Setting device, setting method, and setting program

The setting device optimizes communication anomaly detection by generating network graphs, extracting clusters, and assigning them to models based on calculation costs, addressing the challenge of balancing load and characteristics in existing technologies, resulting in improved detection performance and resource optimization.

WO2024241594A1PCT designated stage expired Publication Date: 2024-11-28NT T INC
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
PCT/JP2023/019581
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-25
Publication Date
2024-11-28

AI Technical Summary

Technical Problem

Existing communication anomaly detection technologies using unsupervised deep learning face challenges in appropriately setting the monitoring range for each model, balancing calculation load and communication characteristics, leading to inefficient resource utilization and detection performance.

Method used

A setting device that generates a graph of network devices and connections, extracts clusters, calculates the calculation cost of each cluster, and assigns clusters to learning models to equalize calculation costs and improve learning accuracy, ensuring appropriate monitoring ranges for anomaly detection models.

Benefits of technology

This approach allows for efficient and balanced resource allocation, enhancing anomaly detection performance and optimizing calculation resources by appropriately setting monitoring ranges for each model, thereby improving detection accuracy and resource utilization.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure JP2023019581_28112024_PF_FP_ABST
    Figure JP2023019581_28112024_PF_FP_ABST
Patent Text Reader

Abstract

A setting device (100) according to the present embodiment has a generation unit (121), an extraction unit (122), a calculation unit (123), and an allocation unit (124). The generation unit (121) generates a graph relating to devices constituting a network and connections between the devices. The extraction unit (122) extracts clusters from the graph on the basis of the graph generated by the generation unit (121). The calculation unit (123) calculates a calculation cost of each cluster extracted by the extraction unit (122). The allocation unit (124) allocates each cluster to one of a plurality of trained models for detecting a communication abnormality of the network on the basis of the calculation costs calculated by the calculation unit (123).
Need to check novelty before this filing date? Find Prior Art

Description

Setting device, setting method, and setting program

[0001] The present invention relates to a setting device, a setting method, and a setting program.

[0002] In recent years, the total number and methods of cyber attacks have been increasing, and it is necessary to respond to new attacks that are constantly being created. For this reason, unsupervised deep learning-based communication anomaly detection technology has attracted attention. This technology automatically learns the characteristics of normal communication from a network, detects data that deviates from these characteristics as an anomaly, and outputs an alert (see, for example, Non-Patent Document 1).

[0003] While unsupervised deep learning-based communication anomaly detection technology has the advantage of being able to detect anomalies with high accuracy, it incurs high computational costs. Therefore, when monitoring large-scale networks, the monitoring range is divided and anomaly detection is performed using multiple deep learning models. The division of the monitoring range takes into consideration factors such as equalizing the computational load for each model, detecting appropriate networks to simplify user management, and allocating similar communication ranges to the same model to improve learning accuracy.

[0004] "Feature extraction of packet payload using BERT", [online], [searched May 12, 2023], Internet: https: / / www.jstage.jst.go.jp / article / pjsai / JSAI2021 / 0 / JSAI2021_1F2GS10a04 / _article / -char / ja /

[0005] However, the above-mentioned conventional technology has a problem in that it may not be possible to appropriately set the monitoring range assigned to each model. For example, in the conventional technology, in order to efficiently utilize the performance of each model, it may not be possible to set the monitoring range of each model taking into account the balance of the calculation amount of each model and the communication characteristics to be learned.

[0006] In order to solve the above-mentioned problems and achieve the objectives, the setting device of the present invention is characterized by having devices that constitute a network, a generation unit that generates a graph related to the connections between the devices, an extraction unit that extracts clusters from the graph based on the graph generated by the generation unit, a calculation unit that calculates the calculation cost of each cluster extracted by the extraction unit, and an assignment unit that assigns each cluster to one of a plurality of learning models that detect communication anomalies in the network based on the calculation cost calculated by the calculation unit.

[0007] According to the present invention, it is possible to appropriately set the monitoring range to be assigned to each model.

[0008] Fig. 1 is a diagram illustrating an overview of a communication anomaly detection system including a setting device according to an embodiment. Fig. 2 is a block diagram illustrating an example of the configuration of the setting device according to an embodiment. Fig. 3 is a diagram illustrating an example of data stored in the setting device according to an embodiment. Fig. 4 is a diagram illustrating a specific example of setting processing according to an embodiment. Fig. 5 is a flowchart illustrating an example of the flow of setting processing according to an embodiment. Fig. 6 is a diagram illustrating an example of a computer that executes a setting program.

