Security channel networking method based on edge computing power

Abstract

The invention provides a secure channel networking method based on edge computing power, and the method comprises the steps: obtaining the security parameters of a link between nodes in a network topology structure based on a selected networking strategy, including link encryption strength, an authentication mechanism and a key updating period, and combining the bandwidth and time delay of the link, obtaining an optimal secure communication route between the nodes through an edge calculation method; issuing the optimized node communication route to each edge computing node, and dynamically adjusting an authentication mechanism and a key management strategy of a secure channel between the nodes according to link security parameters including link encryption strength, a delay jitter threshold and a packet loss rate threshold; and according to the updated networking strategy, dynamically optimizing node roles and link redundancy, updating a network topology structure, and dividing edge nodes into three node function roles of computing nodes, storage nodes and forwarding nodes in combination with node computing capability, storage capacity and energy consumption level.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a secure channel networking method based on edge computing power. Background Art

[0002] Changes in network conditions will affect the connectivity and stability between nodes, which in turn affects the reliability of secure channels. Edge computing nodes need to have the ability to adaptively network and dynamically adjust the networking mode and parameters according to changes in network conditions to maintain network connectivity and stability. In actual heterogeneous network environments, the network conditions of edge computing nodes vary greatly. Some nodes are in an environment with good network conditions, while others are in an environment with poor network conditions. To achieve adaptive networking, edge computing nodes need to be able to accurately perceive changes in network conditions and dynamically adjust networking modes and parameters based on the perception results. How to maintain network stability while being able to respond to changes in network conditions in a timely manner is another difficult problem that adaptive networking technology needs to solve. Summary of the invention

[0003] The present invention provides a secure channel networking method based on edge computing power, which mainly includes:

[0004] Obtain the multi-dimensional network status parameters of each edge computing node, including network topology, link quality and node load, and use the fuzzy comprehensive evaluation method to determine whether the current network status meets the preset stability threshold. If not, select the networking strategy with the highest matching degree with the current network status from the preset networking strategy library and combine the parameters of the networking strategy, including node communication protocol, link redundancy and node functional role;

[0005] Based on the selected networking strategy, the security parameters of the links between nodes in the network topology are obtained, including link encryption strength, authentication mechanism, and key update cycle. In combination with link bandwidth and latency, the optimal secure communication route between nodes is obtained through edge computing methods.

[0006] The optimized node communication route is sent to each edge computing node, and the authentication mechanism and key management strategy of the secure channel between nodes are dynamically adjusted according to the link security parameters, including link encryption strength, delay jitter threshold, and packet loss rate threshold;

[0007] Monitor network status changes and node communication quality in real time, analyze network status change trends and node communication quality scores, and start the adaptive networking process when the scores are lower than the preset threshold for multiple consecutive times, reselect the matching networking strategy to avoid network shocks caused by frequent adjustments;

[0008] According to the updated networking strategy, the node roles and link redundancy are dynamically optimized, and the network topology is updated. The edge nodes are divided into three node functional roles: computing nodes, storage nodes, and forwarding nodes, based on the node computing power, storage capacity, and energy consumption level.

[0009] Continuously monitor network conditions and node communication quality, smooth out network condition fluctuations by setting corresponding perception time windows, avoid network jitters caused by overly sensitive condition perception, respond to network changes while maintaining network stability, perform security scoring against preset thresholds, dynamically trigger network strategy adjustments based on scoring results, and respond to changes in the network environment while maintaining network stability.

[0010] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0011] The present invention discloses a secure channel networking method based on edge computing power. The method obtains multi-dimensional network parameters of edge nodes, uses fuzzy comprehensive evaluation to judge network stability, and selects the best networking strategy from a preset strategy library. Based on the selected strategy, combined with link security parameters, an edge computing method is used to optimize the secure communication routing between nodes. The present invention monitors the network status in real time. When the communication quality score is lower than the threshold, the adaptive networking process is started to reselect the strategy to avoid frequent adjustments causing network shocks. At the same time, according to factors such as node computing power, the edge nodes are dynamically divided into three roles: computing, storage, and forwarding, and the role attributes are adjusted according to load requirements. By setting a perception time window, network fluctuations are smoothed, and environmental changes are responded to while maintaining stability. The present invention realizes adaptive secure networking of edge computing networks, and improves the stability, security, and communication quality of the network. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 This is a flow chart of a secure channel networking method based on edge computing power of the present invention.

[0013] Figure 2 It is a schematic diagram of a secure channel networking method based on edge computing power of the present invention.

[0014] Figure 3 This is another schematic diagram of a secure channel networking method based on edge computing power of the present invention. DETAILED DESCRIPTION

[0015] The following will describe the technical solutions in the embodiments of the present invention in detail in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention.

[0016] like Figure 1-3In this embodiment, a secure channel networking method based on edge computing power may specifically include:

[0017] S101. Obtain multi-dimensional network status parameters of each edge computing node, including network topology, link quality and node load, and use fuzzy comprehensive evaluation method to determine whether the current network status meets the preset stability threshold. If not, select the networking strategy with the highest matching degree with the current network status from the preset networking strategy library and combine the parameters of the networking strategy, including node communication protocol, link redundancy and node functional role.

[0018] Obtain the network topology data, link quality parameters and real-time load information of the edge computing cluster, and construct the multidimensional parameter matrix; wherein the multidimensional parameter matrix reflects the current network status. Apply fuzzy comprehensive evaluation to the multidimensional parameter matrix; wherein the fuzzy comprehensive evaluation includes standardizing the network topology, link quality and node load parameters, establishing a fuzzy relationship matrix, and obtaining a comprehensive evaluation result through fuzzy synthesis operation. Compare the comprehensive evaluation result with the preset stability threshold; if the comprehensive evaluation result does not meet the preset stability threshold, extract all available strategies from the pre-established networking strategy library. Perform vector similarity calculation based on the available strategies and the current network status parameters; wherein the vector similarity calculation is used to select the strategy with the highest similarity as the optimal networking solution.

