Edge computing driven terminal data security acceleration distribution method and system

Abstract

The embodiment of the specification discloses an edge computing driven terminal data security acceleration distribution method and system. The terminal data security acceleration distribution method comprises the following steps: an Internet of Things terminal sends a task request; a current edge computing node receives the task request, executes an edge acceleration distribution process according to the task request, and the edge acceleration distribution process comprises the following steps: a node feature vector and a task feature vector are acquired; a task resilience enhancement strategy identifier is determined based on the node feature vector and the task feature vector; an acceleration distribution channel is determined according to the resilience enhancement strategy identifier, and the acceleration distribution channel is used to execute the task through different resilience enhancement strategies; the current edge computing node or a cooperative node acquires execution data corresponding to the task, and sends the execution data to the Internet of Things terminal. The embodiment of the specification not only can automatically accelerate and distribute the Internet of Things terminal task, but also realizes accurate matching of computing resources and task demand, and significantly improves the task processing efficiency and resource utilization.

Description

TECHNICAL FIELD

[0001] One or more embodiments of the present specification relate to the technical field of edge computing, in particular to an edge computing driven terminal data security acceleration distribution method and system. BACKGROUND

[0002] With the continuous development of information technology, the number of Internet of Things terminals is also growing rapidly, and the application scenarios of Internet of Things terminals are becoming increasingly complex. At present, the Internet of Things terminal generally adopts a centralized processing mode of terminal collection and cloud computing. In this mode, the Internet of Things terminal only acts as a data collector to remotely return the original information to the cloud data center through the network for unified processing and decision-making. However, the Internet of Things terminal information software generally has the following problems when processing tasks: in the traditional centralized cloud computing mode, terminal task data needs to be uploaded to the cloud for processing, which brings high delay and high network load, resulting in high data transmission delay; moreover, in different terminal and application scenarios, the task type, data volume and real-time requirement difference are significant, and the task characteristics are different, so it is difficult to meet the task analysis accuracy requirement by using a single algorithm; in addition, although the existing edge computing can sink part of the calculation, it lacks an intelligent algorithm matching mechanism, and it is difficult to automatically adapt to the optimal acceleration algorithm, so the acceleration matching is difficult; moreover, the data transmission between the edge computing node and the Internet of Things terminal generally depends on the wireless network, and the bandwidth and channel quality of the wireless network have high time-varying characteristics, while the edge computing node is often in a complex and uncontrolled environment, and its computing resources are limited, so when the node is abnormal or the network fluctuates, the task may be interrupted or delayed, and most of the existing systems lack the self-adaptive ability to independently respond to node or network abnormal situations, so the reliability is difficult to meet the key task requirements; moreover, the cross-network transmission efficiency of the task result is low, and the data interaction has privacy leakage and security risks. Therefore, it is an urgent technical problem to be solved to realize the automatic acceleration distribution of different Internet of Things terminal information software tasks, while enhancing the self-adaptive ability, transmission efficiency and security of the system. SUMMARY

[0003] The edge computing driven terminal data security acceleration distribution method and system provided by the embodiments of the present specification have the technical solutions as follows:

[0004] In a first aspect, the embodiments of the present specification provide an edge computing driven terminal data security acceleration distribution method, comprising: establishing a communication link between an Internet of Things terminal and a current edge computing node, and sending a task request; the current edge computing node receives the task request, and executes an edge acceleration distribution process according to the task request, the edge acceleration distribution process comprising: obtaining a node feature vector and a task feature vector corresponding to the task request; determining a resilience enhancement strategy identifier of the task based on the node feature vector and the task feature vector; determining an acceleration distribution channel according to the resilience enhancement strategy identifier, the acceleration distribution channel being used to execute the task through different resilience enhancement strategies, the different resilience enhancement strategies comprising: when the acceleration distribution channel is a first acceleration channel, determining a main target acceleration algorithm according to a preset edge acceleration algorithm library and the task feature vector, and executing the task based on the main target acceleration algorithm; when the acceleration distribution channel is a second acceleration channel, determining a degraded acceleration algorithm according to the task feature vector, and executing the task based on the degraded acceleration algorithm; when the acceleration distribution channel is a cooperative acceleration channel, determining a cooperative node corresponding to the current edge computing node, the cooperative node being used to execute the task according to the edge acceleration distribution process; the current edge computing node or the cooperative node obtains execution data corresponding to the task, and sends the execution data to the Internet of Things terminal.

[0005] In a second aspect, the embodiments of the present specification provide an edge computing driven terminal data security acceleration distribution system, comprising: an Internet of Things terminal, configured to establish a communication link with a current edge computing node, and send a task request; the current edge computing node, configured to receive the task request, and execute an edge acceleration distribution process according to the task request, the current edge computing node comprising a feature acquisition module, an identifier determination module and a channel determination module; the feature acquisition module is configured to obtain a node feature vector and a task feature vector corresponding to the task request; the identifier determination module is configured to determine a resilience enhancement strategy identifier of the task based on the node feature vector and the task feature vector; the channel determination module is configured to determine an acceleration distribution channel according to the resilience enhancement strategy identifier, the acceleration distribution channel being used to execute the task through different resilience enhancement strategies, the different resilience enhancement strategies comprising: when the acceleration distribution channel is a first acceleration channel, determining a main target acceleration algorithm according to a preset edge acceleration algorithm library and the task feature vector, and executing the task based on the main target acceleration algorithm; when the acceleration distribution channel is a second acceleration channel, determining a degraded acceleration algorithm according to the task feature vector, and executing the task based on the degraded acceleration algorithm; when the acceleration distribution channel is a cooperative acceleration channel, determining a cooperative node corresponding to the current edge computing node, the cooperative node being used to execute the task according to the edge acceleration distribution process; the current edge computing node or the cooperative node obtains execution data corresponding to the task, and sends the execution data to the Internet of Things terminal.

[0006] The technical solutions provided by some embodiments of the present specification have at least the following beneficial effects:

[0007] The current edge computing node in the embodiment of the present specification can execute an edge acceleration distribution process according to a task request, that is, obtaining a node feature vector and a task feature vector, and then determining a resilience enhancement strategy identifier of the task based on the node feature vector and the task feature vector. The embodiment of the present specification determines the resilience enhancement strategy identifier of the task by introducing the task feature vector and the node feature vector representing the node state information, realizes the context awareness and precision of different resilience enhancement strategies executing the task, that is, realizes the accurate prediction of whether the current edge computing node environment is sufficient to guarantee the successful execution of the task. Moreover, the embodiment of the present specification can match different acceleration distribution channels according to the resilience enhancement strategy identifier, the acceleration distribution channels being a first acceleration channel, a second acceleration channel or a cooperative acceleration channel, and the first acceleration channel and the second acceleration channel corresponding to a main acceleration channel and a degraded acceleration channel respectively. The main acceleration channel can match a main target acceleration algorithm according to a preset edge acceleration algorithm library and the task feature vector, the degraded acceleration channel can execute the task based on a degraded acceleration algorithm, and the cooperative acceleration channel can update the edge acceleration distribution process based on a cooperative node corresponding to the current edge computing node to execute the task. Then, the current edge computing node or the cooperative node obtains execution data corresponding to the task and sends the execution data to the Internet of Things terminal.