[0009] Hereinafter, embodiments of a setting device, a setting method, and a setting program according to the present application will be described in detail with reference to the accompanying drawings. Note that the setting device, the setting method, and the setting program according to the present application are not limited to these embodiments.

[0010] 1. System Configuration Example First, an overview of the entire intrusion detection system 10 including a setting device 100 according to this embodiment will be described. Fig. 1 is a diagram showing an overview of a communication anomaly detection system including a setting device according to this embodiment.

[0011] In the example of Figure 1, intrusion detection system 10 is a communication anomaly detection system having setting device 100 and anomaly detection models 200-1 to 200-3, and is realized by a cloud system or the like. For example, intrusion detection system 10 receives communication data from a target network and outputs a communication anomaly detection determination result, allowing the user to understand whether or not a communication anomaly exists. Note that the number of anomaly detection models 200 included in intrusion detection system 10 is not limited to the example of Figure 1, and can be increased or decreased at the request of the user, etc.

[0012] The setting device 100 is an information processing device that divides communication data input from a target network into multiple clusters and sets an appropriate monitoring range for each anomaly detection model 200 based on the computational cost of each cluster and the similarity between clusters, and is realized by a PC (Personal Computer) or the like.

[0013] The setting device 100 generates a graph relating to the devices constituting a network and the connections between the devices, and extracts clusters from the graph based on the generated graph. The setting device 100 then calculates the computational cost of each extracted cluster, and assigns each cluster to one of a plurality of learning models for detecting network communication anomalies based on the calculated computational cost.

[0014] The setting device 100 first generates a graph relating to the devices that make up the network and the connections between the devices. For example, the setting device 100 generates graph data in which the devices in the network to be monitored are represented as nodes and connections or communications between the devices are represented as edges.

[0015] Next, the setting device 100 extracts clusters from the graph based on the generated graph. For example, the setting device 100 performs graph clustering on the generated graph data by maximizing modularity, and extracts multiple clusters in which nodes are closely connected to each other within the clusters.

[0016] Then, the setting device 100 calculates the calculation cost of each extracted cluster. For example, the setting device 100 calculates, for each extracted cluster, the calculation cost, which is the processing feature of the devices in the cluster.

[0017] The setting device 100 then assigns each cluster to one of a plurality of learning models for detecting network communication anomalies based on the calculated computational cost. For example, the setting device 100 assigns clusters to each anomaly detection model 200 based on the computational cost calculated for each cluster so that the computational costs of the clusters assigned to each anomaly detection model 200 are equal.

[0018] As a result, the setting device 100 assigns a monitoring range to each learning model so that the computational costs are equal, and the monitoring range of the anomaly detection model 200 is appropriately set, thereby improving detection performance and optimizing computational resources.

[0019] The anomaly detection model 200 is, for example, a learning model using unsupervised deep learning that detects communication anomalies from communication data of a network to be monitored. The anomaly detection model 200 can efficiently detect anomalies by performing anomaly detection processing on data in the monitoring range set by the setting device 100.

[0020] 2. Configuration of the setting device 100 Next, the configuration of the setting device 100 shown in Fig. 1 will be described with reference to Fig. 2. Fig. 2 is a block diagram showing an example configuration of a detection system according to an embodiment. The setting device 100 has a communication unit 110, a control unit 120, and a storage unit 130, and the setting device 100 and anomaly detection models 200-1 to 200-n are connected to each other via wired or wireless communication so as to be able to communicate with each other.

[0021] The communication unit 110 is realized by, for example, a network interface card (NIC). The communication unit 110 is connected to the anomaly detection models 200 via a wired or wireless connection, and transmits and receives information to and from the anomaly detection models 200. The communication unit 110, for example, acquires communication data from the target network and mediates the provision of data divided among the anomaly detection models 200, enabling the transmission and reception of information between the setting device 100 and external networks and devices.

[0022] The storage unit 130 is realized by a storage device such as a RAM (Random Access Memory) or a hard disk, for example. The storage unit 130 stores data and programs necessary for various processes performed by the control unit 120. The storage unit 130 includes a graph data storage unit 131, a cluster information storage unit 132, and a calculation information storage unit 133, which are closely related to the present invention.

[0023] The graph data storage unit 131 stores, for example, graph data generated by the generation unit 121 (described later) in which each device constituting the target network is represented as a node and connections between devices are represented as edges.

[0024] The cluster information storage unit 132 stores, for example, each cluster extracted by the extracting unit 122 (described later) in association with information on the nodes (devices) included in the cluster.