[0019] Specifically, the network topology data, link quality parameters and real-time load information of each node are obtained from the edge computing cluster to construct a multidimensional parameter matrix reflecting the current network status. Fuzzy comprehensive evaluation is applied to the parameter matrix. First, the parameters such as network topology, link quality and node load are standardized, and then a fuzzy relationship matrix is ​​established. The comprehensive evaluation result is obtained through fuzzy synthesis operation. The evaluation result is compared with the preset stability threshold to obtain the judgment result of whether the network status meets the stability requirements. The corresponding networking strategy is selected according to the judgment result. If the network status does not meet the stability requirements, all available strategies are extracted from the pre-established networking strategy library. The vector similarity calculation is performed between the current network status parameters and the applicable conditions of each strategy in the strategy library, and the strategy with the highest similarity is selected as the optimal networking solution. The key parameters of the strategy are extracted, including the type of inter-node communication protocol, network link redundancy configuration and the functional role definition of each node. The edge computing cluster is reconstructed according to the selected networking strategy. The inter-node communication protocol is adjusted according to the strategy parameters, such as switching the protocol to IPv6 or MQTT. The network topology structure is adjusted to increase redundant links and the number of backup links. Reassign the roles of computing nodes, storage nodes, and gateway nodes to balance node loads. Monitor the status of each node through regular sampling during the reconstruction process to ensure that the reconstruction operation will not cause network interruption. After the reconstruction is completed, re-collect network parameters to verify whether the new network status meets the stability requirements. Build a network status history database to record the process data of each network parameter collection, evaluation, and reconstruction. Use the greedy algorithm to mine historical data to identify key factors affecting network stability. Dynamically adjust the stability evaluation criteria based on the mining results and optimize the parameter configuration in the networking strategy library. Use the greedy algorithm to dynamically adjust the node load distribution according to the current network status and the selected strategy to achieve instant optimization. These measures improve the adaptive ability of network management. When obtaining the network topology data, link quality parameters, and real-time load information of each node from the edge computing cluster, the SNMP protocol can be used to collect data regularly. For example, collect indicators such as node CPU usage, memory usage, and network bandwidth every 5 minutes to construct a 100x10 parameter matrix. Apply fuzzy comprehensive evaluation to the parameter matrix. First, standardize each parameter to the 0-1 interval, then establish a 5x10 fuzzy relationship matrix, and set the weight vector to

[0020] [0.3, 0.3, 0.2, 0.1, 0.1], a comprehensive evaluation result of 0.75 is obtained through fuzzy synthesis operation. 0.75 is compared with the preset 0.8 stability threshold to determine that the current network status does not meet the stability requirements. The best solution is selected from the preset 10 networking strategies. The cosine similarity between the current network status vector and the applicable condition vector of each strategy is calculated to obtain a similarity list.

[0021] [0.65, 0.72, 0.58, 0.81, 0.70, 0.63, 0.69, 0.75, 0.67, 0.73]. The strategy No. 4 corresponding to the highest similarity of 0.81 is selected as the optimal networking solution. Extract the parameters of the strategy, the node communication protocol is MQTT, the number of redundant links is 2, and the roles are defined as 3 computing nodes, 2 storage nodes, and 1 gateway node. Reconstruct the edge computing cluster according to the selected strategy. Switch the communication protocol of the 6 nodes from HTTP to MQTT, and set the QoS level to 1. Add 2 redundant links on the basis of the original 1 main link to form a triangular topology. Reassign the node roles, expand the original 2 computing nodes to 3, expand 1 storage node to 2, and keep 1 gateway node unchanged. Use the polling algorithm to sample and monitor the status of each node every 30 seconds. If the sampling results are normal for 10 consecutive times, the reconstruction is considered successful. After reconstruction, collect network parameters again, and calculate the new comprehensive evaluation result of 0.85, which is higher than the stability threshold of 0.8, verifying that the reconstruction is effective. A network status history database containing 1,000 records is constructed. Each record contains 20 fields, covering network parameters, evaluation results, and reconstruction operations. The five key factors that most significantly affect stability are selected from the 20 fields using a feature selection algorithm based on information gain. Based on these five factors, the weight vector of the stability evaluation criteria is dynamically adjusted. For example, the original [0.3, 0.3, 0.2, 0.1, 0.1] is changed to

[0022] Adjusted to [0.25, 0.35, 0.2, 0.15, 0.05]. The greedy algorithm is used to dynamically balance the load of the six nodes, and each time a new task is assigned to the node with the lowest current load, achieving instant optimization. The adaptive ability of network management is improved, and the accuracy of stability evaluation is increased from 85% to 92%.

[0023] S102. Based on the selected networking strategy, the security parameters of the links between nodes in the network topology are obtained, including link encryption strength, authentication mechanism, and key update cycle. In combination with link bandwidth and latency, the optimal secure communication route between nodes is obtained through edge computing methods.

[0024] The network topology data in the network management system is obtained, and the network topology data includes the number of nodes, node types and connection relationships between nodes; the Nmap tool is used to scan the security parameters of the network link to obtain the link encryption algorithm type, key length, authentication protocol version and key update frequency; the bandwidth capacity and end-to-end communication delay of each link are measured by the i Perf measurement tool. According to the link security parameters, a security scoring model is deployed on the edge node; the security scoring model assigns weight values ​​to the encryption strength, authentication mechanism and key update cycle respectively; the security score of each link is calculated according to the weight value. The link bandwidth and delay are normalized to obtain the link performance score; the security score and performance score are synthesized according to a preset ratio to obtain a comprehensive score for each link. A network weighted graph is constructed based on the comprehensive link score, in which the node represents the network device, the edge represents the link, and the weight of the edge is the inverse of the comprehensive link score; the improved Dijkstra algorithm is used to search for the shortest path between any two nodes on the weighted graph to obtain the optimal routing table. The optimal routing table is distributed to each edge node, and a lightweight routing decision program is deployed on the edge node; the lightweight routing decision program performs routing selection according to the locally cached routing table and the real-time network status; if a network topology change or a significant link status change is detected, the routing table update process is triggered.