[0008] The embodiment of the present specification can automatically accelerate the distribution of the Internet of Things terminal task, that is, by introducing the task feature vector, the task feature vector and the preset edge acceleration algorithm library, and matching different acceleration distribution channels according to the resilience enhancement strategy identifier, so as to dynamically match the optimal acceleration algorithm for tasks with different characteristics, realizing the accurate matching of computing resources and task demand, and significantly improving the task processing efficiency and resource utilization.

[0009] The embodiment of the present specification also sets a system resilience guarantee mechanism, that is, matches different resilience enhancement strategies to execute tasks based on different acceleration distribution channels, can monitor and predict the edge computing node load and network quality in real time, and the embodiment of the present specification can automatically trigger algorithm switching or task migration strategy before the system performance exceeds the stable running state, not only enhances the system adaptive ability, transmission efficiency and security, but also guarantees the continuity of task processing and system stability. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present specification, the drawings required to be used in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present specification, and other drawings can be obtained by those skilled in the art without creative labor.

[0011] Figure 1This is a schematic diagram illustrating an application scenario of the edge computing-driven terminal data security acceleration distribution method provided in this manual.

[0012] Figure 2 This is a schematic diagram illustrating an application scenario of the peripheral environment monitoring substation system provided in an embodiment of the present invention.

[0013] Figure 3 This is a flowchart illustrating the edge computing-driven method for accelerating the secure distribution of terminal data, as provided in this manual.

[0014] Figure 4 This is a flowchart illustrating the process of identifying the resilience enhancement strategy for a given task, as provided in this specification.

[0015] Figure 5 This is a flowchart illustrating the process of determining the accelerated distribution channel provided in this manual.

[0016] Figure 6 This is a flowchart illustrating the algorithm for accelerating the determination of the primary target provided in this manual.

[0017] Figure 7 This is a flowchart illustrating the process of determining the degradation acceleration algorithm based on the task feature vector, as provided in this manual.

[0018] Figure 8 This is a flowchart illustrating the process of determining the corresponding collaborative node for the current edge computing node, as provided in this manual.

[0019] Figure 9 This is a schematic diagram of the edge computing-driven terminal data security acceleration distribution system provided in this manual. Detailed Implementation

[0020] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings.

[0021] The terms "first," "second," etc., in the description, claims, and accompanying drawings are used to distinguish different objects and not to describe a particular order. Furthermore, the term "comprising" and any variations thereof are intended to cover a non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such processes, methods, products, or apparatus.

[0022] The edge computing-driven terminal data security acceleration distribution method provided in several embodiments of this specification can be executed by the edge computing-driven terminal data security acceleration distribution system provided in the embodiments of this invention.

[0023] Before this specification elaborates on the edge computing-driven terminal data security acceleration distribution method in conjunction with one or more embodiments, it first introduces the application scenarios of this edge computing-driven terminal data security acceleration distribution method.

[0024] Please see Figure 1 , Figure 1 This is a schematic diagram illustrating an application scenario of the edge computing-driven terminal data security acceleration distribution method provided in this embodiment of the invention. In this embodiment, the edge computing-driven terminal data security acceleration distribution system 100 may include a plurality of IoT terminals 110, a plurality of edge computing nodes 120, etc. In this embodiment, the plurality of IoT terminals 110 are respectively connected to the plurality of edge computing nodes 120.

[0025] In this embodiment, the IoT terminal 110 can be a smart device that sends task requests to edge computing nodes via a network to obtain automated accelerated processing results (i.e., execution data corresponding to the task). The IoT terminal 110 may include industrial sensors, smart cameras, vehicle systems, medical monitoring equipment, and smart home appliances, etc.

[0026] For example, in this embodiment, based on the radiation detection scenario of the nuclear power plant's peripheral environment, the terminal data security acceleration distribution system may include at least one peripheral environment monitoring substation system, which is communicatively connected to several edge computing nodes 120. The peripheral environment monitoring substation system includes several substations, which may include 12 fixed monitoring substations, 2 mobile environmental monitoring vehicles, 1 online effluent monitoring system software interface, and 1 supervisory effluent laboratory software interface, etc. Each substation can be equipped with several IoT terminals. For example, a fixed monitoring substation may include a high-voltage ionization chamber detector, a continuous NAL spectrometer, an ultra-high flow aerosol sampler, a high flow aerosol sampler, an air iodine sampler, an air C14 sampler, an air H3 sampler, a multi-element meteorological monitoring instrument (wind direction, wind speed, rain sensing, rainfall, temperature, humidity, air pressure, etc.), a solar radiation meter, and other IoT terminals. The data acquisition interval can be set to 2 minutes, but can be adjusted to 1 minute in case of a system warning.

[0027] Please see Figure 2 , Figure 2This is a schematic diagram illustrating an application scenario of the peripheral environment monitoring substation system 140 provided in this embodiment of the invention. Each substation 150 in the peripheral environment monitoring substation system 140 is equipped with an industrial control computer (ICC). Each ICC includes front-end software, which is connected to each IoT terminal 110 via a data acquisition device 160. The ICC preprocesses the monitoring data collected by each IoT terminal 110, including data cleaning and normalization, data calibration, invalid data, range averages, maximum and minimum values, etc. Each ICC can set different tasks to be processed based on the preprocessed monitoring data and send task requests wirelessly to the edge computing node 120 via a wired connection or a wireless communication device 170 to obtain the corresponding execution data.

[0028] In this embodiment, the terminal data security acceleration distribution system 100 may further include a central server 130, and several edge computing nodes 120 are respectively communicatively connected to the central server 130. The several edge computing nodes 120 may include a current edge computing node 1200 and several collaborating nodes 1210 communicatively connected to the current edge computing node. The central server 110 may issue control tasks to the edge computing nodes 120 for the IoT terminal 110, and may also collect IoT terminal data uploaded by the edge computing nodes 120 through the edge computing nodes 120. The IoT terminal data includes, but is not limited to, device environmental data and device status data; the device environmental data may at least include temperature, humidity, pressure, illuminance, vibration, sound, and device geographic coordinates, and the device status data may at least include device battery level, CPU temperature, and on / off status.

[0029] In this embodiment, the IoT terminal 110 can establish a communication link with the current edge computing node and send a task request. The current edge computing node 1200 receives the task request and executes an edge acceleration distribution process according to the task request. The edge acceleration distribution process includes: obtaining the node feature vector and the task feature vector corresponding to the task request; determining the resilience enhancement strategy identifier of the task based on the node feature vector and the task feature vector; determining the acceleration distribution channel according to the resilience enhancement strategy identifier. The acceleration distribution channel is used to execute the task through different resilience enhancement strategies. The different resilience enhancement strategies include: when the acceleration distribution channel is the first acceleration channel, determining the primary target acceleration algorithm according to the preset edge acceleration algorithm library and the task feature vector, and executing the task based on the primary target acceleration algorithm; when the acceleration distribution channel is the second acceleration channel, determining the degradation acceleration algorithm according to the task feature vector, and executing the task based on the degradation acceleration algorithm; when the acceleration distribution channel is the collaborative acceleration channel, determining the collaborative node corresponding to the current edge computing node, and the collaborative node is used to execute the task according to the edge acceleration distribution process; the current edge computing node or the collaborative node obtains the execution data corresponding to the task and sends the execution data to the IoT terminal, etc.

