WSN topology control method based on dynamic fuzzy trust and artificial bee colony algorithm

Abstract

The invention discloses a WSN (Wireless Sensor Network) topology control method based on dynamic fuzzy trust and an artificial bee colony algorithm, which only performs security detection when malicious activities exist in a network. Firstly, a sliding time window mechanism is introduced into trust value calculation to ensure the dynamism and timeliness of trust values. And then, by introducing a forwarding delay judgment mechanism, a serious security problem in the network can be quickly identified. And meanwhile, the indirect trust value is calculated by utilizing a fuzzy comprehensive evaluation method, so that the comprehensiveness and the accuracy of trust evaluation are enhanced. And finally, detecting and isolating dishonest recommended nodes by adopting an artificial bee colony algorithm, thereby further improving the accuracy of trust evaluation and the overall security of the network. According to the method, for the large-scale wireless sensor network, unnecessary energy consumption is effectively reduced, the credibility of the malicious node recognition rate is improved, and the misjudgment rate is low. And the comprehensive capability in energy consumption, safety and other performances is better.

Description

Technical Field

[0001] The present invention belongs to the field of routing security of wireless sensor networks, and particularly relates to a WSN topology control method based on dynamic fuzzy trust and artificial bee colony algorithm. Background Technique

[0002] Wireless Sensor Networks (WSN) is a distributed sensing network, which is a wireless network composed of a large number of stationary or mobile sensors in an ad-hoc and multi-hop manner. Currently, it is widely used in military, environmental monitoring, medical, industrial production, traffic control and other fields. Although traditional password verification-based methods perform well in resisting external attacks on the Internet, their ability to prevent internal attacks is insufficient. Therefore, how to effectively deal with internal attackers in WSNs has become a hot research issue in this field. A trust mechanism has been widely introduced into routing protocols to resist internal attacks. By constructing a trust model and organically combining direct and indirect trust values, the credibility of nodes can be more comprehensively evaluated, thus effectively identifying malicious nodes in the network.

[0003] The SRBNT protocol by Wu et al. is based on node trust, sets multiple reference standards, combines various performance indicators, and uses a weighted average method to assign different weights to each indicator according to its importance and perform comprehensive calculations. The TSRF protocol proposed by Duan J et al. incorporates the current behavior and historical behavior of nodes into the trust evaluation category and defines a direct trust value based on these two aspects. At the same time, the protocol also combines the recommended trust value of neighbor nodes and proposes specific trust calculation and derivation methods. The TRPM protocol proposed by Sun B et al. introduces trust evaluation attributes and a time sliding window to deal with on-off attacks.

[0004] However, most trust models rarely consider attacks on the evaluation system itself. These attacks are often diverse and more concealed, such as dishonest recommendation attacks, including malicious language attacks, ballot stuffing attacks, collusion attacks, etc. Dishonest recommendation attacks are selective forwarding attacks based on the security trust model. Malicious nodes forge trust values, spread dishonest recommendations, and hide their identities from other attackers.

[0005] In summary, for existing algorithms for defending against dishonest recommendations, two problems still need to be solved. First, the trust degree between nodes is subjective and uncertain. It is difficult to distinguish malicious nodes and faulty nodes among low-trust nodes. Summary of the Invention

[0006] In order to solve the technical problems mentioned in the above background technique, the present invention proposes a WSN topology control method based on dynamic fuzzy trust and artificial bee colony algorithm.

[0007] To achieve the above technical objectives, the technical solution of the present invention is as follows:

[0008] A WSN topology control method based on dynamic fuzzy trust and artificial bee colony algorithm, comprising the following steps:

[0009] (1) Introduce a sliding time window in the calculation of trust values, periodically update the trust values of nodes, and ensure the timeliness and dynamics of trust values between nodes. And add forwarding delay judgment;

[0010] (2) Use fuzzy comprehensive evaluation to calculate the indirect trust degree, take into account multi-dimensional factors such as the data attributes and physical attributes of the target node in the trust calculation, and improve the accuracy of trust evaluation;

[0011] (3) Use the ABC algorithm to detect and isolate dishonest recommendation nodes to obtain the comprehensive trust degree. It improves the credibility of the malicious node recognition rate and has a low false positive rate.