[0025] Specifically, according to the selected networking strategy, the network topology data, including the number of nodes, node types, and the connection relationship between nodes, is obtained from the network management system. The security parameters of each link are scanned by Nmap to obtain the link encryption algorithm type, key length, authentication protocol version, and key update frequency. i Perf is used to measure the bandwidth capacity and end-to-end communication delay of each link. All collected data are integrated into a link attribute matrix as the input for subsequent route optimization. Data collection programs are deployed on edge nodes to realize distributed data collection and reduce the burden on central nodes. For the security parameters in the link attribute matrix, a security scoring model is deployed on the edge nodes. The encryption strength is assigned a weight of 0.4, the authentication mechanism is assigned a weight of 0.3, and the key update cycle is assigned a weight of 0.3. The encryption strength is graded according to the key length, with 1 point for 128 bits and below, 2 points for 129-256 bits, and 3 points for 257 bits and above. The authentication mechanism is scored according to the security of the protocol, with 1 point for WEP, 2 points for WPA, and 3 points for WPA2. The key update cycle is scored according to the update frequency, with 1 point for more than 1 day, 2 points for 1 day to 1 hour, and 3 points for less than 1 hour. The security score of each link is calculated by weighted summation. The link bandwidth and latency are normalized and converted into a performance score between 0 and 1. The security score and performance score are synthesized in a ratio of 6:4 to obtain the comprehensive score of each link. A network weighted graph is constructed based on the comprehensive score of the link, where the node represents the device in the network, the edge represents the link, and the weight of the edge is the inverse of the comprehensive score of the link. The improved Dijkstra algorithm is used to search for the shortest path between any two nodes on the weighted graph. The algorithm improvements include: when selecting the next hop, not only the cumulative weight is considered, but also the security score of the weakest link on the path, to ensure that there is no link with too low security strength in the path. The routing calculation task is assigned to multiple edge nodes for parallel processing to improve the calculation efficiency. In this way, network performance is taken into account while ensuring communication security. The calculated optimal routing table is distributed to each edge node. A lightweight routing decision program is deployed on the edge node to make routing selection based on the locally cached routing table and real-time network conditions. When a change in network topology or a significant change in link status is detected, the routing table update process is triggered. The routing decision program sends an update request to the central node through the message queue. The central node re-executes the routing optimization algorithm, generates a new routing table and pushes it to the relevant edge nodes. The edge node fine-tunes the routing table according to the local network conditions to achieve rapid response and adaptive optimization. In a network consisting of 10 edge nodes, the network management system first obtains the topology structure and identifies 25 links. The Nmap tool scans each link and finds that 15 links use AES-256 encryption, 8 use AES-128, and 2 use 3DES. In terms of authentication protocols, 20 links use WPA2 and 5 use WPA.The key update cycle distribution is: 10 links are updated every hour, 12 links are updated every day, and 3 links are updated every week. The i Perf test results show that the link bandwidth is distributed between 10Mbps and 1Gbps, and the latency ranges from 5ms to 100ms. The edge nodes process this data in parallel to generate a 25x6 link attribute matrix. The security scoring model runs on the edge nodes and assigns 3 points to AES-256, 2 points to AES-128, 1 point to 3DES, 3 points to WPA2, and 2 points to WPA; 3 points for hourly updates, 2 points for daily updates, and 1 point for weekly updates. After the weights are applied, the security score of each link is obtained. Bandwidth and latency are converted into performance scores through min-max normalization. The comprehensive score calculation takes into account the balance between security and performance in typical scenarios. The constructed network weighted graph contains 10 vertices and 25 edges, and the edge weight is the inverse of the comprehensive score. In addition to accumulating weights, the improved Dijkstra algorithm also maintains a safety threshold (such as 2.5 points) when selecting a path to ensure that links below this threshold are not selected. The 10 edge nodes are responsible for calculating the optimal paths to the other 9 nodes and summarizing the results. The final generated routing table contains 90 optimal paths. The routing decision program of the edge node checks the local network status every 30 seconds. When it detects that the link bandwidth changes by more than 20% or the delay increases by more than 50%, an update request is triggered. After receiving the request, the central node completes the calculation and distribution of the new routing table within 5 seconds. After receiving the new routing table, the edge node fine-tunes the routing weight by ±10% according to the local real-time status to achieve millisecond-level routing optimization response.

[0026] S103. Send the optimized node communication route to each edge computing node, and dynamically adjust the authentication mechanism and key management strategy of the secure channel between nodes according to the link security parameters, including link encryption strength, delay jitter threshold, and packet loss rate threshold.

[0027] The optimized node communication routing table sent by the edge computing management node is obtained, and the optimized node communication routing table is distributed to the edge computing node through the TLS1.3 protocol, and the edge computing node receives the optimized node communication routing table. The security parameters of the link between the edge computing node and the adjacent node are collected according to the preset security parameters; the security parameters include link encryption strength, delay jitter value and packet loss rate value, and it is judged whether the security parameters meet the preset security threshold. If the link encryption strength is lower than the preset security threshold, the encryption algorithm is upgraded; if the delay jitter value or the packet loss rate value exceeds the preset security threshold, the transmission protocol is switched. The current link security status of the edge computing node is obtained; the key update frequency and key length are determined according to the current link security status; if the current link security status is high risk, the key update frequency is increased and the key length is increased. The Shamir's Secret Sharing algorithm is used to store the key in multiple edge computing nodes in a dispersed manner; the key is reconstructed through the collaboration of the multiple edge computing nodes, and it is judged whether the version of the reconstructed key is the latest version.