[0030] It should be noted that, Figure 1 The schematic diagram of the edge computing-driven terminal data security acceleration distribution system shown is merely an example. The edge computing-driven terminal data security acceleration distribution system and scenario described in this embodiment are for the purpose of more clearly illustrating the technical solutions of this embodiment and do not constitute a limitation on the technical solutions provided by this embodiment. As those skilled in the art will know, with the evolution of edge computing-driven terminal data security acceleration distribution systems and the emergence of new scenarios, the technical solutions provided by this embodiment are also applicable to similar technical problems.

[0031] Please see Figure 3 , Figure 3 This is a flowchart illustrating the edge computing-driven terminal data security acceleration distribution method provided in an embodiment of the present invention. This edge computing-driven terminal data security acceleration distribution method can be... Figure 1 The terminal data security acceleration distribution system 100 shown is executed. This edge computing-driven terminal data security acceleration distribution method may include at least the following steps:

[0032] 200. Establish a communication link between the IoT terminal and the current edge computing node, and send a task request;

[0033] 210. The current edge computing node receives the task request and executes the edge acceleration distribution process according to the task request. The edge acceleration distribution process includes:

[0034] 2100. Obtain the node feature vector and the task feature vector corresponding to the task request;

[0035] 2110. Determine the resilience enhancement strategy identifier for a task based on node feature vectors and task feature vectors;

[0036] 2120. Determine the accelerated distribution channel based on the resilience enhancement strategy identifier. The accelerated distribution channel is used to execute tasks through different resilience enhancement strategies, including:

[0037] 2122. When the accelerated distribution channel is the first accelerated channel, the main target acceleration algorithm is determined according to the preset edge acceleration algorithm library and the task feature vector, and the task is executed based on the main target acceleration algorithm.

[0038] 2124. When the accelerated distribution channel is the second accelerated channel, the degradation acceleration algorithm is determined based on the task feature vector, and the task is executed based on the degradation acceleration algorithm.

[0039] 2126. When the accelerated distribution channel is the collaborative acceleration channel, determine the collaborative node corresponding to the current edge computing node. The collaborative node is used to execute tasks according to the edge accelerated distribution process.

[0040] 220. The current edge computing node or collaborative node obtains the execution data corresponding to the task and sends the execution data to the IoT terminal.

[0041] In this embodiment, the task request can be a data structure sent by the IoT terminal to the current edge computing node to request the current edge computing node to provide automated acceleration processing services. The task request may include task feature information, task load data, etc. Task feature information can be data recording the attributes and requirements of the task, and may include task type, task identifier, data return format, etc. The task identifier can be the ID corresponding to the task request; the data return format can be the processing result returned by the specified edge computing node after executing the task in a preset data format, which may include JSON data, compressed images, alarm commands, etc. The task load data can be the raw data sent by the IoT terminal to the edge computing node and that needs to be processed by the edge computing node's acceleration algorithm.

[0042] For example, in a scenario where the IoT terminal is a smart camera, the IoT terminal sends a task request for object recognition to the current edge computing node: to perform object recognition on image A and return the recognized object type in JSON data format within 200ms. Here, the task type corresponding to the task request is object recognition, the data return format is JSON data format, and the task payload data is image A.

[0043] In this embodiment, the task feature vector can be the feature vector obtained after feature extraction of the task request. The task feature vector extraction process can be performed at the current edge computing node. Alternatively, in this embodiment, the task feature vector can be extracted by the IoT terminal and sent to the current edge computing node. In this embodiment, the node feature vector is the feature vector determined by the current edge computing node based on its own state information.

[0044] In this embodiment, the resilience enhancement strategy identifier may include computing resource sufficiency and network resource sufficiency. Computing resource sufficiency can be data obtained by subtracting the computing resource capacity from the task's corresponding computing resource requirement threshold and the node load rate of the edge computing node, based on a preset computing resource capacity. Network resource sufficiency can be data obtained by subtracting the task's corresponding network latency requirement threshold and the predicted network latency of the edge computing node.

[0045] In this embodiment, the current edge computing node can execute an edge acceleration distribution process based on a task request. This involves acquiring node feature vectors and task feature vectors, and then determining the resilience enhancement strategy identifier for the task based on these vectors. By introducing task feature vectors and node feature vectors representing node state information to jointly determine the resilience enhancement strategy identifier, this embodiment achieves context awareness and precision in executing tasks with different resilience enhancement strategies. Specifically, it enables accurate prediction of whether the current edge computing node environment is sufficient to guarantee successful task execution. Furthermore, this embodiment can match different acceleration distribution channels based on the resilience enhancement strategy identifier. These channels include a first acceleration channel, a second acceleration channel, or a collaborative acceleration channel. The first and second acceleration channels correspond to the main acceleration channel and the degraded acceleration channel, respectively. The main acceleration channel matches the primary target acceleration algorithm based on a preset edge acceleration algorithm library and the task feature vector. The degraded acceleration channel executes the task based on a degraded acceleration algorithm, while the collaborative acceleration channel updates the edge acceleration distribution process based on the collaborative node corresponding to the current edge computing node to execute the task. Then, the current edge computing node or collaborative node acquires the execution data corresponding to the task and sends it to the IoT terminal.

[0046] The embodiments in this specification can automatically accelerate the distribution of tasks on IoT terminals. Specifically, by introducing task feature vectors and a preset edge acceleration algorithm library, and matching different acceleration distribution channels according to the resilience enhancement strategy identifier, the optimal acceleration algorithm is dynamically matched for tasks with different characteristics. This achieves precise matching of computing resources and task requirements, significantly improving task processing efficiency and resource utilization.

[0047] The embodiments in this specification also establish a system resilience guarantee mechanism, that is, matching different resilience enhancement strategies to different acceleration distribution channels to execute tasks. This enables real-time monitoring and prediction of edge computing node load and network quality. The embodiments in this specification can automatically trigger algorithm switching or task migration strategies before the system performance exceeds the stable operating state. This not only enhances the system's adaptability, transmission efficiency and security, but also ensures the continuity of task processing and system stability.

[0048] In some embodiments, resilience enhancement policy identifiers may include computational resource sufficiency and network resource sufficiency, etc. See also Figure 4 , Figure 4 This is a flowchart illustrating the process of determining the resilience enhancement strategy identifier for a task according to an embodiment of the present invention. Determining the resilience enhancement strategy identifier for a task based on node feature vectors and task feature vectors includes:

[0049] 300. Based on a preset task resource requirement database, determine the corresponding computing resource requirement threshold and network latency requirement threshold according to the task feature vector;

[0050] 310. Determine the node feature vector corresponding to the current time step;

[0051] 320. Based on the node feature vector corresponding to the current time, determine the node load rate, round-trip latency, and latency jitter value of the current edge computing node in the next time.

[0052] 330. Determine the predicted network latency of the current edge computing node at the next moment based on the round-trip latency and latency jitter value;

[0053] 340. Based on the preset computing resource capacity, the computing resource capacity is compared with the computing resource demand threshold corresponding to the task and the node load rate of the edge computing node at the next moment to obtain the computing resource sufficiency.

