An edge node selection method and device, electronic equipment and storage medium

Abstract

Embodiments of the present application provide an edge node selection method and device, electronic equipment and storage medium. The method comprises: in response to user information representing user network demand, vectorizing the user information to obtain an initial vector; based on the position representing network behavior and the position representing application object in the initial vector, adjusting the initial vector to obtain a fusion vector; extracting features from the fusion vector to determine the selection probability of each security service; and based on the selection probability of each security service and the configuration information of each security service in the candidate edge node, selecting a target edge node matching the user information from the candidate edge node. The embodiments of the present application strengthen the important features of the initial vector, i.e. the network behavior and the application object, expand the semantic expression, improve the semantic differentiation of similar expressions, improve the analysis ability of user network demand, and efficiently, accurately and flexibly select the target edge node.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to an edge node selection method, apparatus, electronic device, and storage medium. Background Technology

[0002] Edge nodes, as network nodes in a cloud computing data center, have fewer intermediate links between themselves and the final user terminals, resulting in better responsiveness and connection speeds for user access. For example, as security atomic capabilities are deployed to the edge cloud, much of the business traffic with security requirements will be securely processed and forwarded to its destination via edge nodes.

[0003] Because different edge nodes have different functions and can provide different security services, users' network needs also vary. Related technologies require security capability surveys to determine the functionality of edge nodes, and professionals then determine the appropriate edge nodes for connection based on user network requirements and the capabilities of the edge nodes.

[0004] However, the process of selecting edge nodes is complex, inefficient, and difficult to accurately select suitable edge nodes. Summary of the Invention

[0005] This application provides an edge node selection method, apparatus, electronic device, and storage medium for efficiently and accurately selecting suitable edge nodes.

[0006] In a first aspect, embodiments of this application provide a first edge node selection method, the method comprising:

[0007] In response to user information representing user network needs, the user information is vectorized to obtain an initial vector;

[0008] Based on the positions representing network behavior and application objects in the initial vector, the initial vector is adjusted to obtain the fusion vector;

[0009] Feature extraction is performed on the fusion vector to determine the selection probability of each security service;

[0010] Based on the selection probability of each security service and the configuration information of each security service in the candidate edge nodes, a target edge node matching the user information is selected from the candidate edge nodes.

[0011] The above scheme, after vectorizing user information to obtain an initial vector, does not directly extract features from the initial vector. Instead, it enhances the features of the initial vector based on the positions representing network behavior and application objects, expanding the semantic expression. Then, based on the fused vector, it extracts features to obtain the selection probability corresponding to each security service. Combined with the configuration information of each security service in the candidate edge nodes, it can select the target edge node that meets the user's network needs from the candidate edge nodes without manual intervention, achieving efficient, accurate, and flexible selection of target edge nodes. In addition, by enhancing the important features in the initial vector, the expression of user network needs is strengthened, the semantic distinguishability of similar expressions is improved, the ability to parse user network needs is enhanced, and the selection accuracy is further improved.

[0012] In some optional implementations, the initial vector is adjusted based on the positions representing network behavior and the positions representing application objects in the initial vector to obtain a fusion vector, including:

[0013] The initial vector and the positions representing network behavior are input into the first sub-network. The weights of the network behavior features in the initial vector are increased through the first sub-network to obtain the first adjustment vector; and

[0014] The initial vector and the position representing the application object are input into the second sub-network. The weights of the application object features in the initial vector are increased through the second sub-network to obtain the second adjustment vector.

[0015] The first adjustment vector and the second adjustment vector are merged to obtain a merged vector.

[0016] The above scheme improves the semantic expression weight of network behavior in the initial vector by weighting the network behavior features based on the location representing network behavior in the first sub-network. It also improves the semantic expression weight of application objects in the initial vector by weighting the application object features based on the location representing application objects in the second sub-network. Since the first and second sub-networks have the same structure but different weights, the first adjustment vector output by the first sub-network and the second adjustment vector output by the second sub-network are different. By further fusing the first and second adjustment vectors, a fusion vector is obtained that simultaneously improves the semantic expression weight of both network behavior and application object.

[0017] In some optional implementations, the first adjustment vector and the second adjustment vector are fused to obtain a fused vector, including:

[0018] Based on the first weighting coefficient, the first adjustment vector and the second adjustment vector are weighted and summed to obtain the fusion vector.

[0019] The above scheme, based on the first weighting coefficient, fuses the first adjustment vector and the second adjustment vector to obtain a fusion vector that simultaneously improves the semantic expression weight of network behavior and the semantic expression weight of application object.

[0020] In some optional implementations, feature extraction is performed on the fusion vector to determine the selection probability corresponding to each security service, including:

[0021] The fused vector is input into a feature extraction model, and features are extracted from the fused vector through the feature extraction model to obtain target features characterizing the fused vector.

[0022] The target features are combined to obtain the selection probability of each security service.

[0023] The above scheme, through a feature extraction model, can efficiently and accurately determine the selection probability of each security service for each fusion vector.

[0024] In some optional implementations, based on the selection probability corresponding to each security service and the configuration information of each security service in the candidate edge nodes, a target edge node matching the user information is selected from the candidate edge nodes, including:

[0025] For any candidate edge node, the product of the configuration information of any security service in the candidate edge node and the selection probability corresponding to the security service is determined as the selection value corresponding to the security service; wherein, if the candidate edge node is configured with the security service, the configuration information of the security service is 1, otherwise the configuration information of the security service is 0.

[0026] The sum of the selection values ​​corresponding to all security services is used to determine the selection value corresponding to the candidate edge node;

[0027] The target edge node is selected based on the selection value corresponding to each candidate edge node.