[0028] Specifically, the optimized node communication routing table is obtained from the edge computing management node, and the routing table is distributed to each edge computing node through the TLS1.3 protocol. During the distribution process, the routing table is digitally signed using the RSA-2048 algorithm and encrypted using the AES-256-GCM algorithm to prevent the routing information from being tampered with or stolen. After receiving the routing table, the edge computing node verifies its integrity and authenticity, and then updates the local routing configuration. The edge computing node regularly collects the security parameters of the link between the adjacent nodes, including link encryption strength, delay jitter, and packet loss rate. A sliding window algorithm with 100 samples is used, updated once per second, and the average and standard deviation of these parameters are calculated to obtain the statistical characteristics of the link status. The calculation results are compared with the preset security threshold to determine whether the link security status meets the requirements. The delay jitter threshold is set to 10 milliseconds, the packet loss rate threshold is set to 1%, and the encryption strength threshold is set to 128-bit key length. According to the judgment result of the link security status, the configuration of the secure channel between nodes is dynamically adjusted. If the link encryption strength is lower than 128 bits, the encryption algorithm is upgraded or the key length is increased to 256 bits; if the delay jitter exceeds 10 milliseconds or the packet loss rate exceeds 1%, TCP is switched to QUI C protocol to improve transmission reliability. At the same time, the strength of the authentication mechanism is adjusted. In high-risk situations, certificate authentication combined with multi-factor authentication using a one-time password is used, and in low-risk situations, only certificate authentication is used to simplify the authentication process. According to the current security status, the key update frequency is preliminarily adjusted to prepare for the next key management optimization. The key management strategy for the secure channel is optimized. Based on the link security status, the key update frequency and key length are dynamically adjusted. In a high-risk environment, the key update frequency is increased to once an hour, and the key length is set to 256 bits; in a low-risk environment, the update frequency is reduced to once a day, and the key length is set to 128 bits. The Shamir's Secret Sharing algorithm is used to implement distributed key management, and the keys are stored in multiple edge nodes in a dispersed manner. When the key is needed, the complete key is reconstructed through collaboration to reduce dependence on the central key server. In addition, a key version control mechanism is implemented to ensure that all nodes use the latest key for communication, improving security and reliability. In a network consisting of 20 edge computing nodes, the edge computing management node generates an optimized routing table of 300 records. Using the TLS1.3 protocol, the routing table is signed using the RSA-2048 algorithm, and then encrypted with AES-256-GCM before distribution. After each edge node receives the encrypted routing table, it first verifies the signature with the preset public key, then decrypts it with its own private key and updates the local routing configuration. Subsequently, the security parameter collection program is started between nodes, and link encryption strength, latency jitter, and packet loss rate data are collected every 100 milliseconds.Using a sliding window of 100 samples, the average encryption strength is calculated to be 192 bits, the average delay jitter is 8 milliseconds, and the average packet loss rate is 0.8%. These values ​​are compared with the preset thresholds (128-bit encryption strength, 10 milliseconds delay jitter, 1% packet loss rate) to determine that the current link is in a medium-risk state. Based on this judgment, the encryption algorithm of 5 nodes is upgraded from AES-128 to AES-256, and 3 high-packet loss rate links are switched from TCP to QUI C protocol. The authentication mechanism is also adjusted accordingly, and 10 high-risk nodes enable two-factor authentication, combining certificate verification and a one-time password with a validity period of 60 seconds. At the same time, the key management strategy changes accordingly, and the key update frequency between high-risk nodes is increased to once every 4 hours, and low-risk nodes are updated daily. Using Shamir's Secret Sharing algorithm, the 256-bit master key is divided into 15 parts and stored in different nodes. As long as 10 of them are collected, the complete key can be reconstructed. This distributed key management method ensures that even if 5 nodes are compromised, the communication security of the entire network can still be maintained.

[0029] S104: monitor network status changes and node communication quality in real time, analyze network status change trends and node communication quality scores, and when the scores are lower than a preset threshold for multiple consecutive times, start the adaptive networking process and reselect a matching networking strategy to avoid network shocks caused by frequent adjustments.

[0030] The network status data and node communication quality indicators sent by the edge computing node are obtained, and the communication quality indicators include link bandwidth, delay, packet loss rate and throughput; according to the network status data and the node communication quality indicators, a preset sliding window algorithm is used to calculate the statistical characteristics of the indicators; it is determined whether the statistical characteristics of the indicators meet the preset conditions, and if so, the node communication quality score is calculated according to the preset weight coefficient; the historical records of the network status data are processed by the pre-established ARI MA model to obtain the network status prediction results for the future preset period; the network status prediction results are compared with the current network status to determine whether the network performance shows a downward trend; it is determined whether the node communication quality score is lower than the preset score threshold for 5 consecutive times, and if so, the adaptive networking process is triggered; from the pre-established networking strategy library, the best matching networking strategy is selected according to the network status prediction results; the networking strategy is executed by a progressive adjustment method, and the strategy change is decomposed into multiple adjustment links; the feedback control mechanism is realized by a pre-set proportional-integral-differential controller, and the network parameters are adjusted according to the deviation, integral and change rate of the network performance indicators; if it is determined that the network performance has declined, the adjustment is rolled back and the strategy switching process is suspended.

[0031] Specifically, network status data and node communication quality indicators, including link bandwidth, latency, packet loss rate, and throughput, are collected from edge computing nodes. A 60-second sliding window algorithm is used, updated every 5 seconds, to calculate the average and standard deviation of these indicators to obtain the statistical characteristics of the network status. According to the preset weight coefficients, multiple indicators are comprehensively calculated to obtain the node communication quality score, specifically: bandwidth accounts for 30% weight, latency accounts for 30% weight, packet loss rate accounts for 20% weight, and throughput accounts for 20% weight. The final communication quality score is obtained by weighted average method. The trend of network status change is modeled using time series analysis methods. The ARI MA (1,1,1) model is used to process historical data and predict network status changes within the next 30 minutes. Among them, the order of the autoregressive term, the difference term, and the moving average term are all 1. The predicted results are compared with the current network status to determine whether the network performance is showing a downward trend. The score threshold is set to 75 points, the full score is 100 points, and the continuous number threshold is set to 5 times to trigger the adaptive networking process. If the node communication quality score is lower than 75 points for five consecutive times, the networking strategy reselection process is started. From the pre-established networking strategy library, the best matching networking strategy is selected according to the current network status and the predicted change trend. The networking strategy library contains multiple strategies such as "high bandwidth priority strategy", "low latency priority strategy" and "load balancing strategy". When selecting a strategy, consider the principle of gradual adjustment to avoid drastic changes. When executing the newly selected networking strategy, adopt a gradual adjustment method. Decompose the strategy change into multiple small steps, and only adjust 10% of the network parameters each time. The feedback control mechanism is implemented through the proportional-integral-differential (PID) controller to adjust the network parameters according to the deviation, integral and change rate of the network performance indicators. If it is found that the network performance has decreased, the adjustment is rolled back and the strategy switching process is suspended. In order to avoid network shock caused by frequent adjustments, a 30-minute cooling period mechanism is introduced, stipulating that there must be at least 30 minutes between two major adjustments. Through this mechanism, while ensuring network performance, unnecessary frequent adjustments are also reduced. In a network consisting of 50 edge computing nodes, each node collects network status data every 5 seconds, including link bandwidth, latency, packet loss rate, and throughput. Using a 60-second sliding window, the average bandwidth is calculated to be 100Mbps, the average latency is 50ms, the average packet loss rate is 1%, and the average throughput is 80Mbps. According to the preset weights, bandwidth 30%, latency 30%, packet loss rate 20%, and throughput 20%, the node communication quality score is calculated to be 85 points. At the same time, the ARI MA (1,1,1) model is used to analyze the data of the past 24 hours, and it is predicted that the network performance may drop by 5% in the next 30 minutes. The adaptive networking process is triggered when the score is lower than 75 points for 5 consecutive times. Select "Load Balancing Strategy" as the new strategy from the networking strategy library and start gradual adjustment.First, 10% of the traffic is transferred from the high-load node to the low-load node, and the PID controller is used to monitor the adjustment effect, where the proportional coefficient Kp = 0.5, the integral coefficient Ki = 0.2, and the differential coefficient Kd = 0.1. After each adjustment, wait for 5 minutes to observe the changes in network performance. If the performance improves, continue with the next round of 10% adjustment; if the performance decreases, roll back the adjustment and pause for 30 minutes. After 3 rounds of adjustments, the network load tends to be balanced, the overall performance improves by 12%, and the score rises to 92 points. In the next 2 hours, it remained stable and no new adjustment process was triggered, effectively avoiding network shocks.