[0054] 350. The network resource sufficiency is obtained by performing difference processing on the network latency requirement threshold corresponding to the task and the network latency prediction value of the edge computing node at the next moment.

[0055] In this embodiment, the task resource requirement database can store task feature vectors corresponding to any task type, computing resource requirement thresholds and network latency requirement thresholds corresponding to any task type, and mapping relationships between task feature vectors corresponding to any task type and computing resource requirement thresholds and network latency requirement thresholds corresponding to any task type.

[0056] In this embodiment, the current time can be the time when the current edge computing node receives the task request, and the next time can be the time after the current time determined based on the time period.

[0057] In this embodiment, the node load rate can be a normalized value determined based on the overall resource pressure of the edge computing node. The node load rate can range from 0 to 1, where a node load rate of 0 indicates that the edge computing node is idle, and a node load rate of 2 indicates that the edge computing node's resources are saturated, meaning it cannot effectively handle new tasks sent by IoT terminals. The node load rate can be calculated by normalizing the CPU utilization, memory utilization, and storage I / O load rate of the edge computing node, and then using a pre-trained weighted fusion model. The storage I / O load rate can be determined by querying the storage device utilization rate corresponding to the edge computing node and normalizing it.

[0058] In this embodiment, the node feature vector is determined by the node operation characteristics corresponding to the edge computing node. Node operation characteristics may include node operating status, network quality data, and task queues. Node operating status may include CPU temperature, utilization, memory usage, and remaining battery power. Network quality data may include bandwidth, round-trip time, latency jitter, and packet loss rate. The task queue may include the number of queued tasks and task type distribution. Round-trip time is the total time required for data to travel from the sender to the receiver and back through the network. The sender can be an IoT terminal or an edge computing node, and the receiver can also be an edge computing node or an IoT terminal.

[0059] In this embodiment, the latency jitter value can be obtained by continuously sending several probe data packets from the current edge computing node to other edge computing nodes within a preset time window, recording the round-trip latency of each data packet, obtaining a round-trip latency series, and calculating the standard deviation of the round-trip latency series.

[0060] In some embodiments, before determining the computational resource requirement threshold and network latency requirement threshold corresponding to a task based on a preset task resource requirement database and task feature vectors, the process includes: acquiring historical task execution data, which includes a set of historical task feature vectors corresponding to several historical tasks and a historical system performance dataset; classifying several historical tasks according to the set of historical task feature vectors to obtain several task types; determining, based on any task type, a peak processor utilization dataset and a peak network latency dataset corresponding to any task type according to the historical system performance dataset; determining the computational resource requirement threshold and network latency requirement threshold corresponding to any task type according to the peak processor utilization dataset and the peak network latency dataset, respectively; and establishing a mapping relationship between the task feature vectors corresponding to any task type and the computational resource requirement thresholds and network latency requirement thresholds corresponding to any task type, and storing this mapping relationship as a task resource requirement database.

[0061] In this embodiment, the historical task feature vector is the task feature vector determined by the edge computing node at a historical moment based on the historical task request sent by the IoT terminal, and the historical system performance data is the system performance data corresponding to when the edge computing node successfully executes the historical task.

[0062] In some embodiments, please refer to Figure 5 , Figure 5 This is a schematic diagram of the process for determining an accelerated distribution channel according to an embodiment of the present invention. Determining the accelerated distribution channel based on the resilience enhancement strategy identifier includes:

[0063] 400. Determine the first sufficiency threshold and the second sufficiency threshold;

[0064] 410. When the computing resource sufficiency is not less than the first sufficiency threshold and the network resource sufficiency is not less than the second sufficiency threshold, the accelerated distribution channel is determined as the first accelerated channel.

[0065] 420. When the computing resource sufficiency is less than the first sufficiency threshold and the network resource sufficiency is not less than the second sufficiency threshold, the accelerated distribution channel is determined as the second accelerated channel.

[0066] 430. When the network resource sufficiency is less than the second sufficiency threshold, the accelerated distribution channel is determined to be a collaborative acceleration channel.

[0067] In this embodiment, different acceleration distribution channels can be matched according to the resilience enhancement strategy identifier. The acceleration distribution channel can be a first acceleration channel, a second acceleration channel, or a collaborative acceleration channel, etc. The first acceleration channel and the second acceleration channel correspond to the main acceleration channel and the degraded acceleration channel, respectively. The main acceleration channel can match the main target acceleration algorithm according to the preset edge acceleration algorithm library and task feature vector. The degraded acceleration channel can execute the task based on the degraded acceleration algorithm. The collaborative acceleration channel can update the edge acceleration distribution process based on the collaborative node corresponding to the current edge computing node to execute the task. Then, the current edge computing node or collaborative node obtains the execution data corresponding to the task and sends the execution data to the IoT terminal.

[0068] In this embodiment, the first sufficiency threshold can be a sufficiency threshold set according to the sufficiency of computing resources, for example, the first sufficiency threshold can be 0; the second sufficiency threshold can be a sufficiency threshold set according to the sufficiency of network resources, for example, the second sufficiency threshold can be 0.

[0069] In some embodiments, please refer to Figure 6 , Figure 6 This is a flowchart illustrating the process of determining the primary target acceleration algorithm according to an embodiment of the present invention. When the acceleration distribution channel is the first acceleration channel, the primary target acceleration algorithm is determined based on a preset edge acceleration algorithm library and task feature vectors, including:

[0070] 500. Based on the set of historical task feature vectors, determine the Euclidean distance between the task feature vector and each historical task feature vector in the set of historical task feature vectors;

[0071] 510. Determine the set of similar historical task instances corresponding to the task feature vectors based on Euclidean distance;

[0072] 520. Determine any edge acceleration algorithm in the edge acceleration algorithm library, and determine an instance subset based on the set of similar historical task instances. The instance subset is a subset of similar historical task instances that execute historical tasks according to any edge acceleration algorithm.

[0073] 530. Determine the prediction processing latency and prediction processing accuracy of any edge acceleration algorithm execution task based on an instance subset;

[0074] 540. Based on the prediction processing latency and prediction processing accuracy of all edge acceleration algorithms, candidate algorithms are determined. The candidate algorithms are the main target acceleration algorithms corresponding to the task.

[0075] In this embodiment, the edge acceleration algorithm library stores several edge acceleration algorithms. These edge acceleration algorithms are software modules located on resource-constrained edge computing nodes and designed to achieve IoT terminal task performance metrics (such as low latency, low power consumption, and high throughput). These software modules can be independently called and managed. Edge acceleration algorithms may include data preprocessing acceleration algorithms, computationally intensive task optimization algorithms, and data transmission optimization algorithms.

[0076] In some embodiments, the instance subset includes several similar historical task instances. Determining the predicted processing latency and predicted processing accuracy of any edge acceleration algorithm executing a task based on the instance subset includes: determining the weight of any similar historical task instance in the instance subset based on a preset bandwidth parameter and Euclidean distance; obtaining the weights of all similar historical task instances in the instance subset; obtaining the actual latency and actual accuracy of all similar historical task instances in the instance subset when executing historical tasks with any edge acceleration algorithm; and determining the predicted processing latency and predicted processing accuracy of any edge acceleration algorithm executing a task based on the actual latency, actual accuracy, and weight of each similar historical task instance.