[0028] In the above scheme, for the configuration of a candidate edge node, it may be configured with a certain security service, in which case the candidate edge node supports that security service, and the configuration information of the security service is 1 (equivalent to the upper limit coefficient for adjusting the selection probability); or it may not be configured with that security service, in which case the candidate edge node does not support that security service, and the configuration information of the security service is 0 (equivalent to the lower limit coefficient for adjusting the selection probability, even if the demand for a certain security service is high, the selection value of this security service is 0). Then, based on the product of the configuration information and the selection probability, the selection value corresponding to the security service can be determined. This selection value represents the probability of selecting the candidate edge node and using the security service. By combining the selection values ​​corresponding to each security service, the selection value corresponding to the candidate edge node is obtained. This selection value represents the probability of selecting the candidate edge node, thereby accurately selecting the target edge node.

[0029] In some optional implementations, the target edge node is selected based on the selection value corresponding to each candidate edge node, including:

[0030] For any candidate edge node, the selection value, network quality parameters, and available load parameters corresponding to the candidate edge node are weighted and summed based on the second weighting coefficient to determine the matching degree corresponding to the candidate edge node.

[0031] The candidate edge node with the highest matching degree is selected as the target edge node.

[0032] The above scheme, based on the second weighting coefficient, performs a weighted summation of the selection value, network quality parameters, and available load parameters corresponding to the candidate edge nodes to obtain the matching degree of the candidate edge nodes. This matching degree reflects the probability of selecting the candidate edge node based on the security service itself, while also taking into account the influence of network quality and load. Based on the matching degree, the scheme accurately selects the target edge nodes that are suitable for the security service, have a good network, and have a high load, which is more in line with the needs of actual applications.

[0033] In some optional implementations, before vectorizing the user information to obtain an initial vector, the following steps are also included:

[0034] A first identifier is added to the characters representing network behavior in the user information, and a second identifier is added to the characters representing application objects in the user information;

[0035] After vectorizing the user information to obtain an initial vector, the process further includes:

[0036] The first identifier is used to determine the position in the initial vector representing network behavior, and the second identifier is used to determine the position in the initial vector representing application objects.

[0037] The above scheme requires knowing the positions representing network behavior and application objects when adjusting the initial vector. By performing character recognition on user information, a first identifier is added to the characters representing network behavior in the user information, and a second identifier, different from the first identifier, is added to the characters representing application objects. After the user information is vectorized, even if the characters become features in the initial vector, the position of the network behavior in the initial vector can be determined based on the first identifier corresponding to the characters representing network behavior; and the position of the application object in the initial vector can be determined based on the second identifier corresponding to the characters representing application objects.

[0038] Secondly, embodiments of this application provide a first edge node selection device, the device comprising:

[0039] The vectorization module is used to vectorize user information representing user network needs to obtain an initial vector.

[0040] The vector processing module is used to adjust the initial vector based on the positions representing network behavior and the positions representing application objects in the initial vector to obtain a fused vector;

[0041] The probability determination module is used to extract features from the fusion vector and determine the selection probability of each security service.

[0042] The node selection module is used to select a target edge node that matches the user information from the candidate edge nodes based on the selection probability of each security service and the configuration information of each security service in the candidate edge nodes.

[0043] In some optional implementations, the vector processing module is specifically used for:

[0044] The initial vector and the positions representing network behavior are input into the first sub-network. The weights of the network behavior features in the initial vector are increased through the first sub-network to obtain the first adjustment vector; and

[0045] The initial vector and the position representing the application object are input into the second sub-network. The weights of the application object features in the initial vector are increased through the second sub-network to obtain the second adjustment vector.

[0046] The first adjustment vector and the second adjustment vector are merged to obtain a merged vector.

[0047] In some optional implementations, the vector processing module is specifically used for:

[0048] Based on the first weighting coefficient, the first adjustment vector and the second adjustment vector are weighted and summed to obtain the fusion vector.

[0049] In some optional implementations, the probability determination module is specifically used for:

[0050] The fused vector is input into a feature extraction model, and features are extracted from the fused vector through the feature extraction model to obtain target features characterizing the fused vector.

[0051] The target features are combined to obtain the selection probability of each security service.

[0052] In some optional implementations, the node selection module is specifically used for:

[0053] For any candidate edge node, the product of the configuration information of any security service in the candidate edge node and the selection probability corresponding to the security service is determined as the selection value corresponding to the security service; wherein, if the candidate edge node is configured with the security service, the configuration information of the security service is 1, otherwise the configuration information of the security service is 0.

[0054] The sum of the selection values ​​corresponding to all security services is used to determine the selection value corresponding to the candidate edge node;

[0055] The target edge node is selected based on the selection value corresponding to each candidate edge node.

[0056] In some optional implementations, the node selection module is specifically used for:

[0057] For any candidate edge node, the selection value, network quality parameters, and available load parameters corresponding to the candidate edge node are weighted and summed based on the second weighting coefficient to determine the matching degree corresponding to the candidate edge node.

[0058] The candidate edge node with the highest matching degree is selected as the target edge node.

[0059] In some alternative implementations, the vectorization module is also used for:

[0060] Before vectorizing the user information to obtain the initial vector, a first identifier is added to the characters representing network behavior in the user information, and a second identifier is added to the characters representing application objects in the user information.

[0061] The vectorization module is also used for:

[0062] After vectorizing the user information to obtain an initial vector, the position representing network behavior in the initial vector is determined based on the first identifier, and the position representing application object in the initial vector is determined based on the second identifier.

[0063] Thirdly, embodiments of this application provide an electronic device including at least one processor and at least one memory, wherein the memory stores a computer program that, when executed by the processor, causes the processor to perform any of the edge node selection methods described in the first aspect above.

[0064] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program executable by a processor, which, when run on the processor, causes the processor to perform any of the edge node selection methods described in the first aspect above. Attached Figure Description

[0065] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0066] Figure 1 A flowchart illustrating the first edge node selection method provided in this application embodiment;

[0067] Figure 2 A flowchart illustrating the second edge node selection method provided in this application embodiment;

[0068] Figure 3 A flowchart illustrating the third edge node selection method provided in this application embodiment;

[0069] Figure 4 A schematic diagram illustrating the adjustment process of the initial vector provided in an embodiment of this application;

[0070] Figure 5 A flowchart illustrating the fourth edge node selection method provided in this application embodiment;

[0071] Figure 6 This is a schematic diagram of the feature extraction process provided in an embodiment of this application;

[0072] Figure 7 A flowchart illustrating the fifth edge node selection method provided in this application embodiment;

[0073] Figure 8A flowchart illustrating the sixth edge node selection method provided in this application embodiment;

[0074] Figure 9 This is a schematic diagram of the structure of the edge node selection device provided in the embodiments of this application;

[0075] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0076] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0077] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0078] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the term "connection" should be interpreted broadly. For example, it can refer to a direct connection, an indirect connection through an intermediate medium, or a connection within two devices. Those skilled in the art can understand the specific meaning of the above term in this application based on the specific circumstances.