[0032] S105. According to the updated networking strategy, dynamically optimize the node roles and link redundancy, and update the network topology. In particular, the edge nodes are divided into three node functional roles: computing nodes, storage nodes, and forwarding nodes, in combination with the node computing power, storage capacity, and energy consumption level.

[0033] The hardware indicators of each edge node are obtained, and the comprehensive performance score of the node is calculated by the weighted average method according to the hardware indicators. The comprehensive performance score includes computing power, storage capacity and energy consumption level. The edge nodes are clustered by the DBSCAN algorithm. If the feature vector of the node meets the clustering condition, the node is divided into a computing node, a storage node or a forwarding node. The feature vector includes the computing power, storage capacity and energy consumption level of the node. According to the role assignment of the node, the improved Dijkstra algorithm is used to recalculate the optimal path in the network. The calculation of the optimal path combines the node role matching factor and the link status. According to the optimal path and the node role, a new network topology is generated. The network topology includes node role information and link redundancy information. The distributed consistency protocol is used to synchronize the network topology information to all edge nodes. The synchronization process propagates information through the Goss ip protocol until the network topology information is consistent among all nodes.

[0034] Specifically, the updated networking strategy is obtained from the edge computing management platform, and the relevant parameters of node role optimization and link redundancy adjustment are extracted. According to the policy requirements, the node role division standard is formulated, including computing power threshold, storage capacity threshold and energy consumption level threshold. The weighted average method is used to quantitatively evaluate the hardware indicators of each edge node, with computing power accounting for 40% of the weight, storage capacity accounting for 30% of the weight, and energy consumption level accounting for 30% of the weight, to obtain the node comprehensive performance score. It is set to evaluate and adjust the node role and link redundancy once an hour, and trigger role redistribution when the node performance or network status changes by more than 20%. The DBSCAN algorithm is used to group the edge nodes and divide the nodes into three categories: computing nodes, storage nodes and forwarding nodes. In the clustering process, the computing power, storage capacity and energy consumption level of the node are used as feature vectors, the neighborhood radius is set to 0.5, and the minimum number of samples is 3. According to the clustering results, the functional role of each node is determined, and a node role allocation table is generated. In the dynamic optimization process, the stability of the existing role allocation is maintained. Only when the performance of the node changes significantly and exceeds the preset threshold of 20%, its role is considered to be adjusted. Based on the new node role allocation, the optimal path in the network is recalculated. The improved Dijkstra algorithm is used to consider the node role and link status and build a weighted graph model. When calculating the path weight, the node role matching factor is introduced, and the path from the computing node to the storage node is given a weight of 0.8, and other paths are given a weight of 0.5. In the path selection process, the connection between the computing node and the storage node is given priority to ensure the efficiency of data processing and storage. At the same time, according to the redundancy requirements in the networking strategy, a backup link is added to the critical path, and the redundancy is set to 2. According to the optimized path and node role, a new network topology is generated. The distributed consistency protocol is used to synchronize the new network topology information to all edge nodes, and their routing tables and neighbor node information are updated. The Goss ip protocol is used for information dissemination. Each node randomly selects neighbor nodes for information exchange until the network topology information is consistent among all nodes. To ensure consistency, a version number is set to identify topology updates, and only newer versions of topology information will be accepted and propagated. In this way, dynamic optimization and timely synchronization of network topology are achieved. In a network consisting of 100 edge nodes, the edge computing management platform performs an assessment of node roles and link redundancy every hour. In the latest assessment, the computing power of node 1 was collected as an 8-core CPU, a storage capacity of 500GB, and an energy consumption level of 50W. The score is calculated using the weighted average method. Similarly, the scores of all nodes are calculated to form a 100x3 feature matrix. The DBSCAN algorithm is applied, the neighborhood radius is set to 0.5, the minimum number of samples is set to 3, and the nodes are clustered into 3 groups: 35 computing nodes, 40 storage nodes, and 25 forwarding nodes. Node 1 is classified as a computing node due to its high computing power.Subsequently, the improved Dijkstra algorithm was used to recalculate the network path. In the path from computing node 1 to storage node 2, a weight factor of 0.8 was assigned, while the path to forwarding node 3 was assigned a weight of 0.5. The algorithm found an optimal path with a total weight of 2.3, 1->4->2, and configured a backup path 1->5->6->2 for the path, achieving the redundancy requirement of 2. The new network topology was propagated through the Gossi p protocol, and each node randomly selected 3 neighbors per second to exchange information. After about 10 rounds of propagation, about 10 seconds, all 100 nodes in the entire network reached a consensus and obtained the latest topology information with version number v2.5. During the entire process, only 5 nodes changed their roles due to performance changes of more than 20%, ensuring the relative stability of the network structure.

[0035] Through the inter-node communication protocol and node load balancing requirements, the node function role is dynamically adjusted, and the adjusted node function role attributes are updated to the network topology.

[0036] The real-time performance data of each node sent by the network monitoring module is obtained, and the comprehensive load score of each node is calculated according to the preset performance index weight. The load information sent by the directly connected neighboring nodes is received through the inter-node communication protocol, and a star-shaped local load distribution diagram centered on itself is constructed. It is judged whether there is load imbalance in the local load distribution diagram. If the load scores of some nodes exceed the preset threshold, while the load scores of other nodes are lower than the preset threshold, the node function role adjustment process is triggered. The greedy algorithm is used for role assignment. The nodes in the local load distribution diagram are sorted from high to low according to the load, and the neighboring node with the lowest load is selected for each overloaded node for role exchange. The dynamic adjustment of the node function role is performed, and a progressive switching strategy is used to assign new tasks to the target role while maintaining the execution of the original task; when the task volume of the new role reaches the preset proportion of the total task volume, the original task is migrated; the task migration process includes data transmission and state synchronization, and the amount of transmitted data is reduced by incremental synchronization; during the role conversion process, the node performance indicators are continuously monitored. If abnormal fluctuations occur, the switching process is suspended and rolled back to the initial state.