[0077] In this embodiment, the weight of any similar historical task instance is an e-exponential function based on a preset bandwidth parameter. When the system calculates the weight, it can first calculate the Euclidean distance between the task feature vector and the historical task feature vector corresponding to any similar historical task instance, and then take the negative number of the Euclidean distance. The ratio between the negative number of the Euclidean distance and the preset bandwidth parameter is used as the exponent of the e-exponential function.

[0078] In this embodiment, the actual latency of any similar historical task instance in the instance subset executing a historical task using any edge acceleration algorithm can be measured by recording the time difference between calling the algorithm and receiving the complete output result of the algorithm. The actual accuracy of any similar historical task instance in the instance subset executing a historical task using any edge acceleration algorithm can be obtained by comparing the output result of the algorithm with a preset result.

[0079] In this embodiment, the system can determine the predicted processing latency of any edge acceleration algorithm execution task based on the actual latency and weights corresponding to each similar historical task instance. The predicted processing latency is... N is the number of all similar historical task instances in the instance subset. Let be the actual latency when the i-th similar historical task instance executes the historical task using any edge acceleration algorithm. Let be the weight of the i-th similar historical task instance.

[0080] In this embodiment, the system can also determine the prediction processing accuracy of any edge acceleration algorithm task based on the actual accuracy and weights corresponding to each similar historical task instance, with a prediction processing latency of [missing information]. , Let be the true accuracy when the i-th similar historical task instance is executed using any edge acceleration algorithm.

[0081] In some embodiments, the preset edge acceleration algorithm library includes several edge acceleration algorithms and their corresponding static resource overhead indices. See also... Figure 7 , Figure 7 This is a flowchart illustrating the process of determining a degradation acceleration algorithm based on a task feature vector, as provided in an embodiment of the present invention. When the acceleration distribution channel is the second acceleration channel, determining the degradation acceleration algorithm based on the task feature vector includes:

[0082] 600. Determine a candidate set of degradation algorithms from the preset edge acceleration algorithm library based on the computing resource requirement threshold corresponding to the task;

[0083] 610. Obtain the accuracy threshold corresponding to the task, and the prediction processing accuracy of each edge acceleration algorithm in the candidate set of degradation algorithms;

[0084] 620. When the prediction accuracy of a certain edge acceleration algorithm in the candidate set of downgrade algorithms is lower than the accuracy threshold, the edge acceleration algorithm is filtered out from the candidate set of downgrade algorithms to obtain the filtered candidate set of downgrade algorithms.

[0085] 630. Determine the edge acceleration algorithm with the smallest static resource overhead index in the filtered candidate set of degradation algorithms. The edge acceleration algorithm with the smallest static resource overhead index is the degradation acceleration algorithm corresponding to the second acceleration channel.

[0086] In this embodiment, the preset edge acceleration algorithm library pre-stores a corresponding static resource cost index for each edge acceleration algorithm. The static resource cost index can be determined by quantifying and normalizing the resource consumption of the algorithm under benchmark testing.

[0087] In this embodiment, the accuracy threshold corresponding to the task can be the minimum accuracy tolerance required by the task. The system in this embodiment can access a task type and accuracy requirement mapping table, which stores the mapping relationship between different task types and their corresponding minimum accuracy tolerances. For example, for a task of type "environmental monitoring data reporting," its minimum accuracy tolerance is 0.85; for a task of type "real-time video structured analysis," its minimum accuracy tolerance is 0.95; and for a task of type "equipment anomaly detection," its minimum accuracy tolerance is 0.99.

[0088] In some embodiments, determining a set of candidate degradation algorithms from a preset edge acceleration algorithm library based on a computing resource requirement threshold corresponding to the task includes: multiplying a preset degradation coefficient with a computing resource requirement threshold corresponding to the task based on a preset degradation coefficient to obtain a degradation resource upper limit; determining a set of candidate degradation algorithms from the preset edge acceleration algorithm library based on the degradation resource upper limit, wherein the static resource overhead index of each edge acceleration algorithm in the degradation algorithm candidate set is less than the degradation resource upper limit.

[0089] In this embodiment, the system can multiply the computing resource requirement threshold corresponding to the task with a preset degradation coefficient to obtain the upper limit of the degradation resources. The preset degradation coefficient can be set to be greater than 0 and less than 1. Then, the system can select edge acceleration algorithms from the preset edge acceleration algorithm library whose static resource overhead index is less than the upper limit of the degradation resources, and form them into a candidate set of degradation algorithms.

[0090] In some embodiments, please refer to Figure 8 , Figure 8 This is a schematic diagram of the process for determining the corresponding collaborative node of the current edge computing node according to an embodiment of the present invention. When the accelerated distribution channel is a collaborative acceleration channel, the collaborative node corresponding to the current edge computing node is determined. The collaborative node is used to execute tasks according to the edge accelerated distribution process, including:

[0091] 700. Determine the node distance between other edge computing nodes and the current edge computing node;

[0092] 710. Determine the set of nearest neighbors of the current edge computing node based on the node distance;

[0093] 720. The current edge computing node sends test data to each of its nearest neighbor nodes, and determines the network quality quantification value and resource sufficiency quantification value corresponding to each nearest neighbor node based on the network test results of the test data.

[0094] 730. Perform weighted fusion processing on the network quality quantification value and resource sufficiency quantification value corresponding to each nearest neighbor node to obtain the node evaluation value corresponding to each nearest neighbor node.

[0095] 740. Determine the corresponding collaborative node for the current edge computing node from among its nearest neighbor nodes based on the node evaluation values;

[0096] 750. The collaborative node is the next current edge computing node to execute the task and executes the task according to the edge acceleration distribution process.

[0097] In this embodiment, the node distance is determined by the coordinate distance between nodes and the network topology distance. The network quality quantification value may include at least bandwidth, round-trip time, and latency jitter value; the resource sufficiency quantification value may include at least CPU load rate, available memory, total memory, and remaining power. The CPU load rate can be the current average CPU utilization of neighboring nodes, the available memory can be the current available memory capacity of neighboring nodes, and the total memory can be the total physical memory capacity of neighboring nodes.

[0098] In this embodiment, the latency jitter value can be obtained by continuously sending several probe data packets from the current edge computing node to other edge computing nodes within a preset time window, recording the round-trip latency of each data packet, obtaining a round-trip latency series, and calculating the standard deviation of the round-trip latency series.

[0099] In this embodiment, the current edge computing node can execute an edge acceleration distribution process based on a task request. This involves acquiring node feature vectors and task feature vectors, and then determining the resilience enhancement strategy identifier for the task based on these vectors. By introducing task feature vectors and node feature vectors representing node state information to jointly determine the resilience enhancement strategy identifier, this embodiment achieves context awareness and precision in executing tasks with different resilience enhancement strategies. Specifically, it enables accurate prediction of whether the current edge computing node environment is sufficient to guarantee successful task execution. Furthermore, this embodiment can match different acceleration distribution channels based on the resilience enhancement strategy identifier. These channels include a first acceleration channel, a second acceleration channel, or a collaborative acceleration channel. The first and second acceleration channels correspond to the main acceleration channel and the degraded acceleration channel, respectively. The main acceleration channel matches the primary target acceleration algorithm based on a preset edge acceleration algorithm library and the task feature vector. The degraded acceleration channel executes the task based on a degraded acceleration algorithm, while the collaborative acceleration channel updates the edge acceleration distribution process based on the collaborative node corresponding to the current edge computing node to execute the task. Then, the current edge computing node or collaborative node acquires the execution data corresponding to the task and sends it to the IoT terminal.