[0079] Edge nodes, as network nodes in a cloud computing data center, have fewer intermediate links between themselves and the final user terminals, resulting in better responsiveness and connection speeds for user access. For example, as security atomic capabilities are deployed to the edge cloud, much of the business traffic with security requirements will be securely processed and forwarded to its destination via edge nodes.

[0080] Because different edge nodes have different functions and can provide different security services, users' network needs also vary. Related technologies require security capability surveys to determine the functionality of edge nodes, and professionals then determine the appropriate edge nodes for connection based on user network requirements and the capabilities of the edge nodes.

[0081] However, the process of selecting edge nodes is complex, inefficient, and makes it difficult to accurately select suitable edge nodes. In particular, it cannot respond promptly to changes in edge node functionality or user network requirements.

[0082] With the advancement of intent-driven networks, user intent can be intelligently perceived through the network, and plans can be automatically formulated and tasks can be automatically executed.

[0083] However, the interpretation of user intent by intent-driven networks is affected by factors such as small sample size and high semantic representation similarity, resulting in problems such as low recognition accuracy and inaccuracy.

[0084] In view of this, embodiments of this application propose an edge node selection method, apparatus, electronic device, and storage medium. The method includes: responding to user information representing user network needs, vectorizing the user information to obtain an initial vector; adjusting the initial vector based on the positions representing network behavior and application objects in the initial vector to obtain a fused vector; extracting features from the fused vector to determine the selection probability corresponding to each security service; and selecting a target edge node matching the user information from the candidate edge nodes based on the selection probability corresponding to each security service and the configuration information of each security service in the candidate edge nodes.

[0085] The above scheme, after vectorizing user information to obtain an initial vector, does not directly extract features from the initial vector. Instead, it enhances the features of the initial vector based on the positions representing network behavior and application objects, expanding the semantic expression. Then, based on the fused vector, it extracts features to obtain the selection probability corresponding to each security service. Combined with the configuration information of each security service in the candidate edge nodes, it can select the target edge node that meets the user's network needs from the candidate edge nodes without manual intervention. This achieves efficient, accurate, and flexible selection of target edge nodes and can respond promptly to changes in edge node functions or user network needs. In addition, by enhancing the important features in the initial vector, the expression of user network needs is strengthened, the semantic distinguishability of similar expressions is improved, the ability to parse user network needs is enhanced, and the selection accuracy is further improved.

[0086] The technical solution of this application and how it solves the above-mentioned technical problems will be described in detail below with reference to the accompanying drawings and specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0087] Figure 1 A flowchart illustrating the first edge node selection method provided in this application embodiment is shown below. Figure 1 As shown, it includes the following steps:

[0088] Step S101: In response to user information representing user network needs, the user information is vectorized to obtain an initial vector.

[0089] Since manually configuring edge nodes is tedious and prone to errors, this embodiment uses intent parsing for intelligent point selection; therefore, user information needs to be vectorized to obtain an initial vector.

[0090] This embodiment does not specifically limit the triggering method of user information. For example, when a new user requests to connect to an edge node for the first time, or when the user's needs change, user information representing the user's current network needs can be triggered to dynamically adapt to changes in edge nodes and user needs.

[0091] For example, user information can be in the form of text, images, or voice. If the information is in text format, it can be directly vectorized; if it is in other formats, it can be converted into text before vectorization. This embodiment does not impose specific limitations on this.

[0092] This embodiment does not limit the specific implementation method of vectorizing user information, such as the word-based distributed expression method (Word2vec), text vectorization (doc2vec), etc.

[0093] Step S102: Based on the positions representing network behavior and application objects in the initial vector, adjust the initial vector to obtain the fusion vector.

[0094] In practice, when configuring edge nodes, the number of samples is usually limited and the similarity of user information is high. Directly extracting features from the initial vector may result in edge nodes that are poorly adapted to the user's network requirements.

[0095] Based on this, this embodiment does not directly extract features from the initial vector. Instead, it enhances the features of the two important features in the initial vector, namely the positions representing network behavior and the positions representing application objects, to expand the semantic expression. By enhancing the important features in the initial vector, the expression of user network requirements is strengthened in the fused vector.

[0096] The aforementioned network behaviors correspond to network services, and in security services, these are protective measures such as blocking and detection.

[0097] The aforementioned application objects are the objects to which the network service is applied, that is, the business systems that require the network service.

[0098] Step S103: Extract features from the fusion vector to determine the selection probability of each security service.

[0099] In this embodiment, the expression of user network needs is enhanced in the fusion vector, and the semantic distinguishability of similar expressions is improved. Based on the fusion vector, features that accurately represent user network needs can be extracted, thereby determining the probability of selecting each security service on the demand side.

[0100] Step S104: Based on the selection probability of each security service and the configuration information of each security service in the candidate edge nodes, select the target edge node that matches the user information from the candidate edge nodes.

[0101] In this embodiment, the selection probability is based on the user's demand for each security service. It also needs to consider the configuration information of each security service in the candidate edge nodes, that is, the ability of the candidate edge nodes to provide each security service. Only by combining these two pieces of information can the target edge node that meets the user's network needs be selected.