[0012] Furthermore, in step (1), the specific calculation steps of the direct trust degree are as follows:

[0013] (101) Forwarding probability F i,j (t): Considering the above-mentioned trust metric indicators, a probability-based trust model is adopted here to calculate and evaluate the forwarding probability of the target node recorded by the node. That is, within a communication cycle, using the Beta distribution as the calculation model and representing it by the successful forwarding ratio, the forwarding probability calculation formula is as follows:

[0014]

[0015] Among them, F i,j (t) represents the forwarding probability of the target node N j , s i,j is the number of successful forwards of node N j , f i,j is the number of forwarding failures of N j ;

[0016] (102) Historical trust degree HT i,j(t): The trust value between nodes has significant timeliness and dynamics, and its current state changes continuously due to the actual performance of the nodes. Since the network environment and node behavior may fluctuate over time, relying solely on historical trust values cannot comprehensively reflect the current situation of the nodes or predict their future behavior. The sliding time window dynamically updates the trust value of nodes by limiting the effective range of historical data. Under this mechanism, only the node interaction records within the sliding time window are retained as the valid data for calculating the trust value, while the historical records outside the window range are discarded. The sliding time window consists of multiple time slots, each corresponding to a fixed time period for recording the trust value of the node during that period. In the process of updating the historical trust value, the older the record, the smaller its contribution should be. In this section, a time decay function is adopted to determine the historical trust degree weight, and the formula is as follows:

[0017]

[0018] where λ is the decay factor, with a value between 0 and 1, and t k is the k-th time slot of the window.

[0019] Assume that the historical trust degree sequence saved in each time slot within the sliding time window at the current moment is Then the historical trust degree calculation formula is as follows:

[0020]

[0021] (103) Direct trust degree DT i,j (t): The evaluating node establishes the trust relationship between nodes based on the forwarding probability and forwarding delay of the target node. However, the complexity and dynamic changes of the network environment may lead to fluctuations in node behavior, which makes a single piece of current behavior data insufficient to comprehensively evaluate the true trustworthiness of the node. Therefore, historical trust values are introduced into the direct trust mechanism. The calculation formula is as follows:

[0022] DT i,j (t) = (w1 * F i,j (t) + w2 * HT i,j (t)) × FD#(1 - 4)

[0023] Among them, FD is the forwarding delay, which is used to measure the performance of nodes in the data forwarding process. When the forwarding delay time exceeds the preset security threshold, it is considered that there are serious security risks in this node. The value of FD is directly set to 0, indicating that this node shows abnormalities in terms of forwarding delay. On the contrary, the value of FD is set to 1. If the forwarding delay of a node is determined to be abnormal (the FD value is 0), its direct trust value will be directly reduced to 0. The direct trust value of a node is obtained based on the forwarding probability and the historical trust value through a weighted algorithm, where w1 + w2 = 1 and can be flexibly adjusted. Here, we take w1 = w2 = 0.5.

[0024] Furthermore, in step (2), the specific steps of the fuzzy comprehensive evaluation are as follows:

[0025] (201) Determine the factor set U: U = {u1, u2, u3}, where u i (i = 1, 2, 3) represents the i-th factor in the factor set. u1 is the energy factor, which represents the remaining energy of the target node. u2 is the transmission rate factor, which represents the rate at which the node sends data packets within a certain period of time. u3 is the neighbor factor, which represents the number of neighbors within one hop of the target node;

[0026] (202) Determine the evaluation set V: V = {v1, v2, v3}, where v j j = 1, 2, 3) represents the j-th evaluation level. v1 represents untrusted, v2 represents generally trusted, and v3 represents highly trusted;

[0027] (203) Construct the fuzzy judgment matrix R: The fuzzy judgment matrix R is the relationship matrix between the factor set U and the evaluation set V. For the factor u i perform a single-factor judgment evaluation to determine its membership degree to the evaluation subset v j (expressed as r ij ), and obtain a set of single-factor judgment factors r i = (r i1 , r i2 , r i3 ). The judgment factors of the three factors construct the fuzzy judgment matrix R.

