S (t) / s _k (t) (15) Among them, s _k (t) represents the degree of deviation between the recommended trust value provided by neighbor node k calculated by formula (13) and the recommended trust value provided by other nodes; S(t) represents the degree of deviation calculated by formula (14) The total deviation value of the uncertain trust recommendation class; Normalize d _k (t) to get the weight w′ _k (t) of node k; Among them, d _k (t) is the total deviation value calculated by formula (15). 7. A kind of wireless sensor network dynamic trust evaluation method based on multiple factors as claimed in claim 1, is characterized in that, the dynamic weight distribution algorithm of described step 3, this algorithm specifically comprises the following processing: The input is the acceptable error range e of the current environment, and the number of interactions m; The output is the composite weight β of the direct trust value; The weight β, the error range e and the number of interactions m satisfy the equation; 2exp(-2t ² m)＝1-β (17) Among them, the weight β represents the composite weight of the direct trust value; m represents the number of interactions; transformation equation; β＝1-2exp(-2t ² m) (18) Among them, the weight β represents the composite weight of the direct trust value; m represents the number of interactions; Bounded treatment of β; Among them, the weight range of the direct trust value in the comprehensive trust synthesis process is determined by the actual application environment, and the value range is assumed to be [c,d], where 0<c<d<1.