[0102] The above scheme, after vectorizing user information to obtain an initial vector, does not directly extract features from the initial vector. Instead, it enhances the features of the initial vector based on the positions representing network behavior and application objects, expanding the semantic expression. Then, based on the fused vector, it extracts features to obtain the selection probability corresponding to each security service. Combined with the configuration information of each security service in the candidate edge nodes, it can select the target edge node that meets the user's network needs from the candidate edge nodes without manual intervention, achieving efficient, accurate, and flexible selection of target edge nodes. In addition, by enhancing the important features in the initial vector, the expression of user network needs is strengthened, the semantic distinguishability of similar expressions is improved, the ability to parse user network needs is enhanced, and the selection accuracy is further improved.

[0103] Figure 2 A flowchart illustrating the second edge node selection method provided in this application embodiment is shown below. Figure 2 As shown, it includes the following steps:

[0104] Step S201: In response to user information representing user network needs, add a first identifier to the characters representing network behavior in the user information, and add a second identifier to the characters representing application objects in the user information.

[0105] Since adjusting the initial vector requires knowing the location representing the network behavior and the location representing the application object, it is difficult to identify the location directly from the initial vector. However, if there are some special markers in the initial vector, the location can be easily identified.

[0106] For example, character recognition is performed on user information to determine characters representing network behavior (first characters) and characters representing application objects (second characters); a first identifier is added to the first character in the user information, and a second identifier, distinct from the first identifier, is added to the second character in the user information. Subsequently, positions are identified from the initial vector based on these identifiers.

[0107] Step S202: The user information is vectorized to obtain an initial vector.

[0108] The specific implementation of step S202 can be found in the above embodiments, and will not be repeated here.

[0109] Step S203: Determine the position representing network behavior in the initial vector based on the first identifier, and determine the position representing application object in the initial vector based on the second identifier.

[0110] In practice, after user information is vectorized, the characters become features in the initial vector, but the first and second identifiers mentioned above will not change.

[0111] Based on this, this embodiment determines the position of the network behavior in the initial vector based on the first identifier corresponding to the character representing the network behavior; and determines the position of the application object in the initial vector based on the second identifier corresponding to the character representing the application object.

[0112] Step S204: Based on the positions representing network behavior and application objects in the initial vector, adjust the initial vector to obtain the fusion vector.

[0113] Step S205: Extract features from the fusion vector to determine the selection probability of each security service.

[0114] Step S206: Based on the selection probability of each security service and the configuration information of each security service in the candidate edge nodes, select the target edge node that matches the user information from the candidate edge nodes.

[0115] The specific implementation of steps S204 to S206 can be found in other embodiments, and will not be repeated here.

[0116] The above scheme requires knowing the positions representing network behavior and application objects when adjusting the initial vector. By performing character recognition on user information, a first identifier is added to the characters representing network behavior in the user information, and a second identifier, different from the first identifier, is added to the characters representing application objects. After the user information is vectorized, even if the characters become features in the initial vector, the position of the network behavior in the initial vector can be determined based on the first identifier corresponding to the characters representing network behavior; and the position of the application object in the initial vector can be determined based on the second identifier corresponding to the characters representing application objects.

[0117] Figure 3 A flowchart illustrating the third edge node selection method provided in this application embodiment is shown below. Figure 3 As shown, it includes the following steps:

[0118] Step S301: In response to user information representing user network needs, the user information is vectorized to obtain an initial vector.

[0119] The specific implementation of step S301 can be found in the above embodiments, and will not be repeated here.

[0120] Step S302: Input the initial vector and the position representing the network behavior into the first sub-network, and increase the weight of the network behavior feature in the initial vector through the first sub-network to obtain the first adjustment vector; and input the initial vector and the position representing the application object into the second sub-network, and increase the weight of the application object feature in the initial vector through the second sub-network to obtain the second adjustment vector.

[0121] In this embodiment, a pseudo twin network (PSN) is set up. The PSN includes a first subnetwork and a second subnetwork. The first subnetwork and the second subnetwork have the same structure but different weights. The first subnetwork corresponds to network behavior, and the second subnetwork corresponds to application objects.

[0122] See Figure 4 As shown, the first sub-network performs feature weighting on the network behavior features in the initial vector A0 based on the position representing the network behavior, thereby increasing the weight of the semantic expression of network behavior that determines the user's network needs, and correspondingly reducing the weight of other features, thus obtaining the first adjustment vector A1;

[0123] The second sub-network, based on the location of the application object, performs feature weighting on the application object features in the initial vector A0, increasing the semantic expression weight of the application object that determines the user's network requirements, and correspondingly reducing the weight of other features, thus obtaining the second adjustment vector A2.

[0124] The aforementioned PSN enables users with similar expressions to have greater differentiation.

[0125] Step S303: Merge the first adjustment vector and the second adjustment vector to obtain a merged vector.

[0126] Since the first subnetwork and the second subnetwork have the same structure but different weights, the first adjustment vector A1 output by the first subnetwork and the second adjustment vector A2 output by the second subnetwork are different.

[0127] As mentioned above Figure 4 For example, A1 and A2 need to be further fused to obtain a fusion vector A3 that simultaneously improves the semantic expression weight of network behavior and the semantic expression weight of application object.

[0128] Step S304: Extract features from the fusion vector to determine the selection probability of each security service.

[0129] Step S305: Based on the selection probability of each security service and the configuration information of each security service in the candidate edge nodes, select the target edge node that matches the user information from the candidate edge nodes.

[0130] The specific implementation of steps S304 to S305 can be found in other embodiments, and will not be repeated here.

[0131] The above scheme improves the semantic expression weight of network behavior in the initial vector by weighting the network behavior features based on the location representing network behavior in the first sub-network. It also improves the semantic expression weight of application objects in the initial vector by weighting the application object features based on the location representing application objects in the second sub-network. Since the first and second sub-networks have the same structure but different weights, the first adjustment vector output by the first sub-network and the second adjustment vector output by the second sub-network are different. By further fusing the first and second adjustment vectors, a fusion vector is obtained that simultaneously improves the semantic expression weight of both network behavior and application object.

[0132] In some optional implementations, step S303 above can be implemented in, but is not limited to, the following ways:

[0133] Based on the first weighting coefficient, the first adjustment vector and the second adjustment vector are weighted and summed to obtain the fusion vector.