[0028]

[0029] Among them, r ij represents the membership degree of the i-th factor to the j-th evaluation. For example, r 13 represents the membership degree of the energy factor in the highly trusted fuzzy subset;

[0030] (204) Determine the weight set A: A = {a1, a2, a3}, where A represents the weight coefficients of the importance of each factor for the comprehensive evaluation, and a1 + a2 + a3 = 1. According to the three-scale method, each weight coefficient is calculated as a1 = 0.10, a2 = 0.65, a3 = 0.25;

[0031] (205) Perform fuzzy operation to obtain the result vector B: The fuzzy comprehensive evaluation results of each evaluation object are as follows:

[0032]

[0033] Among them, b j is the membership degree of the node to the evaluation subset v j ;

[0034] (206) Use the weighted average method to calculate the indirect trust degree IT j of the target node N i,j (t), and the calculation formula is as follows:

[0035]

[0036] Among them, DT i,j (t) is the direct trust degree of the evaluation node N i calculated for the target node N j according to formula (4-4);

[0037] (207) In the one-to-one trust evaluation, the finally calculated indirect trust degree of the target node is denoted as λ. If λ ≥ γ, it means the node is trustworthy; if λ ≤ ξ, it means the node is a malicious node, and this node is added to the blacklist and removed from the network; when ξ < λ < γ, the node is regarded as a suspected node and needs further observation and evaluation. Among them, γ and ξ respectively represent the trust threshold and the malicious threshold, which are used to distinguish the behavior status of the node.

[0038] Furthermore, in step (3), the specific process of the comprehensive trust evaluation is as follows:

[0039] (301) Employed bee stage: In our algorithm, the food source corresponds to the target node, each employed bee corresponds to a recommended node, and the employed bees no longer select the food source for recommendation according to the greedy algorithm, but all employed bees evaluate and recommend the food source. If the evaluation node B i wants to evaluate the trust degree of the target node n j , it will send a request to the current one-hop neighbors of the target node. After the neighbor node N k receives the request, they work as employed bees and send the recommendation information about the target node (food source) to the evaluation node. Among them, the recommendation information includes its identification ID, direct trust, and indirect trust;

[0040] (302) Fitness function: Evaluate node B i Merge the received recommendation information, and then calculate the deviation function of the direct trust of these recommendations to detect false information. Recommend node N k Regarding the target node n j Recommendation deviation The calculation formula is as follows:

[0041]

[0042] Among them, can be regarded as the fitness function of the recommendation deviation, and m is the number of recommendation nodes.

[0043] For the recommendation node N k The weighted linear fitness function fit k The formula is as follows:

[0044]

[0045] (303) Observation bee stage: The evaluation node classifies these recommendation nodes according to the probability value p k If p k is less than the default threshold, then regard the recommendation node N k as a dishonest node, and then the evaluation node adds it to the blacklist. The nodes on the blacklist cannot interact and transfer data with other nodes, that is, regard it as a dead node. The calculation formula of the probability value p k is as follows:

[0046]

[0047] Evaluate node B i Merge the recommendation information of the recommendation nodes not on the blacklist to obtain the comprehensive trust degree of the target node n j The calculation formula is as follows: The calculation formula is as follows:

[0048]

[0049] Among them, It means that the recommendation nodes with smaller deviation values have greater reference weights, and y is the number of honest recommendation nodes;

[0050] (304) Scout bee stage: When the comprehensive trust degree of node n j is greater than the set security threshold, then the node is trustworthy, and the scout bee is used to collect and record data and perform trust calculation. Conversely, node n jIf the comprehensive trust value is lower than the threshold, it will be added to the blacklist and broadcast to ensure network security. At the same time, the evaluation node will enter a new cycle to request trust evaluation of other target nodes.

[0051] Beneficial effects brought by adopting the above technical solutions:

[0052] (1) The present invention proposes a WSN security routing protocol DFTSRP based on dynamic fuzzy trust and ABC algorithm. Considering the limited network energy, the base station adopts a reactive trust mechanism. Only when there is malicious activity in the network will the trust evaluation be enabled, effectively reducing unnecessary energy consumption.