[0025] The calculation information storage unit 133 stores the calculation cost φ and the similarity S between clusters calculated by the calculation unit 123, which will be described later. A specific example of data stored in the calculation information storage unit 133 will now be described with reference to Fig. 3. Fig. 3 is a diagram showing an example of data stored in the setting device according to the embodiment.

[0026] The calculation information storage unit 133 stores the calculation cost φ calculated for each extracted cluster in order, in association with the cluster, as shown in Fig. 3A. Also, as shown in Fig. 3B, the calculation information storage unit 133 stores the similarity S calculated between two clusters, such as "cluster 1-cluster 2" and "cluster 1-cluster 3," in order.

[0027] The control unit 120 is realized by a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or the like executing various programs stored in a storage device within the device using RAM as a work area. The control unit 120 is also realized by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array). The control unit 120 has a generation unit 121, an extraction unit 122, a calculation unit 123, and an allocation unit 124.

[0028] The generator 121 generates a graph relating to devices constituting a network and connections between the devices. The generator 121 may also generate graph data in which devices are represented as nodes and connections between devices are represented as edges.

[0029] For example, the generation unit 121 generates graph data for a network to be monitored, in which each device constituting the network is represented as a node and the communication connections between the devices are represented as edges, and stores the generated graph data in the graph data storage unit 131. Note that communication between devices may be wired or wireless. The generation unit 121 may generate the graph data based on pre-stored network configuration information, or may generate the graph data based on information contained in communication data acquired from the network to be monitored.

[0030] The extraction unit 122 extracts clusters from the graph based on the graph generated by the generation unit 121. Furthermore, the extraction unit 122 may extract clusters from the graph so that modularity, which exhibits a high value when nodes in a set of extracted clusters are connected to each other and are loosely connected to nodes in other clusters, is maximized.

[0031] For example, the extraction unit 122 extracts clusters from the graph data stored in the graph data storage unit 131 so as to maximize the modularity Q shown in the following equation (1), and stores data about the extracted clusters in the cluster information storage unit 132.

[0032]

[0033] Here, in the above formula (1), "C" represents a set of extracted clusters, and "e ii " indicates the ratio of the number of edges connecting nodes belonging to cluster i to the number of edges in the entire graph. ij " indicates the ratio of the number of edges connecting nodes belonging to cluster i to nodes in cluster j to the total number of edges in the entire graph. In other words, the modularity Q in the above formula (1) is a function that exhibits a high value when the set C of clusters extracted from the graph is densely connected to nodes within the cluster and sparsely connected to nodes in other clusters.

[0034] The extraction unit 122 performs graph clustering so as to maximize the modularity Q of the above formula (1), thereby being able to extract from the graph data a cluster structure in which nodes within the cluster are closely connected to one another. Note that the cluster extraction process by the extraction unit 122 is not limited to a method using the modularity Q of the above formula (1), and various graph clustering methods (e.g., Louvain method), algorithms, etc. that have been proposed up to now may be used as appropriate.

[0035] The calculation unit 123 calculates the calculation cost of each cluster extracted by the extraction unit 122. Furthermore, the calculation unit 123 may calculate the calculation cost using a processing feature per predetermined time output by a device included in the cluster.

[0036] For example, the calculation unit 123 refers to the information of each cluster stored in the cluster information storage unit 132, calculates the calculation cost φ using the processing features per unit time for anomaly detection output by the devices in each cluster, and stores the calculation cost φ in the calculation information storage unit 133.

[0037] Here, the processing feature is a value calculated from communication data acquired from the network to be monitored, such as the number of packets or the number of sessions, and differs depending on the content of the anomaly detection technology being applied.

[0038] Furthermore, the calculation unit 123 may further calculate the similarity between the clusters extracted by the extraction unit 122. For example, the calculation unit 123 refers to the information of each cluster stored in the cluster information storage unit 132, calculates the similarity S between the clusters shown in the following formula (2), and stores the calculated similarity S in the calculation information storage unit 133.

[0039]

[0040] Here, in the above formula (2), "S i,j " indicates the similarity between cluster i and cluster j. "P(i)" indicates the set of protocols flowing in cluster i, and "r p "(i)" indicates the proportion of protocol p flowing in cluster i. In other words, the above formula (2) indicates that the more identical the protocols flowing in cluster i and cluster j, the higher the value becomes, and also indicates that this is an index that becomes higher the more protocols that match with each other, showing a larger proportion.