[0037] Specifically, the real-time performance data of each node is obtained from the network monitoring module, including CPU usage, memory usage, network throughput, and the number of tasks currently carried. According to the preset performance indicator weights, CPU usage 30%, memory usage 25%, network throughput 25%, and number of tasks 20%, the comprehensive load score of each node is calculated. The load information of directly connected neighboring nodes is collected through the inter-node communication protocol, and a star topology local load distribution diagram centered on itself is constructed. If load imbalance is detected, that is, the load score of some nodes exceeds the preset threshold of 80 points, the full score is 100 points, and other nodes are less than 50 points, the node function role adjustment process is triggered. Based on the current network topology and the characteristics of the inter-node communication protocol, the cost of node role conversion is calculated. The cost factors include the time required for role switching, the degree of impact on network communication, and the overhead of resource reallocation. A greedy algorithm is used for role allocation: nodes are sorted from high to low according to load, and the neighboring node with the lowest load is selected for each overloaded node for role exchange. For nodes using the MQTT protocol, the subscription topic is reconfigured when the role is converted; for nodes using HTTP, the API routing table is updated. Dynamic adjustment of node function roles is performed. For nodes that need to change roles, a gradual switching strategy is adopted. First, new tasks are assigned to the target role while maintaining the execution of the original tasks. When the task volume of the new role reaches 70% of the total task volume, the original tasks are migrated. The task migration process includes data transmission and state synchronization, and the amount of transmitted data is reduced by incremental synchronization. During the role conversion process, the node performance indicators are continuously monitored. If there is an abnormal fluctuation (such as a sudden increase of 20% in CPU usage), the switching process is suspended and rolled back to the initial state. The adjusted node function role attributes are updated to the network topology. The version number mechanism is used to manage the update of the topology structure, and a new version number is generated for each modification. The updated topology information is synchronized between all nodes using the consistent hashing algorithm: the node ID is used as the hash key, the topology information is sharded and stored on different nodes, and the node responsible for storing each shard is found through consistent hashing to achieve fast and load-balanced information synchronization. According to the new node role distribution, the communication protocol parameters between nodes are dynamically adjusted, such as updating the routing table and adjusting the data distribution strategy to adapt to the new network structure. In a network consisting of 50 edge computing nodes, the node monitoring module collects performance data every 30 seconds. The latest data of node A shows that the CPU utilization rate is 85%, the memory usage is 70%, the network throughput is 60Mbps, and the current number of tasks is 15. The comprehensive load score is calculated based on the weight. Node A broadcasts its load information to the five adjacent nodes through the MQTT protocol, and receives the load data of these five nodes at the same time, constructing a star-shaped local load distribution diagram centered on itself and covering 6 nodes. It was detected that the score of node A exceeded the threshold of 80 points, while the score of the adjacent node B was only 45 points, triggering the role adjustment process.The greedy algorithm chooses to transfer part of the computing tasks of node A to node B. Considering that node A uses the MQTT protocol, the subscription topics of nodes A and B are reconfigured, and 50% of the new task topics originally sent to A are redirected to B. After the task migration begins, the performance indicators of A and B are continuously monitored. When the amount of tasks undertaken by node B reaches 70% of the total, the original task increments on node A are synchronized to node B, and 10% of the task data and status information are transmitted each time. In this process, the CPU utilization rate of node A dropped from 85% to 65%, while that of node B increased from 40% to 75%. After completing the role adjustment, a new topology structure version number v2.5 is generated. Using the consistent hashing algorithm, the updated topology information is divided into 50 shards and stored on 50 nodes respectively. Each node is responsible for storing and synchronizing the topology information shards within the range of 2 adjacent hash values, ensuring the rapid propagation of information and load balancing. Finally, the entire network completed the update and synchronization of topology information within 3 seconds, and all nodes obtained the latest role allocation information.

[0038] S106. Continuously monitor network conditions and node communication quality, smooth out network condition fluctuations by setting corresponding perception time windows, avoid network jitters caused by overly sensitive condition perception, and respond to network changes while maintaining network stability. Perform security scoring against preset thresholds, dynamically trigger network strategy adjustments based on scoring results, and respond to changes in the network environment while maintaining network stability.

[0039] The network status and node communication quality data collected regularly by the network monitoring module are obtained, including link bandwidth, delay, packet loss rate and node load indicators; based on the network status and node communication quality data, each indicator is subjected to exponential weighted moving average processing using a sliding time window to obtain smoothed network status data; it is determined whether the smoothed network status data is within a normal range, wherein the normal range is dynamically determined based on the mean and standard deviation of each indicator within a certain period of time in the past; if the smoothed network status data is within the normal range, the smoothed network status data is scored according to a preset security scoring model to obtain a network security score; if the network security score is lower than a preset threshold for multiple consecutive times, the policy that best matches the current network status is selected from the predefined policy library, and the network parameters are gradually adjusted until the network security score is increased to above the preset threshold.