[0053] (2) A sliding time window mechanism is introduced in the trust value calculation to ensure the dynamics and timeliness of the trust value. By introducing a forwarding delay judgment mechanism, serious security problems in the network can be quickly identified. At the same time, the fuzzy comprehensive evaluation method is used to calculate the indirect trust value, enhancing the comprehensiveness and accuracy of the trust evaluation. Finally, the artificial bee colony algorithm is used to detect and isolate dishonest recommendation nodes, thereby further improving the accuracy of the trust evaluation and the overall security of the network. Description of the Drawings

[0054] Figure 1 is the sliding time window diagram of the present invention;

[0055] Figure 2 is the comprehensive trust evaluation model diagram of the present invention; Detailed Embodiments

[0056] The technical solutions of the present invention will be described in detail below with reference to the accompanying drawings.

[0057] A WSN topology control method based on dynamic fuzzy trust and artificial bee colony algorithm includes the following steps:

[0058] Step 1: Introduce a sliding time window in the calculation of the trust value, update the trust value of the node periodically, and ensure the timeliness and dynamics of the trust value between nodes. And add a forwarding delay judgment;

[0059] Step 2: Use fuzzy comprehensive evaluation to calculate the indirect trust degree, taking into account multi-dimensional factors such as the data attributes and physical attributes of the target node in the trust calculation, and improving the accuracy of the trust evaluation;

[0060] Step 3: Use the ABC algorithm to detect and isolate dishonest recommendation nodes to obtain the comprehensive trust value. It improves the credibility of the malicious node recognition rate and has a low false positive rate. The comprehensive trust evaluation model is as Figure 2 shown;

[0061] In this embodiment, the following preferred solutions can be adopted to implement the above Step 1:

[0062] 101. Forwarding probability F i,j (t): Based on the above trust metric indicators, a probability-based trust model is adopted here to calculate the forwarding probability of the target node recorded by the evaluation node. That is, within a communication cycle, using the Beta distribution as the calculation model, represented by the successful forwarding ratio, the forwarding probability calculation formula is as follows:

[0063]

[0064] Among them, F i,j (t) represents the forwarding probability of the target node N j , s i,j is the number of successful forwards of node N j , f i,j is the number of failed forwards of N j ;

[0065] 102. Historical trust degree HT i,j (t): The trust value between nodes has significant timeliness and dynamics, and its current state will change continuously due to the actual performance of the nodes. Since the network environment and node behavior may fluctuate over time, relying solely on historical trust values cannot fully reflect the current situation of the nodes or predict their future behavior. The sliding time window dynamically updates the trust value of the nodes by limiting the effective range of historical data. Under this mechanism, only the node interaction records within the sliding time window are retained as the effective data for calculating the trust value, while the historical records outside the window range are discarded. The sliding time window consists of multiple time slots, each corresponding to a fixed time period for recording the trust value of the node within that time period. As Figure 1 shown, it can be seen from the figure that the right time slot stores the historical trust degree closer to the current time point. During the update process of the historical trust value, the older the record, the smaller its contribution should be. In this section, a time decay function is adopted to determine the historical trust degree weight, and the formula is as follows:

[0066]

[0067] Among them, λ is the decay factor, with a value between 0 and 1, and t k is the k-th time slot of the window.

[0068] Assume that the historical trust degree sequence stored in each time slot within the sliding time window at the current moment is Then the historical trust degree calculation formula is as follows:

[0069]

[0070] 103. Direct trust degree DT i,j(t): The evaluation node establishes the trust relationship between nodes based on the forwarding probability and forwarding delay of the target node. However, the complexity and dynamic changes of the network environment may lead to fluctuations in node behavior, making a single piece of current behavior data insufficient to comprehensively evaluate the true trustworthiness of the node. Therefore, historical trust values are introduced into the direct trust mechanism. The calculation formula is as follows:

[0071] DT i,j (t) = (w1 * F i,j (t) + w2 * HT i,j (t)) × FD#(1 - 4)

[0072] Among them, FD is the forwarding delay, which is used to measure the performance of the node in the data forwarding process. When the forwarding delay time exceeds the preset security threshold, it is considered that the node has serious security risks. The value of FD is directly set to 0, indicating that the node behaves abnormally in terms of forwarding delay. On the contrary, the value of FD is set to 1. If the forwarding delay of the node is determined to be abnormal (FD value is 0), its direct trust value will be directly reduced to 0. The direct trust value of the node is obtained based on the weighted algorithm of the forwarding probability and the historical trust value, and w1 + w2 = 1, which can be flexibly adjusted. Here, we take w1 = w2 = 0.5.