[0134] The aforementioned first weighting coefficient includes the weighting coefficient corresponding to the first adjustment vector and the weighting coefficient corresponding to the second adjustment vector. These two coefficients can be set according to actual needs, and this embodiment does not limit them.

[0135] For example, the fusion vector A3 = α*A1 + β*A2; where α is the weighting coefficient corresponding to the first adjustment vector, A1 is the first adjustment vector, β is the weighting coefficient corresponding to the second adjustment vector, and A1 is the second adjustment vector.

[0136] The above scheme, based on the first weighting coefficient, fuses the first adjustment vector and the second adjustment vector to obtain a fusion vector that simultaneously improves the semantic expression weight of network behavior and the semantic expression weight of application object.

[0137] Figure 5 A flowchart illustrating the fourth edge node selection method provided in this application embodiment is shown below. Figure 5 As shown, it includes the following steps:

[0138] Step S501: In response to user information representing user network needs, the user information is vectorized to obtain an initial vector.

[0139] Step S502: Based on the positions representing network behavior and application objects in the initial vector, adjust the initial vector to obtain the fusion vector.

[0140] The specific implementation of steps S501 to S502 can be referred to the above embodiments, and will not be repeated here.

[0141] Step S503: Input the fused vector into the feature extraction model, and extract features from the fused vector through the feature extraction model to obtain the target features characterizing the fused vector.

[0142] See Figure 6 As shown, after inputting the fused vector into the feature extraction model, semantic information is extracted through the convolutional layer of the feature extraction model to obtain the target features.

[0143] Step S504: Perform feature synthesis on the target features to obtain the selection probability corresponding to each security service.

[0144] As mentioned above Figure 6 For example, features are synthesized through the fully connected layer of the feature extraction model to obtain the selection probability of each security service.

[0145] Step S505: Based on the selection probability of each security service and the configuration information of each security service in the candidate edge nodes, select the target edge node that matches the user information from the candidate edge nodes.

[0146] The specific implementation of step S505 can be found in other embodiments, and will not be repeated here.

[0147] The above scheme, through a feature extraction model, can efficiently and accurately determine the selection probability of each security service for each fusion vector.

[0148] The following is a specific example to illustrate this:

[0149] The initial model includes an initial behavior network, an initial object network, and an initial neural network;

[0150] The initial behavioral network is fed into a sample vector and the positions in that sample vector that represent the network behavior.

[0151] The input to the initial object network is the sample vector mentioned above, and the position of the application object in the sample vector;

[0152] The output of the initial behavior network is the first sample vector, and the output of the initial object network is the second sample vector;

[0153] The initial neural network input is the fused sample vector obtained by fusing the first sample vector and the second sample vector; the output is the predicted selection probability of each security service.

[0154] The similarity between the predicted selection probabilities of each security service output by the initial neural network and their corresponding calibration probabilities is used as an optimization condition to train the initial model.

[0155] After achieving the optimization requirements, the target model is obtained; the target model includes the first sub-network obtained from the initial behavior network training, the second sub-network obtained from the initial object network training, and the feature extraction model obtained from the initial neural network training.

[0156] After receiving user information, it is vectorized to obtain an initial vector;

[0157] The initial vector and the position representing the network behavior are input into the first sub-network, and the first sub-network outputs the first adjustment vector.

[0158] The initial vector and the position representing the application object are input into the second sub-network, and the second sub-network outputs the second adjustment vector.

[0159] The first adjustment vector and the second adjustment vector are merged to obtain the merged vector;

[0160] The fused vector is input into the feature extraction model, which outputs the selection probability for each security service.

[0161] Figure 7 A flowchart illustrating the fifth edge node selection method provided in this application embodiment is shown below. Figure 7 As shown, it includes the following steps:

[0162] Step S701: In response to user information representing user network needs, the user information is vectorized to obtain an initial vector.

[0163] Step S702: Based on the positions representing network behavior and application objects in the initial vector, adjust the initial vector to obtain the fusion vector.

[0164] Step S703: Extract features from the fusion vector to determine the selection probability of each security service.

[0165] The specific implementation of steps S701 to S703 can be found in the above embodiments, and will not be repeated here.

[0166] Step S704: For any candidate edge node, the product of the configuration information of any security service in the candidate edge node and the selection probability corresponding to the security service is determined as the selection value corresponding to the security service.

[0167] If the candidate edge node is configured with the security service, the configuration information of the security service is 1; otherwise, the configuration information of the security service is 0.

[0168] For example, the selection probability is the degree of demand for each security service from the user's perspective; the configuration information of each security service in the candidate edge nodes is the ability of the candidate edge nodes to provide each security service from the perspective of the candidate edge nodes.

[0169] For the configuration of a candidate edge node, it may be configured with a certain security service. In that case, the candidate edge node supports the security service, and the configuration information of the security service is 1 (which is equivalent to the upper limit coefficient of the selection probability adjustment).

[0170] For the alternative edge node configuration, it may not have the security service configured. In that case, the alternative edge node does not support the security service, and the security service configuration information is 0 (which is equivalent to the lower limit coefficient of the selection probability adjustment. Even if the demand for a certain security service is high, the selection value of this security service is 0).

[0171] Based on the product of configuration information and selection probability, the selection value corresponding to the security service can be determined (representing the probability of selecting the candidate edge node and using the security service).

[0172] For example, for candidate edge node j, the selection value E corresponding to security service i. ij =P i *C ij ;

[0173] Among them, P i The selection probability corresponding to security service i; Cij This refers to the configuration information for security service i for candidate edge node j; if candidate edge node j is configured with security service i, C ij The value is 1; if the candidate edge node j is not configured with security service i, the configuration information of the security service is 0; 1≤i≤M, where M is the total number of security services; 1≤j≤N, where N is the total number of candidate edge nodes.

[0174] Step S705: The sum of the selection values ​​corresponding to all security services is used to determine the selection value corresponding to the candidate edge node.