[0040] Specifically, the network status and node communication quality data are collected from the network monitoring module every 5 seconds, including indicators such as link bandwidth, latency, packet loss rate, and node load. A 60-second sliding time window is set, and exponential weighted moving average processing is applied to each indicator separately, giving the latest data a weight of 0.3 and the historical data a weight of 0.7 to smooth short-term fluctuations. The processed data is stored in the InfluxDB time series database as the basis for subsequent analysis. The EWMA control chart method is used to dynamically adjust the judgment criteria for network status. The EWMA value μ and standard deviation σ of each indicator in the past 24 hours are calculated, and μ±3σ is used as the upper and lower control limits of the normal range. If the current smoothed indicator value falls within this range, the network status is considered normal; otherwise, the anomaly detection process is triggered. In this way, the judgment criteria are automatically adjusted with the long-term changes of the network. The anomaly detection results are passed as input to the security scoring module to adjust the scoring weight. According to the preset security scoring model, the smoothed network status data is scored once every 5 minutes. The scoring model considers factors such as link encryption strength, weight 0.4, authentication mechanism security, weight 0.3, and data integrity, weight 0.3, and calculates the total score by weighted summation. Set 90 points, out of 100 points, as the security threshold. When the score is lower than the threshold for three consecutive times, i.e., 15 minutes, the networking policy adjustment process is triggered. Start the progressive networking policy adjustment mechanism. First, select the policy that best matches the current network status from the predefined policy library. The policy library includes "high security policy", "low latency policy", and "load balancing policy". Then, divide the new policy into multiple small steps, and adjust only 10% of the network parameters each time. After each small step is executed, wait 30 seconds to observe the changes in network status. If the network score improves, continue to the next step; if the score decreases, roll back the current step and suspend the adjustment process. Introduce a 30-minute cooling period mechanism, stipulating that there must be at least 30 minutes between two major adjustments to prevent frequent policy switching from causing network jitter. In this way, while maintaining network stability, gradually adapt to changes in the network environment. In a network consisting of 100 edge computing nodes, the monitoring module collects data every 5 seconds. Taking node A as an example, the latest collection shows that its link bandwidth is 95Mbps, the delay is 20ms, the packet loss rate is 0.5%, and the CPU load is 75%. Apply a 60-second sliding window to perform exponentially weighted moving average processing on each indicator. Taking bandwidth as an example, the current smoothing value is 92Mbps, the weight of the latest data is 0.3, and the updated smoothing value is: 920.7+950.3=92.9Mbps. Other indicators are processed simultaneously and stored in InfluxDB. The EWMA control chart method is based on the data of the past 24 hours, and the bandwidth mean μ=90Mbps, the standard deviation σ=5Mbps, and the control limit is set to 75~105Mbps.The current 92.9Mbps is within the range and is considered normal. The security scoring module scores every 5 minutes. The current link uses AES-256 encryption, with a score of 95, two-factor authentication, with a score of 90, and a data integrity check pass rate of 99.9%, with a score of 98. The weighted total score is 94.4 points, which is higher than the 90-point threshold, and the network security status is good. Within 15 minutes, the node score was always higher than the threshold, and no policy adjustment was triggered. However, node B scored below the threshold for three consecutive times, triggering the adjustment process. Select "High Security Policy" from the policy library and divide it into 10 steps. First, increase the encryption strength from AES-128 to AES-256, and the score will increase from 88 to 92. After waiting for 30 seconds, execute the next step to increase the authentication complexity. The entire adjustment process took 5 minutes, and the network score stabilized above 95 points. Mark the adjustment timestamp and start a 30-minute cooling period. During this period, even if a slight fluctuation is detected, it will not trigger a new large-scale adjustment, effectively avoiding network jitter.

[0041] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A secure channel networking method based on edge computing power, characterized in that: The method comprises: Obtain the multi-dimensional network status parameters of each edge computing node, including network topology, link quality and node load, and use the fuzzy comprehensive evaluation method to determine whether the current network status meets the preset stability threshold. If not, select the networking strategy with the highest matching degree with the current network status from the preset networking strategy library and combine the parameters of the networking strategy, including node communication protocol, link redundancy and node functional role; Based on the selected networking strategy, the security parameters of the links between nodes in the network topology are obtained, including link encryption strength, authentication mechanism, and key update cycle. In combination with link bandwidth and latency, the optimal secure communication route between nodes is obtained through edge computing methods. The optimized node communication route is sent to each edge computing node, and the authentication mechanism and key management strategy of the secure channel between nodes are dynamically adjusted according to the link security parameters, including link encryption strength, delay jitter threshold, and packet loss rate threshold; Monitor network status changes and node communication quality in real time, analyze network status change trends and node communication quality scores, and start the adaptive networking process when the scores are lower than the preset threshold for multiple consecutive times, reselect the matching networking strategy to avoid network shocks caused by frequent adjustments; According to the updated networking strategy, the node roles and link redundancy are dynamically optimized, and the network topology is updated. The edge nodes are divided into three node functional roles: computing nodes, storage nodes, and forwarding nodes, based on the node computing power, storage capacity, and energy consumption level. Continuously monitor network conditions and node communication quality, smooth out network condition fluctuations by setting corresponding perception time windows, avoid network jitters caused by overly sensitive condition perception, respond to network changes while maintaining network stability, perform security scoring against preset thresholds, dynamically trigger network strategy adjustments based on scoring results, and respond to changes in the network environment while maintaining network stability.

2. The method according to claim 1, characterized in that The multi-dimensional network status parameters of each edge computing node are obtained, including network topology, link quality and node load, and a fuzzy comprehensive evaluation method is used to determine whether the current network status meets the preset stability threshold. If not, the networking strategy with the highest matching degree with the current network status is selected from the preset networking strategy library and combined with the parameters of the networking strategy, including node communication protocol, link redundancy and node function role, including: Obtain network topology data, link quality parameters, and real-time load information of the edge computing cluster, and construct the multi-dimensional parameter matrix; Wherein, the multi-dimensional parameter matrix reflects the current network status; Applying fuzzy comprehensive evaluation to the multidimensional parameter matrix; The fuzzy comprehensive evaluation includes standardizing the network topology, link quality and node load parameters, establishing a fuzzy relationship matrix, and obtaining a comprehensive evaluation result through fuzzy synthesis operation; Comparing the comprehensive evaluation result with a preset stability threshold; If the comprehensive evaluation result does not meet the preset stability threshold, all available strategies are extracted from the pre-established networking strategy library; Calculate vector similarity based on available strategies and current network status parameters; The vector similarity calculation is used to select the strategy with the highest similarity as the optimal networking solution.

3. The method according to claim 1, characterized in that Based on the selected networking strategy, the security parameters of the links between nodes in the network topology are obtained, including link encryption strength, authentication mechanism, key update cycle, and the optimal secure communication route between nodes is obtained by edge computing method in combination with link bandwidth and latency, including: Acquire network topology data in the network management system, wherein the network topology data includes the number of nodes, the type of nodes, and the connection relationship between the nodes; Use the Nmap tool to scan the security parameters of the network link to obtain the link encryption algorithm type, Key length and authentication protocol version and key update frequency; The bandwidth capacity and end-to-end communication delay of each link are measured using the iPerf measurement tool; Deploy a security scoring model on the edge node based on link security parameters; The security scoring model assigns weight values ​​to encryption strength, authentication mechanism and key update cycle respectively; Calculate the security score of each link according to the weight value; Normalize the link bandwidth and delay to obtain the link performance score; The security score and the performance score are combined according to a preset ratio to obtain a comprehensive score for each link; A network weighted graph is constructed based on the comprehensive link score, where nodes represent network devices, edges represent links, and the weight of the edge is the inverse of the comprehensive link score; The improved Dijkstra algorithm is used to search the shortest path between any two nodes on the weighted graph to obtain the optimal routing table; Distributing the optimal routing table to each edge node and deploying a lightweight routing decision program on the edge node; The lightweight routing decision maker makes routing selections based on the locally cached routing table and real-time network conditions; If a network topology change or a significant link status change is detected, the routing table update process is triggered.