[0100] The embodiments in this specification can automatically accelerate the distribution of tasks on IoT terminals. Specifically, by introducing task feature vectors and a preset edge acceleration algorithm library, and matching different acceleration distribution channels according to the resilience enhancement strategy identifier, the optimal acceleration algorithm is dynamically matched for tasks with different characteristics. This achieves precise matching of computing resources and task requirements, significantly improving task processing efficiency and resource utilization.

[0101] The embodiments in this specification also establish a system resilience guarantee mechanism, that is, matching different resilience enhancement strategies to different acceleration distribution channels to execute tasks. This enables real-time monitoring and prediction of edge computing node load and network quality. The embodiments in this specification can automatically trigger algorithm switching or task migration strategies before the system performance exceeds the stable operating state. This not only enhances the system's adaptability, transmission efficiency and security, but also ensures the continuity of task processing and system stability.

[0102] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0103] Please see Figure 9 , Figure 9 This is a schematic diagram of the structure of an edge computing-driven terminal data security acceleration distribution system provided in the embodiments of this specification.

[0104] like Figure 9 As shown, this edge computing-driven terminal data security acceleration distribution system may include at least an IoT terminal 800, a current edge computing node 810, and a collaborative node 820, wherein:

[0105] The IoT terminal 800 is used to establish a communication link with the current edge computing node and send task requests;

[0106] The current edge computing node 810 is used to receive task requests and execute the edge acceleration distribution process according to the task requests. The current edge computing node includes a feature acquisition module 8100, an identifier determination module 8110, and a channel determination module 8120.

[0107] The feature acquisition module 8100 is used to acquire node feature vectors and task feature vectors corresponding to task requests;

[0108] The identifier determination module 8110 is used to determine the resilience enhancement strategy identifier of the task based on the node feature vector and the task feature vector;

[0109] The channel determination module 8120 is used to determine the accelerated distribution channel based on the resilience enhancement strategy identifier. The accelerated distribution channel is used to execute tasks through different resilience enhancement strategies, including:

[0110] When the accelerated distribution channel is the first accelerated channel, the main target acceleration algorithm is determined according to the preset edge acceleration algorithm library and task feature vector, and the task is executed based on the main target acceleration algorithm;

[0111] When the accelerated distribution channel is the second accelerated channel, the degraded acceleration algorithm is determined based on the task feature vector, and the task is executed based on the degraded acceleration algorithm;

[0112] When the accelerated distribution channel is the collaborative acceleration channel, the collaborative node 820 corresponding to the current edge computing node 810 is determined. The collaborative node 820 is used to execute tasks according to the edge accelerated distribution process.

[0113] The current edge computing node 810 or collaborative node 820 obtains the execution data corresponding to the task and sends the execution data to the IoT terminal 800.

[0114] In some embodiments, the resilience enhancement strategy identifier includes computing resource sufficiency and network resource sufficiency. The identifier determination module 8110 includes an identifier determination submodule, which is used to: determine the computing resource requirement threshold and network latency requirement threshold corresponding to the task based on a preset task resource requirement database and the task feature vector; determine the node feature vector corresponding to the current time; determine the node load rate, round-trip latency, and latency jitter value of the current edge computing node in the next time based on the node feature vector corresponding to the current time; determine the network latency prediction value of the current edge computing node in the next time based on the round-trip latency and latency jitter value; perform difference processing on the computing resource capacity and the computing resource requirement threshold and the node load rate of the edge computing node in the next time based on a preset computing resource capacity to obtain the computing resource sufficiency; and perform difference processing on the network latency requirement threshold and the network latency prediction value of the current edge computing node in the next time to obtain the network resource sufficiency.

[0115] In some embodiments, the identifier determination submodule further includes a database establishment module, which is configured to: acquire historical task execution data, including a set of historical task feature vectors corresponding to several historical tasks and a historical system performance dataset; classify several historical tasks according to the set of historical task feature vectors to obtain several task types; determine, based on any task type, a peak processor utilization dataset and a peak network latency dataset corresponding to any task type according to the historical system performance dataset; determine, based on the peak processor utilization dataset and the peak network latency dataset, a computing resource requirement threshold and a network latency requirement threshold corresponding to any task type, respectively; and establish a mapping relationship between the task feature vectors corresponding to any task type and the computing resource requirement thresholds and network latency thresholds corresponding to any task type, and store them as a task resource requirement database.

[0116] In some embodiments, the historical task feature vector is a task feature vector determined by the edge computing node at a historical moment based on the historical task request sent by the IoT terminal, and the historical system performance data is the system performance data corresponding to when the edge computing node successfully executes the historical task.

[0117] In some embodiments, the channel determination module 8120 includes a determination module, which is configured to: determine a first sufficiency threshold and a second sufficiency threshold; determine the accelerated distribution channel as a first accelerated channel when the computing resource sufficiency is not less than the first sufficiency threshold and the network resource sufficiency is not less than the second sufficiency threshold; determine the accelerated distribution channel as a second accelerated channel when the computing resource sufficiency is less than the first sufficiency threshold and the network resource sufficiency is not less than the second sufficiency threshold; and determine the accelerated distribution channel as a collaborative accelerated channel when the network resource sufficiency is less than the second sufficiency threshold.

[0118] In some embodiments, the channel determination module 8120 includes a first channel determination module, which is configured to: determine the Euclidean distance between the task feature vector and each historical task feature vector in the historical task feature vector set based on the historical task feature vector set; determine a set of similar historical task instances corresponding to the task feature vector based on the Euclidean distance; determine any edge acceleration algorithm in the edge acceleration algorithm library, and determine an instance subset based on the set of similar historical task instances, wherein the instance subset is a subset of similar historical task instances that execute a historical task according to any edge acceleration algorithm; determine the prediction processing latency and prediction processing accuracy of the task executed by any edge acceleration algorithm based on the instance subset; and determine candidate algorithms based on the prediction processing latency and prediction processing accuracy of the task executed by all edge acceleration algorithms, wherein the candidate algorithms are the main target acceleration algorithms corresponding to the task.

[0119] In some embodiments, the instance subset includes several similar historical task instances, and the first channel determination module includes a prediction module. The prediction module is used to: determine the weight of any similar historical task instance in the instance subset based on a preset bandwidth parameter and Euclidean distance; obtain the weights of all similar historical task instances in the instance subset; obtain the actual latency and actual accuracy of all similar historical task instances in the instance subset when executing historical tasks with any edge acceleration algorithm; and determine the prediction processing latency and prediction processing accuracy of any edge acceleration algorithm for executing tasks based on the actual latency, actual accuracy, and weights of each similar historical task instance.