[0175] In implementation, it is not possible to refer to only a specific security service. The selection values ​​corresponding to each security service must be considered in order to obtain the selection values ​​corresponding to the candidate edge nodes (representing the probability of selecting the candidate edge node).

[0176] For example, for candidate edge node j, its corresponding selection value Among them, E ij Let i be the selection value corresponding to security service i for candidate edge node j; 1≤i≤M, where M is the total number of security services; 1≤j≤N, where N is the total number of candidate edge nodes.

[0177] Step S706: Select the target edge node based on the selection value corresponding to each candidate edge node.

[0178] As mentioned above, the selection value corresponding to the candidate edge node represents the probability of selecting that candidate edge node. Therefore, based on the selection value corresponding to the candidate edge node, the target edge node can be selected more accurately.

[0179] In the above scheme, for the configuration of a candidate edge node, it may be configured with a certain security service, in which case the candidate edge node supports that security service, and the configuration information of the security service is 1 (equivalent to the upper limit coefficient for adjusting the selection probability); or it may not be configured with that security service, in which case the candidate edge node does not support that security service, and the configuration information of the security service is 0 (equivalent to the lower limit coefficient for adjusting the selection probability, even if the demand for a certain security service is high, the selection value of this security service is 0). Then, based on the product of the configuration information and the selection probability, the selection value corresponding to the security service can be determined. This selection value represents the probability of selecting the candidate edge node and using the security service. By combining the selection values ​​corresponding to each security service, the selection value corresponding to the candidate edge node is obtained. This selection value represents the probability of selecting the candidate edge node, thereby accurately selecting the target edge node.

[0180] Figure 8 A flowchart illustrating the sixth edge node selection method provided in this application embodiment is shown below. Figure 8 As shown, it includes the following steps:

[0181] Step S801: In response to user information representing user network needs, the user information is vectorized to obtain an initial vector.

[0182] Step S802: Based on the positions representing network behavior and application objects in the initial vector, adjust the initial vector to obtain the fusion vector.

[0183] Step S803: Extract features from the fusion vector to determine the selection probability of each security service.

[0184] Step S804: For any candidate edge node, the product of the configuration information of any security service in the candidate edge node and the selection probability corresponding to the security service is determined as the selection value corresponding to the security service.

[0185] Step S805: The sum of the selection values ​​corresponding to all security services is determined as the selection value corresponding to the candidate edge node.

[0186] The specific implementation of steps S801 to S805 can be found in other embodiments, and will not be repeated here.

[0187] Step S806: For any candidate edge node, based on the second weighting coefficient, perform a weighted summation of the selection value, network quality parameters, and available load parameters corresponding to the candidate edge node to determine the matching degree corresponding to the candidate edge node.

[0188] During implementation, the network environment (such as network latency, packet loss, jitter, etc.) between user terminals and edge nodes may vary significantly. If an edge node with a poor network is selected, it may affect data transmission. The load conditions of edge nodes (connected users, remaining bandwidth, available resources, etc.) may also vary significantly. If an edge node with insufficient available load is selected, it may affect data processing.

[0189] In other words, besides the impact on the security service itself, other factors may also influence the selection of target edge nodes. For example, a candidate edge node may have a high selection value but poor network quality or low available load. Ultimately, a candidate edge node with a lower selection value but better network quality and higher available load may be selected.

[0190] Based on this, this embodiment uses a second weighting coefficient to perform a weighted summation of the selection value, network quality parameters, and available load parameters corresponding to the candidate edge node, to obtain the matching degree corresponding to the candidate edge node. This matching degree reflects the probability of selecting the candidate edge node based on the security service itself, while also taking into account the influence of network quality and load; and then selects the candidate edge node with the highest matching degree as the target edge node.

[0191] This embodiment does not limit the specific implementation method for determining network quality parameters; as long as the network environment is better, the corresponding network quality parameters should be larger. For example, the weighted sum of network latency, packet loss rate, and jitter can be determined, and the reciprocal of this weighted sum can be used as the network quality parameter.

[0192] This embodiment does not limit the specific implementation method for determining the available load parameter. As long as the fewer the connected loads and the higher the load capacity, the larger the corresponding available load parameter will be. For example, the weighted sum of the reciprocal of the number of connected users, the remaining bandwidth, and the available resources can be determined as the available load parameter.

[0193] Step S807: Select the candidate edge node with the highest matching degree as the target edge node.

[0194] For example, for candidate edge node j, its corresponding matching degree S j =θ1*Ej+θ2*Fj+θ3*Hj; where Ej is the selection value corresponding to candidate edge node j; Fj is the network quality parameter corresponding to candidate edge node j; Hj is the available load parameter corresponding to candidate edge node j; 1≤j≤N, and N is the total number of candidate edge nodes.

[0195] From matching degree S1 to S N Select the node with the highest matching degree and use its corresponding candidate edge node as the target edge node.

[0196] The above scheme, based on the second weighting coefficient, performs a weighted summation of the selection value, network quality parameters, and available load parameters corresponding to the candidate edge nodes to obtain the matching degree of the candidate edge nodes. This matching degree reflects the probability of selecting the candidate edge node based on the security service itself, while also taking into account the influence of network quality and load. Based on the matching degree, the scheme accurately selects the target edge nodes that are suitable for the security service, have a good network, and have a high load, which is more in line with the needs of actual applications.

[0197] It is understandable that the above Figure 8 This is just one possible way to select the target edge node. In practice, other methods can also be used, such as directly selecting the candidate edge node with the highest selection value as the target edge node.

[0198] like Figure 9 As shown, this application embodiment provides an edge node selection device 900, which includes:

[0199] The vectorization module 901 is used to vectorize the user information representing the user's network needs in response to the user information to obtain an initial vector.

[0200] The vector processing module 902 is used to adjust the initial vector based on the positions representing network behavior and the positions representing application objects in the initial vector to obtain a fused vector;

[0201] The probability determination module 903 is used to extract features from the fusion vector and determine the selection probability corresponding to each security service;

[0202] The node selection module 904 is used to select a target edge node that matches the user information from the candidate edge nodes based on the selection probability corresponding to each security service and the configuration information of each security service in the candidate edge nodes.

