A WSN Topology Control Method Based on Grey Clustering and Dynamic Slicing

Abstract

The present invention discloses a WSN topology control method based on grey clustering and dynamic slicing, which is used to improve the security of data during transmission. The method includes: evaluating the trust values of adjacent cluster head nodes in a way of inter-cluster security awareness, and the cluster head nodes adopt dynamic slicing technology to slice data packets and send them to candidate cluster head nodes. The present invention makes certain improvements in data slicing. It uses a clustering algorithm to divide the security states of each candidate cluster head, and defines the number and proportion of data slices of the cluster head nodes according to the security states and location information of the candidate cluster heads, reducing the threat of malicious nodes to data during network communication and also controlling the energy consumption at a relatively low level.

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 grey clustering and dynamic slicing. Background Art

[0002] A wireless sensor network is a distributed sensor network in which sensor nodes communicate with each other wirelessly. A large number of sensor nodes are usually deployed in the network to receive environmental data, form an ad-hoc wireless network, and route the sensed data to the sink node in a single-hop or multi-hop manner, and finally the sink node transmits it to the user. The wireless sensor network originated in the military field. With the in-depth research, its application fields have gradually expanded to multiple fields such as ecology, healthcare, and smart cities.

[0003] Regarding the security and energy consumption problems in wireless sensor network topology control, domestic and foreign researchers have proposed many improvement methods. The EECS protocol improves the algorithm of randomly selecting cluster heads in the LEACH protocol, selects cluster heads based on weighted parameters, and the clustering result is more reasonable, improving the energy utilization efficiency of the network. The PEGASIS protocol establishes a chain-like topology, stipulating that all sensor nodes need to exchange data with the sink node, balancing the network energy consumption and avoiding the emergence of hot spots in the network. W. Wang et al. designed a trust-based secure topology model LEACH-TM, and established a reliable clustering structure by analyzing the trust status and remaining energy of nodes, and selected the optimal topology by calculating the average trust value of LEACH-TM. This method reduces the difficulty of collecting and processing trust information to a certain extent, but the author did not consider the situation where malicious nodes communicate with each other, which is prone to such attacks. R. Roman et al. proposed a method for sensor nodes to use the situation awareness mechanism to determine whether there are abnormal events around them, and applied the concept of network situation awareness to sensor nodes. This method takes the situation awareness mechanism as a basic service and embeds it in the intrusion detection system to achieve security monitoring of the network. He et al. proposed a method for sensor nodes to use multivariate time series association rules to analyze the received data to achieve the security awareness of the surrounding environment of the nodes. A. Mehmood et al. placed the task of network situation awareness at the sink node, detected node intrusion based on the context awareness method, and reduced the storage pressure of the cluster head nodes by storing the knowledge base at the sink node, extending the network life cycle.

[0004] In summary, designing a secure and efficient topology control method has become a research hotspot in wireless sensor networks in recent years. How to balance energy consumption, extend the network survival period, and take into account the security of nodes has become an urgent problem to be solved in this field. Summary of the Invention

[0005] To solve the technical problems mentioned in the above background art, the present invention proposes a WSN topology control method based on grey clustering and dynamic slicing.

[0006] To achieve the above technical purpose, the technical solution of the present invention is as follows:

[0007] A WSN topology control method based on grey clustering and dynamic slicing, comprising the following steps:

[0008] (1) Adopt the grey relational analysis method. According to the characteristic that the sensing data of sensor nodes has spatial correlation, dynamically cluster in the network, and the cluster head nodes are selected by combining the distance vector method and the polynomial fitting method.

[0009] (2) The protocol divides the network into several independent cluster regions, records the relevant attributes of each cluster region, including the number of nodes in the cluster region, the area of the cluster region, the energy of each node, and the sensing data of each node, and judges the security state of the cluster region through the grey clustering analysis method.

[0010] (3) The nodes in the cluster send the data to their respective cluster head nodes. Considering the key role of the cluster head nodes in data forwarding, process the data slicing of the cluster head nodes, and then send the data to the candidate cluster heads.