[0073] In this embodiment, the following preferred solution can be adopted to implement the above step 2:

[0074] 201. Determine the factor set U: U = {u1, u2, u3}, where u i (i = 1, 2, 3) represents the i-th factor in the factor set. u1 is the energy factor, which represents the remaining energy of the target node. u2 is the transmission rate factor, which represents the rate at which the node sends data packets within a period of time. u3 is the neighbor factor, which represents the number of neighbors within one hop of the target node;

[0075] 202. Determine the evaluation set V: V = {v1, v2, v3}, where v j j = 1, 2, 3) represents the j-th evaluation level. v1 represents untrusted, v2 represents generally trusted, and v3 represents highly trusted;

[0076] 203. Construct the fuzzy judgment matrix R: The fuzzy judgment matrix R is the relationship matrix between the factor set U and the evaluation set V. Perform a single-factor judgment and evaluation on the factor u i to determine its membership degree to the evaluation subset v j (expressed as r ij ), and obtain a set of single-factor judgment factors r i = (r i1 , r i2 , r i3 ). The judgment factors of the three factors construct the fuzzy judgment matrix R.

[0077]

[0078] Among them, r ij represents the membership degree of the i-th factor to the j-th evaluation. For example, r 13 represents the membership degree of the energy factor in the highly credible fuzzy subset;

[0079] 204. Determine the weight set A: A = {a1, a2, a3}, where A represents the weight coefficients of the importance of each factor for the comprehensive evaluation, and a1 + a2 + a3 = 1. According to the three-scale method, each weight coefficient is calculated as a1 = 0.10, a2 = 0.65, a3 = 0.25;

[0080] 205. Perform fuzzy operation to obtain the result vector B: The fuzzy comprehensive evaluation results of each evaluation object are as follows:

[0081]

[0082] Among them, b j is the membership degree of the node to the evaluation subset v j ;

[0083] 206. Use the weighted average method to calculate the indirect trust degree IT j of the target node N i,j (t), and the calculation formula is as follows:

[0084]

[0085] Among them, DT i,j (t) is the direct trust degree of the evaluation node N i calculated for the target node N j according to formula (4-4);

[0086] 207. In the one-to-one trust evaluation, the finally calculated indirect trust degree of the target node is denoted as λ. If λ ≥ γ, it means the node is trustworthy; if λ ≤ ξ, it means the node is a malicious node, add this node to the blacklist and remove it from the network; when ξ < λ < γ, the node is regarded as a suspected node and needs further observation and evaluation. Among them, γ and ξ respectively represent the trust threshold and the malicious threshold, which are used to distinguish the behavior status of the node.

[0087] In this embodiment, the above step 3 can be implemented by adopting the following preferred scheme:

[0088] 301. Employed Bee Phase: In our algorithm, the food source corresponds to the target node, and each employed bee corresponds to a recommended node. The employed bees no longer select the food source for recommendation according to the greedy algorithm. Instead, all employed bees evaluate and recommend the food source. If the evaluation node B i wants to evaluate the trustworthiness of the target node n j , it will send a request to the current one-hop neighbors of the target node. After the neighbor node N k receives the request, they work as employed bees and send the recommendation information about the target node (food source) to the evaluation node. Among them, the recommendation information includes its identification ID, direct trust, and indirect trust;

[0089] 302. Fitness Function: The evaluation node B i merges the received recommendation information, and then calculates the deviation function of the direct trust of these recommendations to detect false information. The recommendation deviation k of the recommendation node N j about the target node n is calculated as follows:

[0090]

[0091] Among them, can be regarded as the fitness function of the recommendation deviation, and m is the number of recommendation nodes.