[0203] In some optional implementations, the vector processing module 902 is specifically used for:

[0204] The initial vector and the positions representing network behavior are input into the first sub-network. The weights of the network behavior features in the initial vector are increased through the first sub-network to obtain the first adjustment vector; and

[0205] The initial vector and the position representing the application object are input into the second sub-network. The weights of the application object features in the initial vector are increased through the second sub-network to obtain the second adjustment vector.

[0206] The first adjustment vector and the second adjustment vector are merged to obtain a merged vector.

[0207] In some optional implementations, the vector processing module 902 is specifically used for:

[0208] Based on the first weighting coefficient, the first adjustment vector and the second adjustment vector are weighted and summed to obtain the fusion vector.

[0209] In some optional implementations, the probability determination module 903 is specifically used for:

[0210] The fused vector is input into a feature extraction model, and features are extracted from the fused vector through the feature extraction model to obtain target features characterizing the fused vector.

[0211] The target features are combined to obtain the selection probability of each security service.

[0212] In some optional implementations, the node selection module 904 is specifically used for:

[0213] For any candidate edge node, the product of the configuration information of any security service in the candidate edge node and the selection probability corresponding to the security service is determined as the selection value corresponding to the security service; wherein, if the candidate edge node is configured with the security service, the configuration information of the security service is 1, otherwise the configuration information of the security service is 0.

[0214] The sum of the selection values ​​corresponding to all security services is used to determine the selection value corresponding to the candidate edge node;

[0215] The target edge node is selected based on the selection value corresponding to each candidate edge node.

[0216] In some optional implementations, the node selection module 904 is specifically used for:

[0217] For any candidate edge node, the selection value, network quality parameters, and available load parameters corresponding to the candidate edge node are weighted and summed based on the second weighting coefficient to determine the matching degree corresponding to the candidate edge node.

[0218] The candidate edge node with the highest matching degree is selected as the target edge node.

[0219] In some optional implementations, the vectorization module 901 is further configured to:

[0220] Before vectorizing the user information to obtain the initial vector, a first identifier is added to the characters representing network behavior in the user information, and a second identifier is added to the characters representing application objects in the user information.

[0221] Vectorization module 901 is also used for:

[0222] After vectorizing the user information to obtain an initial vector, the position representing network behavior in the initial vector is determined based on the first identifier, and the position representing application object in the initial vector is determined based on the second identifier.

[0223] Since this device is the same as the device in the method of this application embodiment, and the principle of the device in solving the problem is similar to that of the method, the implementation of the device can be referred to the implementation of the method, and the repeated parts will not be described again.

[0224] Based on the same technical concept, this application also provides an electronic device 1000, such as... Figure 10As shown, it includes at least one processor 1001 and a memory 1002 connected to at least one processor. In this embodiment, the specific connection medium between the processor 1001 and the memory 1002 is not limited. Figure 10 Taking the connection between processor 1001 and memory 1002 via bus 1003 as an example. The bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, Figure 10 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0225] The processor 1001 is the control center of the electronic device, capable of connecting various parts of the electronic device via various interfaces and lines. It performs data processing by running or executing instructions stored in the memory 1002 and accessing data stored in the memory 1002. Optionally, the processor 1001 may include one or more processing units. The processor 1001 may integrate an application processor and a modem processor. The application processor primarily handles the operating system, user interface, and applications, while the modem processor primarily handles issuing instructions. It is understood that the modem processor may not be integrated into the processor 1001. In some embodiments, the processor 1001 and the memory 1002 may be implemented on the same chip; in other embodiments, they may be implemented on separate chips.

[0226] Processor 1001 can be a general-purpose processor, such as a CPU, digital signal processor, application-specific integrated circuit (ASIC), field-programmable gate array or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the edge node selection method can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0227] Memory 1002, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. Memory 1002 may include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic memory, magnetic disk, optical disk, etc. Memory 1002 can be any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer, but is not limited thereto. In the embodiments of this application, memory 1002 can also be a circuit or any other device capable of implementing storage functions for storing program instructions and / or data.

[0228] In this embodiment, the memory 1002 stores a computer program, which, when executed by the processor 1001, causes the processor 1001 to perform the following:

[0229] In response to user information representing user network needs, the user information is vectorized to obtain an initial vector;

[0230] Based on the positions representing network behavior and application objects in the initial vector, the initial vector is adjusted to obtain the fusion vector;

[0231] Feature extraction is performed on the fusion vector to determine the selection probability of each security service;

[0232] Based on the selection probability of each security service and the configuration information of each security service in the candidate edge nodes, a target edge node matching the user information is selected from the candidate edge nodes.

[0233] In some optional implementations, processor 1001 specifically performs:

[0234] The initial vector and the positions representing network behavior are input into the first sub-network. The weights of the network behavior features in the initial vector are increased through the first sub-network to obtain the first adjustment vector; and

[0235] The initial vector and the position representing the application object are input into the second sub-network. The weights of the application object features in the initial vector are increased through the second sub-network to obtain the second adjustment vector.

[0236] The first adjustment vector and the second adjustment vector are merged to obtain a merged vector.

[0237] In some optional implementations, processor 1001 specifically performs:

[0238] Based on the first weighting coefficient, the first adjustment vector and the second adjustment vector are weighted and summed to obtain the fusion vector.

[0239] In some optional implementations, processor 1001 specifically performs:

[0240] The fused vector is input into a feature extraction model, and features are extracted from the fused vector through the feature extraction model to obtain target features characterizing the fused vector.

[0241] The target features are combined to obtain the selection probability of each security service.

[0242] In some optional implementations, processor 1001 specifically performs:

[0243] For any candidate edge node, the product of the configuration information of any security service in the candidate edge node and the selection probability corresponding to the security service is determined as the selection value corresponding to the security service; wherein, if the candidate edge node is configured with the security service, the configuration information of the security service is 1, otherwise the configuration information of the security service is 0.