[0011] Further, in step (1), the method of dynamic clustering and selecting cluster heads is as follows:

[0012] (101) According to the distribution and coordinates of the nodes, determine the neighbor node situation of each node, calculate the density ρ of each node, and then calculate the correlation degree λ between each node and the surrounding nodes based on the received sensing data i ;

[0013] (102) Combine the energy and distance factors of the nodes to calculate the decision value of the nodes. All nodes are sorted in descending order of the decision value, and the largest M nodes are selected as the peak nodes.

[0014] (103) For the other nodes that are not selected as peak nodes, calculate the Euclidean distance from all peak nodes, and select the peak node with the closest distance to join the cluster. All nodes repeat the execution to ensure that all nodes join the cluster.

[0015] Further, in step (2), the judgment of the security state of the cluster region is as follows:

[0016] (201) According to the energy values of the nodes in the candidate cluster, calculate the ratio r between the energy value of the candidate cluster head node and the average energy of the nodes in the candidate cluster;

[0017] (202) According to the sensing data of the nodes in the candidate cluster, calculate the gap d between the sensing data of the candidate cluster head node and the average data of the nodes in the candidate cluster;

[0018] (203) The aggregation node calculates the average number of nodes within a cluster and the average cluster area of the network according to the distribution state of the nodes in the network. The cluster head node obtains the average number of nodes within a cluster and the average cluster area, and then equalizes the number of nodes num and the area area within each candidate cluster;

[0019] (204) Design the grey classes of the candidate clusters into three security classes, namely "sufficiently secure and suitable as the next-hop forwarding area", "moderately secure and moderately suitable as the next-hop forwarding area", and "unsafe and not suitable as the next-hop forwarding area";

[0020] (205) For r, d, num, and area obtained in (201) to (203), use them as evaluation indicators of the grey clustering analysis method, and design the possibility functions based on the degree to which each evaluation indicator belongs to each grey class;

[0021] (206) According to the possibility functions in (205), calculate the clustering coefficients of the candidate cluster areas belonging to each grey class, and the grey class with the largest coefficient is its final belonging grey class;

[0022] (207) According to the remaining energy E of the candidate cluster head res , the distance d from the aggregation node s and the included angle θ between the line connecting the node and the candidate cluster head and the line connecting the node and the aggregation node, calculate the decision value f, and the cluster head with the largest decision value is used as the ideal next-hop node;

[0023] (208) The cluster head node excludes the clusters that are not suitable as the next-hop forwarding area. Based on its own energy consumption considerations, it selects the cluster heads in the remaining candidate clusters whose distances from itself are less than or equal to the distance from the ideal next-hop node to form a set of next-hop nodes and enters the data slicing stage.

[0024] Further, in step (3), the data slicing processing method for the cluster head node is as follows:

[0025] (301) Drawing on the idea of the D-SMART protocol, select to slice the data of the cluster head node, and the nodes within the cluster directly send the data to their respective cluster head nodes;

[0026] (302) Let the set of next-hop nodes CH = {x i | i = 0, 1, …, n}, where x i represents the next-hop cluster head node, and the number of next-hop nodes is n + 1. Let the grey class to which the next-hop node belongs be GC = {c j | j = 0, 1, 2}, then for each next-hop node, the calculation method of the slice ratio it receives is Equation (3-1):

[0027] ​

[0028] where c i The larger the value, the higher the security of the node.

[0029] (303) The cluster head node transmits the data slices after data fusion according to the reception slice ratios of the nodes in its next-hop node set.

[0030] Beneficial effects brought by adopting the above technical solutions:

[0031] (1) The present invention proposes to slice the data sent by sensor nodes. When a malicious node launches a selective forwarding attack, if the sent data is not sliced, the malicious node will discard all the received data, resulting in data loss. After the data sent by the node is sliced, the malicious node only has part of the sliced data. Even if the node loses all of it, the resulting loss will be smaller. Therefore, the packet transmission rate and the security of the network are improved.