[0092] For the weighted linear fitness function fit k of the recommendation node N k the formula is as follows:

[0093]

[0094] 303. Scout Bee Phase: The evaluation node classifies these recommendation nodes according to the probability value p k . If p k is less than the default threshold, the recommendation node N k is regarded as a dishonest node, and then the evaluation node adds it to the blacklist. The nodes on the blacklist cannot interact and transfer data with other nodes, that is, they are regarded as dead nodes. The calculation formula of the probability value p k is as follows:

[0095]

[0096] The evaluation node B i merges the recommendation information of the recommendation nodes not on the blacklist to obtain the comprehensive trustworthiness j of the target node n Its calculation formula is as follows:

[0097]

[0098] Among them, it means that the recommended nodes with smaller deviation values have greater reference weights, and y is the number of honest recommended nodes;

[0099] 304. Scout bee stage: When the comprehensive trust degree of node n j is greater than the set security threshold, then the node is trustworthy, and it collects and records data through scout bees and conducts trust calculation. On the contrary, if the comprehensive trust degree of node n j is lower than the threshold, it will be added to the blacklist and broadcast to ensure network security. At the same time, the evaluation node will enter a new cycle to request the trust evaluation of other target nodes.

Claims

1. A WSN topology control method based on dynamic fuzzy trust and artificial bee colony algorithm, characterized in that: The following steps are involved: (1) A sliding time window is introduced in the calculation of trust values ​​to periodically update the trust values ​​of nodes to ensure the timeliness and dynamics of the trust values ​​between nodes. A forwarding delay judgment is also added to quickly discover serious security issues. The specific steps for calculating direct trust are as follows: (101) Forwarding probability F i,j (t): Based on the above trust metrics, a probability-based trust model is used here to calculate the forwarding probability of the target node recorded by the evaluation node. That is, within a communication cycle, the Beta distribution is used as the calculation model and expressed as the successful forwarding ratio. The forwarding probability calculation formula is as follows: Among them, F i,j (t) represents the target node N j The forwarding probability, s i,j For node N j Number of successful forwardings, f i,j N j Number of forwarding failures; (102) Historical Trust HT i,j (t): The trust value between nodes has significant timeliness and dynamics, and its current state will continue to change due to the actual performance of the node. Since the network environment and node behavior may fluctuate over time, relying solely on historical trust values ​​cannot fully reflect the current status of the node or predict its future behavior. The sliding time window dynamically updates the trust value of the node by limiting the effective range of historical data. Under this mechanism, only the node interaction records within the sliding time window are retained as valid data for calculating the trust value, and historical records outside the window range will be discarded. The sliding time window consists of multiple time slots, each of which corresponds to a fixed time period, which is used to record the trust value of the node within the time period. In the process of updating the historical trust value, the older the record, the smaller the contribution should be. This section uses a time decay function to determine the historical trust weight, and the formula is as follows: Among them, λ is the attenuation factor, which ranges from 0 to 1, t k is the kth time slot of the window. Assume that the historical trust sequence stored in each time slot in the sliding time window at the current moment is The historical trust calculation formula is as follows: (103) Direct Trust DT i,j (t): The evaluation node establishes a trust relationship between nodes based on the forwarding probability and forwarding delay of the target node. However, the complexity and dynamic changes of the network environment may cause fluctuations in node behavior, which makes the single current behavior data insufficient to fully evaluate the true trustworthiness of the node. For this reason, the historical trust value is introduced into the direct trust mechanism. The calculation formula is as follows: DT i,j (t)=(w1*F i,j (t)+w2*HT i,j (t))×FD# (1-4) Among them, FD is the forwarding delay, which is used to measure the performance of the node in the data forwarding process. When the forwarding delay time exceeds the preset safety threshold, the node is considered to have serious security risks. The value of FD is directly set to 0, indicating that the node performs abnormally in terms of forwarding delay. On the contrary, the value of FD is set to 1. If the forwarding delay of the node is judged to be abnormal (FD value is 0), its direct trust value will be directly reduced to 0. The direct trust value of the node is obtained by the forwarding probability and the historical trust value based on a weighted algorithm, w1+w2=1, which can be flexibly adjusted. Here we take w1=w2=0.5; (2) Fuzzy comprehensive evaluation is used to calculate indirect trust, taking into account multi-dimensional factors such as the data attributes and physical attributes of the target node in the trust calculation