[0120] In some embodiments, the preset edge acceleration algorithm library includes several edge acceleration algorithms and static resource overhead indices corresponding to the several edge acceleration algorithms; the channel determination module 8120 includes a degradation module, which is used to: determine a set of degradation algorithm candidates from the preset edge acceleration algorithm library according to the computing resource requirement threshold corresponding to the task; obtain the accuracy threshold corresponding to the task and the prediction processing accuracy of each edge acceleration algorithm in the degradation algorithm candidate set for executing the task; when the prediction processing accuracy corresponding to a certain edge acceleration algorithm in the degradation algorithm candidate set is lower than the accuracy threshold, filter the certain edge acceleration algorithm from the degradation algorithm candidate set to obtain a filtered set of degradation algorithm candidates; determine the edge acceleration algorithm with the smallest static resource overhead index in the filtered set of degradation algorithm candidates, and the edge acceleration algorithm with the smallest static resource overhead index is the degradation acceleration algorithm corresponding to the second acceleration channel.

[0121] In some embodiments, the degradation module includes a candidate set determination module, which is used to: multiply the preset degradation coefficient with the computing resource requirement threshold corresponding to the task based on the preset degradation coefficient to obtain the degradation resource upper limit value; and determine a degradation algorithm candidate set from a preset edge acceleration algorithm library according to the degradation resource upper limit value, wherein the static resource overhead index of each edge acceleration algorithm in the degradation algorithm candidate set is less than the degradation resource upper limit value.

[0122] In some embodiments, the channel determination module 8120 includes a coordination module, which is configured to: determine the node distance between other edge computing nodes and the current edge computing node, wherein the node distance is determined by the coordinate distance between nodes and the network topology distance; determine the set of nearest neighbors corresponding to the current edge computing node based on the node distance; send test data to each nearest neighbor in the set of nearest neighbors respectively, and determine the network quality quantification value and resource adequacy quantification value corresponding to each nearest neighbor based on the network test results of the test data; perform weighted fusion processing on the network quality quantification value and resource adequacy quantification value corresponding to each nearest neighbor to obtain the node evaluation value corresponding to each nearest neighbor; determine the coordination node corresponding to the current edge computing node from each nearest neighbor based on the node evaluation value; the coordination node is the next current edge computing node to execute the task, and executes the task according to the edge acceleration distribution process.

[0123] Based on the edge computing-driven terminal data security acceleration distribution system described in several embodiments of this specification, it can be seen that the embodiments of this specification can automatically accelerate the distribution of IoT terminal tasks. Specifically, by introducing task feature vectors and a preset edge acceleration algorithm library, and matching different acceleration distribution channels according to resilience enhancement strategy identifiers, the optimal acceleration algorithm is dynamically matched for tasks with different characteristics. This achieves precise matching of computing resources and task requirements, significantly improving task processing efficiency and resource utilization. Furthermore, the embodiments of this specification also establish a system resilience guarantee mechanism, matching different resilience enhancement strategies to different acceleration distribution channels for task execution. This allows for real-time monitoring and prediction of edge computing node load and network quality. Before system performance exceeds a stable operating state, the embodiments of this specification can automatically trigger algorithm switching or task migration strategies, enhancing not only the system's adaptability, transmission efficiency, and security, but also ensuring the continuity of task processing and system stability.

[0124] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. In particular, the embodiment of the edge computing-driven terminal data security acceleration distribution system is relatively simple in description because it is fundamentally similar to the embodiment of the edge computing-driven terminal data security acceleration distribution method. Relevant parts can be referred to the description of the method embodiment.

[0125] This specification also provides a computer-readable storage medium storing instructions that, when executed on a computer or processor, cause the computer or processor to perform the above-described instructions. Figures 2 to 7One or more steps in the illustrated embodiment. If the constituent modules of the above-described electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0126] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this specification is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., Digital Versatile Discs (DVDs)), or semiconductor media (e.g., Solid State Disks (SSDs)).

[0127] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks. Unless otherwise specified, the technical features of this embodiment and its implementation can be combined arbitrarily.

[0128] The above embodiments are merely preferred embodiments described in this specification and are not intended to limit the scope of this specification. Any modifications and improvements made by those skilled in the art to the technical solutions of this specification without departing from the spirit of this specification should fall within the protection scope defined by the claims of this specification.

Claims

1. An edge computing-driven method for accelerating secure distribution of terminal data, including: The IoT terminal establishes a communication link with the current edge computing node and sends a task request; The current edge computing node receives the task request and executes an edge acceleration distribution process according to the task request. The edge acceleration distribution process includes: Obtain the node feature vector and the task feature vector corresponding to the task request; The resilience enhancement strategy identifier of the task is determined based on the node feature vector and the task feature vector; An accelerated distribution channel is determined based on the resilience enhancement strategy identifier. The accelerated distribution channel is used to execute the task through different resilience enhancement strategies, including: When the accelerated distribution channel is the first accelerated channel, the main target acceleration algorithm is determined according to the preset edge acceleration algorithm library and the task feature vector, and the task is executed based on the main target acceleration algorithm; When the accelerated distribution channel is the second accelerated channel, a degradation acceleration algorithm is determined based on the task feature vector, and the task is executed based on the degradation acceleration algorithm; When the accelerated distribution channel is a collaborative acceleration channel, a collaborative node corresponding to the current edge computing node is determined, and the collaborative node is used to execute the task according to the edge accelerated distribution process. The current edge computing node or the collaborating node obtains the execution data corresponding to the task and sends the execution data to the IoT terminal; The resilience enhancement strategy identifier includes computational resource sufficiency and network resource sufficiency. The step of determining the task's resilience enhancement strategy identifier based on the node feature vector and the task feature vector includes: Based on a preset task resource requirement database, the computing resource requirement threshold and network latency requirement threshold corresponding to the task are determined according to the task feature vector. Determine the node feature vector corresponding to the current time. Based on the node feature vector corresponding to the current time, determine the node load rate, round-trip latency, and latency jitter value of the current edge computing node in the next time. The network latency prediction value of the current edge computing node at the next moment is determined based on the round-trip latency and the latency jitter value. Based on the preset computing resource capacity, the computing resource capacity is compared with the computing resource demand threshold corresponding to the task and the node load rate of the edge computing node at the next moment to obtain the computing resource sufficiency. The network resource sufficiency is obtained by performing a difference processing on the network latency requirement threshold corresponding to the task and the network latency prediction value of the current edge computing node in the next moment.

2. The method according to claim 1, before determining the computing resource requirement threshold and network latency requirement threshold corresponding to the task based on the task feature vector using a preset task resource requirement database, the method includes: Acquire historical task execution data, which includes a set of historical task feature vectors corresponding to several historical tasks and a historical system performance dataset. Based on the set of historical task feature vectors, the historical tasks are classified to obtain several task types; Based on any task type, determine the peak processor utilization dataset and peak network latency dataset corresponding to the historical system performance dataset. Based on the processor utilization peak dataset and the network latency peak dataset, determine the computing resource requirement threshold and the network latency requirement threshold corresponding to any task type, respectively. Establish a mapping relationship between the task feature vector corresponding to any task type and the computing resource requirement threshold and network latency requirement threshold corresponding to any task type, and store it as the task resource requirement database; The historical task feature vector is the task feature vector determined by the edge computing node at a historical moment based on the historical task request sent by the IoT terminal, and the historical system performance data is the system performance data corresponding to when the edge computing node successfully executes the historical task.