[0032] (2) Based on single-hop between clusters, the present invention adopts the grey clustering analysis method. According to the attribute information such as the number of nodes, node energy, area of the region, and node data in each cluster region, it analyzes the security of each cluster region, classifies the security categories, and considers energy consumption and security categories when selecting the next hop for data forwarding to obtain an optimal solution, improving the packet transmission rate of network data while reducing network energy consumption. Description of the Drawings

[0033] Figure 1 is a schematic diagram of the selection of the next hop of the cluster head in the present invention;

[0034] Figure 2 is a schematic diagram of data slice transmission in the present invention; Detailed Embodiment

[0035] The technical solution of the present invention will be described in detail below with reference to the drawings.

[0036] A WSN topology control method based on grey clustering and dynamic slicing includes the following steps:

[0037] (1) Adopt the grey correlation analysis method. According to the characteristic that the sensing data of sensor nodes has spatial correlation, dynamically cluster in the network, and the cluster head node is selected by combining the distance vector method and the polynomial fitting method.

[0038] (2) The protocol divides the network into several independent cluster regions, records the relevant attributes of each cluster region, including the number of nodes in the cluster region, the area of the cluster region, the energy of each node, etc., and judges the security state of the cluster region through the grey clustering analysis method.

[0039] (3) The nodes within the cluster send the data to their respective cluster heads. Considering the key role of the cluster heads in data forwarding, the data of the cluster heads is sliced and then sent to the candidate cluster heads.

[0040] In this embodiment, the following preferred solution can be adopted to implement the above (1):

[0041] (101) According to the distribution and coordinates of the nodes, determine the neighbor node situation of each node, calculate the density ρ of each node, and then based on the received sensing data, calculate the association degree λ between each node and the surrounding nodes. i ;

[0042] (102) Combining the energy and distance factors of the nodes, calculate the decision value of the nodes. All nodes are sorted in descending order according to the decision value, and the largest M nodes are selected as the peak nodes.

[0043] (103) For the other nodes that are not elected as peak nodes, calculate the Euclidean distance from all peak nodes, and select the peak node with the closest distance to join the cluster. All nodes repeat the execution to ensure that all nodes join the cluster.

[0044] In this embodiment, the following preferred solution can be adopted to implement the above (2):

[0045] (201) According to the energy values of the nodes within the candidate cluster, calculate the ratio r between the energy value of the candidate cluster head node and the average energy of the nodes within the candidate cluster.

[0046] (202) According to the sensing data of the nodes within the candidate cluster, calculate the gap d between the sensing data of the candidate cluster head node and the average data of the nodes within the candidate cluster.

[0047] Since the data packets sent by each node contain data for several time periods, for each attribute, calculate the average value of the data sensed by the node over all time periods, and fuse the data of each node into a set of numerical values.

[0048] The cluster head node calculates the distance d between its own data and the data of the nodes within the cluster using the distance vector method with equation (2-1). ij

[0049]

[0050] Wherein, represents the vector pointing to node x i from node x k ; node x k is the reference node, represents the modulus of the vector .

[0051] The cluster head node establishes an original data table based on the sensed data of the nodes within the cluster. Calculate the correlation degree between each attribute, obtain one or more groups of attributes with a relatively large correlation degree, and for the obtained attribute combinations, use the method of polynomial fitting to calculate the mutual dependence between attributes to obtain the fitting function.

[0052] Calculate the prediction factor pd of each node i , that is, whether its data change conforms to the prediction of the fitting function, and the calculation formula is as follows:

[0053]

[0054] Combine the distance d ij and the remaining energy E of the node i , judge the suspicion coefficient μ of the node i , and update the trust value t of the node i .

[0055]

[0056] Among them, mid(d ij ) represents the median of the distance factor of the nodes within the cluster, represents the average remaining energy of the nodes within the cluster, τ, ω is a coefficient, satisfying δ is the threshold of the suspicion coefficient.