to improve the accuracy of trust evaluation. The specific steps of fuzzy comprehensive evaluation are as follows: (201) Determine the factor set U: U = {u1,u2,u3}, where u i (i=1,2,3) represents the i-th factor in the factor set, u1 is the energy factor, which represents the remaining energy of the target node, u2 is the transmission rate factor, which represents the rate at which the node sends data packets within a period of time, and u3 is the neighbor factor, which represents the number of neighbors within one hop of the target node; (202) Determine the evaluation set V: V = {v1, v2, v3}, where v j j=1,2,3) represents the jth evaluation level. v1 represents untrustworthy, v2 represents generally trustworthy, and v3 represents highly trustworthy; (203) Construct the fuzzy judgment matrix R: The fuzzy judgment matrix R is the relationship matrix between the factor set U and the evaluation set V. i Perform single factor judgment evaluation to determine its effect on the evaluation subset v j The membership degree (denoted as r ij ), and obtain a set of single factor judgment factors r i =(r i1 ,r i2 ,r i3 ). The judgment factors of the three factors construct the fuzzy judgment matrix R. Among them, r ij represents the membership of the i-th factor to the j-th evaluation. For example, r 13 Indicates the membership of the energy factor in the highly credible fuzzy subset; (204) Determine the weight set A: A = {a1, a2, a3}, where A represents the weight coefficient of each factor for the comprehensive evaluation, a1 + a2 + a3 = 1. According to the three-scale method, each weight coefficient is calculated to be a1 = 0.10, a2 = 0.65, a3 = 0.25; (205) Fuzzy operation is performed to obtain the result vector B: The fuzzy comprehensive evaluation results of each evaluation object are as follows: Among them, b j For node pair evaluation subset v j The degree of membership; (206) Calculate the target node N using the weighted average method j Indirect trust in IT i,j (t), the calculation formula is as follows: Among them, DT i,j (t) is the evaluation node N i The target node N calculated according to formula (4-4) j Direct trust in (207) In the one-to-one trust evaluation, the indirect trust of the target node is finally calculated and recorded as λ. If λ≥γ, it means that the node is trustworthy; if λ≤ξ, it means that the node is a malicious node, and the node is added to the blacklist and removed from the network; and when ξ<λ<γ, the node is regarded as a suspect node and needs further observation and evaluation. Among them, γ and ξ represent the trust threshold and malicious threshold, respectively, which are used to distinguish the behavior status of the node; (3) The ABC algorithm is used to detect and isolate dishonest recommendation nodes to obtain comprehensive trust. The credibility of the malicious node identification rate is improved, and the misjudgment rate is lower. The specific process of comprehensive trust evaluation is as follows: (301) Hired bee stage: In our algorithm, the food source corresponds to the target node, and each hired bee corresponds to a recommended node. In addition, the hired bees no longer select food sources for recommendation based on the greedy algorithm. Instead, all hired bees evaluate and recommend food sources. If the evaluation node B i Want to evaluate the target node n j If the trust level of the target node is higher, it will send a request to the current one-hop neighbor of the target node, the neighbor node N k After receiving the request, they work as hired bees and send recommendation information about the target node (food source) to the evaluation node. The recommendation information includes its identification ID, direct trust, and indirect trust; (302) Fitness function: Evaluate node B i Merge the received recommendation information, and then calculate the deviation function of the direct trust of these recommendations to detect false information, recommending node N k About the target node n j Recommended deviation The calculation formula is as follows: in, It can be regarded as the fitness function of the recommendation deviation, and m is the number of recommended nodes. For the recommended node N k The weighted linear fitness function fit k The formula is as follows: (303) Observation bee stage: evaluate the node according to the probability value p k Classify these recommended nodes. If p k If it is less than the default threshold, the recommended node N k The node is considered as a dishonest node, and then the evaluation node adds it to the blacklist. The blacklisted node cannot interact with other nodes and transmit data, that is, it is considered a dead node. The probability value p k The calculation formula is as follows: Evaluate Node B i Merge the recommendation information of the recommended nodes that are not on the blacklist to get the target node n j The overall trust The calculation formula is as follows: in, It means that the recommended nodes with smaller deviation values ​​have greater reference weights, and y is the number of honest recommended nodes; (304) Scouting Bee Phase: When node n j When the comprehensive trust of node n is greater than the set security threshold, the node is trustworthy and collects and records data through scout bees and performs trust calculation. j If the comprehensive trust level is lower than the threshold, it will be added to the blacklist and broadcast to ensure network security. At the same time, the evaluation node will enter a new cycle and request trust evaluation for other target nodes.

Images