[0244] The sum of the selection values ​​corresponding to all security services is used to determine the selection value corresponding to the candidate edge node;

[0245] The target edge node is selected based on the selection value corresponding to each candidate edge node.

[0246] In some optional implementations, processor 1001 specifically performs:

[0247] For any candidate edge node, the selection value, network quality parameters, and available load parameters corresponding to the candidate edge node are weighted and summed based on the second weighting coefficient to determine the matching degree corresponding to the candidate edge node.

[0248] The candidate edge node with the highest matching degree is selected as the target edge node.

[0249] In some optional implementations, before vectorizing the user information to obtain an initial vector, the processor 1001 further performs:

[0250] A first identifier is added to the characters representing network behavior in the user information, and a second identifier is added to the characters representing application objects in the user information;

[0251] After vectorizing the user information to obtain an initial vector, the processor 1001 further executes:

[0252] The first identifier is used to determine the position in the initial vector representing network behavior, and the second identifier is used to determine the position in the initial vector representing application objects.

[0253] Since the electronic device is the same as the electronic device in the method of this application embodiment, and the principle of the electronic device in solving the problem is similar to that of the method, the implementation of the electronic device can refer to the implementation of the method, and the repeated parts will not be described again.

[0254] Based on the same technical concept, embodiments of this application also provide a computer-readable storage medium storing a computer program executable by a processor, which, when run on the processor, causes the processor to perform the steps of the edge node selection method described above.

[0255] In some alternative implementations, various aspects of the edge node selection method provided in this application can also be implemented as a program product containing computer-executable instructions that, when the program product is run on a computer device, cause the computer device to perform the steps of the edge node selection method according to the various exemplary embodiments of this application described above.

[0256] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0257] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0258] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0259] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0260] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0261] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. An edge node selection method, characterized in that, The method includes: In response to user information representing a user's network needs, a first identifier is added to the characters representing network behavior in the user information, and a second identifier is added to the characters representing the application object in the user information. The user information is vectorized to obtain an initial vector; the position representing network behavior in the initial vector is determined based on the first identifier, and the position representing application object in the initial vector is determined based on the second identifier; The initial vector and the position representing the network behavior are input into a first sub-network. The weights of the network behavior features in the initial vector are increased and the weights of other features in the initial vector are decreased through the first sub-network to obtain a first adjustment vector. The initial vector and the position representing the application object are input into a second sub-network. The weights of the application object features in the initial vector are increased and the weights of other features in the initial vector are decreased through the second sub-network to obtain a second adjustment vector. Here, the network behavior is the behavior corresponding to the network service, and the application object is the object applied to by the network service. The first and second sub-networks have the same structure but different weights. The first adjustment vector and the second adjustment vector are merged to obtain a fused vector; Feature extraction is performed on the fusion vector to determine the selection probability of each security service; For any candidate edge node, the product of the configuration information of any security service in the candidate edge node and the selection probability corresponding to the security service is determined as the selection value corresponding to the security service; wherein, if the candidate edge node is configured with the security service, the configuration information of the security service is 1, otherwise the configuration information of the security service is 0. The sum of the selection values ​​corresponding to all security services is determined as the selection value corresponding to the candidate edge node, and the target edge node is selected based on the selection value corresponding to each candidate edge node.

2. The method as described in claim 1, characterized in that, The first adjustment vector and the second adjustment vector are merged to obtain a merged vector, including: Based on the first weighting coefficient, the first adjustment vector and the second adjustment vector are weighted and summed to obtain the fusion vector.

3. The method as described in claim 1, characterized in that, Feature extraction is performed on the fusion vector to determine the selection probability of each security service, including: The fused vector is input into a feature extraction model, and features are extracted from the fused vector through the feature extraction model to obtain target features characterizing the fused vector. The target features are combined to obtain the selection probability of each security service.

4. The method as described in claim 1, characterized in that, Based on the selection values ​​corresponding to each candidate edge node, the target edge node is selected, including: For any candidate edge node, the selection value, network quality parameters, and available load parameters corresponding to the candidate edge node are weighted and summed based on the second weighting coefficient to determine the matching degree corresponding to the candidate edge node. The candidate edge node with the highest matching degree is selected as the target edge node.

5. An edge node selection device, characterized in that, The device includes: A vectorization module is used to respond to user information representing user network needs by adding a first identifier to characters representing network behavior in the user information and adding a second identifier to characters representing application objects in the user information; vectorizing the user information to obtain an initial vector; determining the position representing network behavior in the initial vector based on the first identifier and determining the position representing application objects in the initial vector based on the second identifier; A vector processing module is used to input the initial vector and the position representing the network behavior into a first sub-network, increase the weight of the network behavior feature in the initial vector through the first sub-network, and decrease the weight of other features in the initial vector besides the network behavior feature to obtain a first adjusted vector; and to input the initial vector and the position representing the application object into a second sub-network, increase the weight of the application object feature in the initial vector through the second sub-network, and decrease the weight of other features in the initial vector besides the application object feature to obtain a second adjusted vector; wherein the network behavior is the behavior corresponding to the network service, the application object is the object applied by the network service, the first sub-network and the second sub-network have the same structure but different weights; and to fuse the first adjusted vector and the second adjusted vector to obtain a fused vector; The probability determination module is used to extract features from the fusion vector and determine the selection probability of each security service. The node selection module is used to determine the selection value of any security service by multiplying the configuration information of any security service in the candidate edge node with the selection probability of the security service for any candidate edge node; wherein, if the candidate edge node is configured with the security service, the configuration information of the security service is 1, otherwise the configuration information of the security service is 0. The sum of the selection values ​​corresponding to all security services is determined as the selection value corresponding to the candidate edge node, and the target edge node is selected based on the selection value corresponding to each candidate edge node.

6. An electronic device, characterized in that, It includes at least one processor and at least one memory, wherein the memory stores a computer program that, when executed by the processor, causes the processor to perform the method as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, It stores a computer program executable by a computer, which, when run on the computer, causes the computer to perform the method as described in any one of claims 1 to 4.

Images