[0057] (203) The sink node calculates the average number of nodes within the cluster and the average cluster area in the network according to the distribution state of the nodes in the network. The cluster head node obtains the average number of nodes within the cluster and the average cluster area, and then equalizes the number of nodes num and the area area of each candidate cluster;

[0058] (204) Design the gray classes of the candidate clusters into three security classes, namely "sufficiently secure and suitable as the next-hop forwarding area", "generally secure and generally suitable as the next-hop forwarding area", and "unsafe and not suitable as the next-hop forwarding area";

[0059] (205) Use r, d, num, and area obtained through (201) to (203) as the evaluation indicators of the gray clustering analysis method, and design the possibility function according to which each evaluation indicator belongs to each gray class;

[0060] (206) According to the possibility function in (205), calculate the clustering coefficient of the candidate cluster area belonging to each gray class, and the gray class with the largest coefficient is its final belonging gray class;

[0061] (207) According to the remaining energy E of the candidate cluster head res and the distance d from the sink node sAnd the included angle θ between the connection line of the node and the candidate cluster head and the connection line of the node and the sink node is calculated to obtain the decision value f. The cluster head with the largest decision value is used as the ideal next-hop node. The calculation formula is as shown in (2-5):

[0062]

[0063] The cluster head node excludes the clusters that are not suitable as the next-hop forwarding area. Considering its own energy consumption, it selects the cluster heads in the remaining candidate clusters whose distance from itself is less than or equal to the distance from the ideal next-hop node to form the next-hop node set and enters the data slicing stage. As Figure 1 shown, for node CH0, the ideal node is CH2, and CH1 is the candidate cluster head with a closer distance.

[0064] In this embodiment, the following preferred solution can be adopted to implement the above (3):

[0065] (301) Drawing on the idea of the D-SMART protocol, the data of the cluster head node is selected for slicing. The nodes within the cluster directly send the data to their respective cluster head nodes. The cluster head slicing scheme is as Figure 2 shown:

[0066] Node CH0 is the current cluster head node. Nodes CH1-CH4 are within the communication range of CH0. Therefore, nodes CH1 to CH4 are the neighbor nodes of node CH0;

[0067] Node CH0 classifies nodes CH1-CH4 based on the clustering analysis method. Assume that the categories to which nodes CH1-CH4 belong are "sufficiently secure and suitable as the next-hop forwarding area", "sufficiently secure and suitable as the next-hop forwarding area", and "generally secure and generally suitable as the next-hop forwarding area", and "unsafe and not suitable as the next-hop forwarding area". Node CH4 exits the selection process of candidate nodes;

[0068] Assume that node CH2 is the ideal node. The distances between CH1 and CH3 and CH0 are both less than the distance between CH2 and CH0. Therefore, the next-hop candidate set is the set composed of CH1, CH2, and CH3;

[0069] According to the clustering results, the gray classes to which nodes CH1, CH2, and CH3 belong are c1 = 2, c2 = 2, c3 = 1;

[0070] The slicing ratios of nodes CH1, CH2, and CH3 in the next-hop candidate set are p1 = 0.4, p2 = 0.4, p3 = 0.2.

[0071] (302) The next-hop node set CH = {x i |i = 0, 1, …, n}, where x iIndicates the next-hop cluster head node. The number of next-hop nodes is n + 1. Let the gray class to which the next-hop nodes belong be GC = {c j |j = 0, 1, 2}. Then, for each next-hop node, the calculation method of the received slice ratio is as follows:

[0072]

[0073] where the larger the value of c i , the higher the security level of the node.

[0074] (303) The cluster head node transmits the data slices that have undergone data fusion according to the received slice ratios of the nodes in its next-hop node set.