3. The method according to claim 2, wherein determining the accelerated distribution channel based on the resilience enhancement strategy identifier comprises: Determine the first adequacy threshold and the second adequacy threshold; When the computing resource sufficiency is not less than the first sufficiency threshold and the network resource sufficiency is not less than the second sufficiency threshold, the accelerated distribution channel is determined to be the first accelerated channel. When the computing resource sufficiency is less than the first sufficiency threshold and the network resource sufficiency is not less than the second sufficiency threshold, the accelerated distribution channel is determined to be the second accelerated channel. When the network resource sufficiency is less than the second sufficiency threshold, the accelerated distribution channel is determined to be a collaborative acceleration channel.

4. The method according to claim 1, wherein when the accelerated distribution channel is a first accelerated channel, determining the main target acceleration algorithm based on a preset edge acceleration algorithm library and the task feature vector includes: Based on the set of historical task feature vectors, determine the Euclidean distance between the task feature vector and each historical task feature vector in the set of historical task feature vectors. The set of similar historical task instances corresponding to the task feature vector is determined based on the Euclidean distance. Determine any edge acceleration algorithm in the edge acceleration algorithm library, and determine an instance subset based on the set of similar historical task instances, wherein the instance subset is a subset of similar historical task instances that execute historical tasks according to the arbitrary edge acceleration algorithm; Based on the aforementioned subset of instances, determine the prediction processing latency and prediction processing accuracy of any edge acceleration algorithm executing the task. Candidate algorithms are determined based on the prediction processing latency and prediction processing accuracy of all edge acceleration algorithms for the task, and the candidate algorithms are the main target acceleration algorithms corresponding to the task.

5. The method according to claim 4, wherein the instance subset includes several similar historical task instances, and the step of determining the prediction processing latency and prediction processing accuracy of any one edge acceleration algorithm executing the task based on the instance subset includes: Based on preset bandwidth parameters, the weight of any similar historical task instance in the instance subset is determined according to Euclidean distance; Obtain the weights of all similar historical task instances in the subset of instances; Obtain the actual latency and actual accuracy of all similar historical task instances in the instance subset when executing the historical task using any of the edge acceleration algorithms; Based on the actual latency, actual accuracy, and weights of each similar historical task instance, the prediction processing latency and prediction processing accuracy of any edge acceleration algorithm executing the task are determined.

6. The method according to claim 3, wherein the preset edge acceleration algorithm library includes a plurality of edge acceleration algorithms and static resource overhead indices corresponding to the plurality of edge acceleration algorithms respectively; When the accelerated distribution channel is the second accelerated channel, the degradation acceleration algorithm is determined based on the task feature vector, including: Based on the computing resource requirement threshold corresponding to the task, a set of candidate degradation algorithms is determined from a preset edge acceleration algorithm library; Obtain the accuracy threshold corresponding to the task, and the prediction processing accuracy of each edge acceleration algorithm in the candidate set of the degradation algorithm for executing the task; When the prediction accuracy of a certain edge acceleration algorithm in the candidate set of degradation algorithms is lower than the accuracy threshold, the certain edge acceleration algorithm is filtered out from the candidate set of degradation algorithms to obtain a filtered candidate set of degradation algorithms. The edge acceleration algorithm with the smallest static resource overhead index in the filtered candidate set of degradation algorithms is determined, and the edge acceleration algorithm with the smallest static resource overhead index is the degradation acceleration algorithm corresponding to the second acceleration channel.

7. The method according to claim 6, wherein determining a candidate set of degradation algorithms from a preset edge acceleration algorithm library based on the computing resource requirement threshold corresponding to the task includes: Based on a preset degradation coefficient, the preset degradation coefficient is multiplied by the computing resource requirement threshold corresponding to the task to obtain the upper limit of degradation resources. Based on the degradation resource cap, a set of degradation algorithm candidates is determined from a preset edge acceleration algorithm library, wherein the static resource overhead index of each edge acceleration algorithm in the degradation algorithm candidate set is less than the degradation resource cap.

8. The method according to claim 1, wherein when the accelerated distribution channel is a collaborative acceleration channel, determining the collaborative node corresponding to the current edge computing node, the collaborative node being used to execute the task according to the edge accelerated distribution process, includes: Determine the node distance between other edge computing nodes and the current edge computing node, wherein the node distance is determined by the coordinate distance between nodes and the network topology distance; The set of nearest neighbors corresponding to the current edge computing node is determined based on the node distance; The current edge computing node sends test data to each of the nearest neighbor nodes in the nearest neighbor node set, and determines the network quality quantification value and resource sufficiency quantification value corresponding to each nearest neighbor node based on the network test results of the test data. The network quality quantification value and resource sufficiency quantification value corresponding to each neighbor node are weighted and fused to obtain the node evaluation value corresponding to each neighbor node. Based on the node evaluation values, determine the collaborating node corresponding to the current edge computing node from each of the neighboring nodes; The collaborative node is the next current edge computing node to execute the task, and executes the task according to the edge acceleration distribution process.

9. An edge computing-driven terminal data security acceleration distribution system, including: IoT terminals are used to establish communication links with current edge computing nodes and send task requests; The current edge computing node is used to receive the task request and execute the edge acceleration distribution process according to the task request. The current edge computing node includes a feature acquisition module, an identifier determination module, and a channel determination module. The feature acquisition module is used to acquire node feature vectors and task feature vectors corresponding to the task request; The identifier determination module is used to determine the resilience enhancement strategy identifier of a task based on the node feature vector and the task feature vector. The resilience enhancement strategy identifier includes computational resource sufficiency and network resource sufficiency. Determining the resilience enhancement strategy identifier based on the node feature vector and the task feature vector includes: determining the computational resource requirement threshold and network latency requirement threshold corresponding to the task based on a preset task resource requirement database and the task feature vector; determining the node feature vector corresponding to the current time; determining the node load rate, round-trip latency, and latency jitter value of the current edge computing node at the next time based on the node feature vector corresponding to the current time; determining the predicted network latency value of the current edge computing node at the next time based on the round-trip latency and the latency jitter value; performing difference processing on the computational resource capacity and the computational resource requirement threshold corresponding to the task, and the node load rate of the edge computing node at the next time, based on a preset computational resource capacity, to obtain computational resource sufficiency; and performing difference processing on the network latency requirement threshold corresponding to the task and the predicted network latency value of the current edge computing node at the next time, to obtain network resource sufficiency. The channel determination module is used to determine an accelerated distribution channel based on the resilience enhancement strategy identifier. The accelerated distribution channel is used to execute the task through different resilience enhancement strategies, including: When the accelerated distribution channel is the first accelerated channel, the main target acceleration algorithm is determined according to the preset edge acceleration algorithm library and the task feature vector, and the task is executed based on the main target acceleration algorithm; When the accelerated distribution channel is the second accelerated channel, a degradation acceleration algorithm is determined based on the task feature vector, and the task is executed based on the degradation acceleration algorithm; When the accelerated distribution channel is a collaborative acceleration channel, a collaborative node corresponding to the current edge computing node is determined, and the collaborative node is used to execute the task according to the edge accelerated distribution process. The current edge computing node or the collaborating node obtains the execution data corresponding to the task and sends the execution data to the IoT terminal.

Images