4. The method according to claim 1, characterized in that The optimized node communication route is sent to each edge computing node, and the authentication mechanism and key management strategy of the secure channel between nodes are dynamically adjusted according to the link security parameters, including link encryption strength, delay jitter threshold, and packet loss rate threshold, including: Obtaining the optimized node communication routing table sent by the edge computing management node, distributing the optimized node communication routing table to the edge computing node through the TLS1.3 protocol, and the edge computing node receiving the optimized node communication routing table; Collect security parameters of the link between the edge computing node and the adjacent node according to the preset security parameters; The security parameters include link encryption strength, delay jitter value and packet loss rate value, and it is determined whether the security parameters meet the preset security threshold; If the link encryption strength is lower than the preset security threshold, the encryption algorithm is upgraded; if the delay jitter value or packet loss rate value exceeds the preset security threshold, the transmission protocol is switched; Get the current link security status of the edge computing node; Determine the key update frequency and key length based on the current link security status; If the current link security status is high risk, increase the key update frequency and increase the key length; Use Shamir's Secret Sharing algorithm to store keys in multiple edge computing nodes; The key is reconstructed through collaboration of the multiple edge computing nodes to determine whether the version of the reconstructed key is the latest version.

5. The method according to claim 1, characterized in that The real-time monitoring of network status changes and node communication quality, analyzing network status change trends and node communication quality scores, and starting the adaptive networking process when the scores are lower than the preset threshold for multiple consecutive times, reselecting the matching networking strategy to avoid network shocks caused by frequent adjustments, including: Obtain network status data and node communication quality indicators sent by edge computing nodes, wherein the communication quality indicators include link bandwidth, latency, packet loss rate, and throughput; According to the network status data and node communication quality indicators, the preset sliding window algorithm is used to calculate the statistical characteristics of the indicators; Determine whether the statistical characteristics of the indicator meet the preset conditions. If so, calculate the node communication quality score according to the preset weight coefficient; Use the pre-established ARIMA model to process the historical records of network status data to obtain the network status forecast results for the preset time period in the future; Compare the network status prediction result with the current network status to determine whether the network performance is showing a downward trend; Determine whether the node communication quality score is lower than the preset score threshold for 5 consecutive times. If so, trigger the adaptive networking process; Select the best matching networking strategy from the pre-established networking strategy library according to the network status prediction results; Use a progressive adjustment method to implement networking strategies, breaking down strategy changes into multiple adjustment steps; A feedback control mechanism is implemented through a pre-set proportional-integral-derivative controller to adjust network parameters according to the deviation, integral and rate of change of network performance indicators; If it is determined that the network performance has degraded, the adjustment will be rolled back and the policy switching process will be suspended.

6. The method according to claim 1, characterized in that According to the updated networking strategy, the node roles and link redundancy are dynamically optimized, and the network topology is updated. In combination with the node computing power, storage capacity, and energy consumption level, the edge nodes are divided into three node functional roles: computing nodes, storage nodes, and forwarding nodes, including: Obtain the hardware indicators of each edge node, and calculate the comprehensive performance score of the node using the weighted average method based on the hardware indicators; The comprehensive performance score includes computing power, storage capacity and energy consumption level; The DBSCAN algorithm is used to cluster edge nodes. If the feature vector of the node meets the clustering condition, the node is divided into a computing node and a storage node or a forwarding node; The feature vector includes the computing power and storage capacity of the node and the energy consumption level; According to the role allocation of nodes, the improved Dijkstra algorithm is used to recalculate the optimal path in the network; The calculation of the optimal path combines the node role matching factor and the link status; Generate a new network topology based on the optimal path and node roles; The network topology includes node role information and link redundancy information; Use distributed consistency protocol to synchronize network topology information to all edge nodes; The synchronization process propagates information through the Gossip protocol until the network topology information is consistent among all nodes; It also includes: dynamically adjusting the node function role through the inter-node communication protocol and the node load balancing requirements, and updating the adjusted node function role attributes into the network topology structure.

7. The method according to claim 6, characterized in that The method of dynamically adjusting the node function role through the inter-node communication protocol and the node load balancing requirements, and updating the adjusted node function role attributes to the network topology structure, includes: Obtain the real-time performance data of each node sent by the network monitoring module, and calculate the comprehensive load score of each node according to the preset performance indicator weight; Receive the load information sent by directly connected neighbor nodes through the inter-node communication protocol, and build a star topology local load distribution map centered on itself; Determine whether there is load imbalance in the local load distribution diagram. If the load scores of some nodes exceed the preset threshold, while the load scores of other nodes are lower than the preset threshold, the node function role adjustment process is triggered; A greedy algorithm is used for role assignment. The nodes in the local load distribution graph are sorted from high to low according to the load. The neighboring node with the lowest load is selected for each overloaded node for role exchange. Dynamic adjustment of the functional roles of execution nodes, using a progressive switching strategy to assign new tasks to the target roles while maintaining the execution of the original tasks; When the task load of the new role reaches the preset proportion of the total task load, start migrating the original tasks; The task migration process includes data transmission and state synchronization, and the amount of transmitted data is reduced through incremental synchronization; During the role switching process, the node performance indicators are continuously monitored. If abnormal fluctuations occur, the switching process is paused and rolled back to the initial state.

8. The method according to claim 1, characterized in that: The continuous monitoring of network conditions and node communication quality, smoothing the volatility of network conditions by setting corresponding perception time windows, avoiding network jitter caused by overly sensitive condition perception, responding to network changes while maintaining network stability, performing security scoring against preset thresholds, dynamically triggering network strategy adjustments based on scoring results, and coping with changes in the network environment while maintaining network stability, including: Obtain network status and node communication quality data regularly collected by the network monitoring module, including link bandwidth, latency, packet loss rate, and node load indicators; According to the network status and node communication quality data, a sliding time window is used to perform exponentially weighted moving average processing on each indicator to obtain smoothed network status data; Determine whether the smoothed network status data is within a normal range, where the normal range is dynamically determined based on the mean and standard deviation of various indicators within a certain period of time in the past; If the smoothed network status data is within the normal range, the smoothed network status data is scored according to a preset security scoring model to obtain a network security score; If the network security score is lower than the preset threshold for multiple consecutive times, the policy that best matches the current network status is selected from the predefined policy library, and the network parameters are gradually adjusted until the network security score is raised above the preset threshold.

Images