Claims

1. A WSN topology control method based on grey clustering and dynamic slicing, characterized in that Including the following steps: (1) Adopt the grey relational analysis method. According to the characteristic that the sensing data of sensor nodes has spatial correlation, dynamically cluster in the network. The selection of cluster head nodes combines the distance vector method and the polynomial fitting method. The specific steps are as follows: (101) Determine the neighbor node situation of each node according to the distribution and coordinates of the nodes, calculate the density ρ of each node, and then calculate the association degree λ between each node and its surrounding nodes based on the received sensing data. i ; (102) Combine the energy and distance factors of the nodes to calculate the decision value of the nodes. All nodes are sorted in descending order according to the decision value, and the largest M nodes are selected as the peak nodes. (103) For other nodes that are not elected as peak nodes, calculate the Euclidean distance from all peak nodes, and select the peak node with the closest distance to join the cluster. All nodes repeat this operation to ensure that all nodes join the cluster. (2) Divide the network into several independent cluster regions, record the relevant attributes of each cluster region, including the number of nodes in the cluster region, the area of the cluster region, the energy of each node, and the sensing data of each node. Judge the security status of the cluster region through the grey clustering analysis method. The specific steps are as follows: (201) According to the energy values of the nodes in the candidate cluster, calculate the ratio r between the energy value of the candidate cluster head node and the average energy of the nodes in the candidate cluster. (202) According to the sensing data of the nodes in the candidate cluster, calculate the difference d between the sensing data of the candidate cluster head node and the average data of the nodes in the candidate cluster. Since the data packets sent by each node contain data for several time periods, for each attribute, calculate the average value of the data sensed by the node in all time periods, and fuse the data of each node into a set of values. The cluster head node calculates the distance d between its own data and the data of the nodes within the cluster using the distance vector method with Equation (2-1). ij Among them, represents node x i points to node x k 's vector, and node x k is the reference node. represents the modulus of the vector ; The cluster head node establishes an original data table based on the sensing data of the nodes in the cluster, calculates the correlation degree between each attribute, obtains one or more groups of attributes with a larger correlation degree. For the obtained attribute combination, adopt the polynomial fitting method to calculate the mutual dependence between the attributes, and obtain the fitting function. Calculate the prediction factor pd of each node i , that is, whether its data change conforms to the prediction of the fitting function. The calculation formula is as follows: Combined distance d ij and the remaining energy E of the node i , judge the suspicion coefficient μ of the node i , and update the trust value t of the node i ; Among them, mid(d ij ) represents the median of the distance factor of the nodes within the cluster, represents the average remaining energy of the nodes within the cluster, τ, ω is a coefficient that satisfies δ is the threshold of the suspicion coefficient; (203) The sink node calculates the average number of nodes in the cluster and the average area of the cluster region according to the distribution state of the nodes in the network. The cluster head node obtains the average number of nodes in the cluster and the average area of the cluster region, and then performs a mean value processing on the number of nodes num and the area area in each candidate cluster. (204) Design the grey classes of the candidate clusters into three security classes, namely "sufficiently secure and suitable as the next-hop forwarding area", "generally secure and generally suitable as the next-hop forwarding area", and "unsafe and not suitable as the next-hop forwarding area". (205) For r, d, num, and area obtained in (201) to (203), use them as evaluation indicators of the grey clustering analysis method, and design the possibility function based on the degree to which each evaluation indicator belongs to each grey class. (206) According to the possibility function in (205), calculate the clustering coefficient of the candidate cluster region belonging to each grey class. The grey class with the largest coefficient is its final belonging grey class. (207) Calculate the decision value f based on the remaining energy E of the candidate cluster head res , the distance d from the sink node s and the angle θ between the line connecting the node and the candidate cluster head and the line connecting the node and the sink node. The cluster head with the largest decision value is used as the ideal next-hop node. The calculation formula is as shown in (2-5): (208) The cluster head node excludes the clusters that are not suitable as the next-hop forwarding area. Considering its own energy consumption, select the cluster head whose distance from itself is less than or equal to the distance from the ideal next-hop node among the remaining candidate clusters to form the next-hop node set and enter the data slicing stage. (3) The nodes in the cluster send the data to their respective cluster head nodes. Considering the key role of the cluster head node in data forwarding, perform data slicing processing on the data of the cluster head node, and then send the data to the candidate cluster head. The data slicing processing method for the cluster head node is as follows: (301) Denote the set of next-hop nodes as CH = {x i | i = 0, 1, …, n}, where x i represents the next-hop cluster head node, and the number of next-hop nodes is n + 1. Let the gray class to which the next-hop nodes belong be GC = {c j | j = 0, 1, 2}. Then, for each next-hop node, the calculation method of the received slice ratio is shown in Equation (3-1): where c i The larger the value is, the higher the security of the node is; (302) The cluster head node transmits the data slices that have undergone data fusion based on the reception slice ratios of the nodes in its next-hop node set.

Images