[0041] The calculation unit 123 can calculate the degree of protocol agreement between clusters as the degree of similarity between clusters using the above formula (2). Note that the calculation of the degree of similarity between clusters is not limited to the calculation of the degree of protocol agreement using the above formula (2), and is set appropriately for each anomaly detection technology.

[0042] The allocation unit 124 allocates each cluster to one of a plurality of learning models for detecting network communication anomalies, based on the calculation cost calculated by the calculation unit 123. For example, the allocation unit 124 refers to the calculation cost φ for each cluster stored in the calculation information storage unit 133, and allocates clusters such that the calculation costs φ of the anomaly detection models 200 for detecting communication anomalies are equal.

[0043] Furthermore, the allocation unit 124 may assign each cluster to one of a plurality of learning models for detecting network communication anomalies, based on the computational cost and the similarity between the clusters calculated by the calculation unit 123. For example, the allocation unit 124 refers to the computational cost φ and the similarity S between the clusters for each cluster stored in the calculation information storage unit 133, and assigns clusters to each anomaly detection model 200 so that the following formula (3) is minimized.

[0044]

[0045] Here, in the above formula (3), "M" indicates a set of anomaly detection models 200, and "Cm" indicates a set of clusters assigned to model m. Furthermore, "w1" and "w2" indicate hyperparameters that adjust the weights of each term, and are appropriately set by the user.

[0046] In the above formula (3), the first half term multiplied by w1 calculates the sum of the differences in the calculation costs assigned to each model, and the second half term multiplied by w2 calculates the sum of the similarities between the clusters assigned to each model. Note that because the second half term is multiplied by a minus, the smaller the second half term, the higher the similarity of the clusters assigned to the model.

[0047] In other words, the above equation (3) shows the allocation of clusters so that the computational costs allocated to each model are equal and the similarity between the clusters assigned to each model is high, by minimizing the sum of the differences in computational costs allocated to each model and the sum of the negative similarities between clusters.

[0048] By performing allocation so as to minimize the above formula (3), the allocation unit 124 can allocate clusters to each anomaly detection model 200 so that the calculation costs are uniform and the similarity between clusters allocated to the same model is high. Note that the above formula (3) is just an example, and the allocation process is not limited to the above formula (3).

[0049] 3. Specific Example of Setting Process Now, a specific example of setting process by the setting device 100 will be described using the example of Fig. 4. Fig. 4 is a diagram showing a specific example of setting process according to the embodiment. In the example of Fig. 4, four clusters are extracted from the generated graph data, and the calculation costs and similarities between the clusters are calculated, after which each cluster is assigned to two anomaly detection models 200-1 and 200-2.

[0050] First, the generation unit 121 generates graph data from the configuration information of the network to be monitored, with devices as nodes and connections between devices as edges. Then, the extraction unit 122 performs graph clustering on the generated graph data so as to maximize the modularity Q shown in the above formula (1), and extracts clusters 1 to 4 in which nodes within the clusters are considered to be closely connected to each other.

[0051] Next, the calculation unit 123 calculates the calculation cost φ for each of the extracted clusters 1 to 4. At the same time, the calculation unit 123 also calculates the inter-cluster similarity S shown in the above formula (2). Specifically, between any two clusters among clusters 1 to 4, six inter-cluster similarities S are calculated in the order "1-2", "1-3", and so on.

[0052] Then, the allocation unit 124 uses the calculated computational costs and the similarities between the clusters to assign clusters 1 and 3 to the anomaly detection model 200-1, and clusters 2 and 4 to the anomaly detection model 200-2, based on the results of equation (3) above.

[0053] 4. Example of Setting Process Next, the setting process by the setting device 100 will be described with reference to Fig. 5. Fig. 5 is a flowchart showing an example of the flow of the setting process according to the embodiment. First, the setting device 100 acquires information regarding the connections between devices in the network (step S101).

[0054] If information regarding the connections between devices in the network has been acquired (step S101; Yes), the generation unit 121 generates graph data indicating the configuration of the network (step S102). On the other hand, if information regarding the connections between devices in the network has not been acquired (step S101; No), the setting device 100 waits until information regarding the connections between devices in the network has been acquired.

[0055] The extraction unit 122 then extracts clusters in which nodes within the clusters are closely connected from the generated graph data (step S103). The calculation unit 123 then calculates the computational cost of each cluster and the similarity between the clusters (step S104). The assignment unit 124 then assigns the clusters to the respective learning models based on the calculated computational cost and the similarity between the clusters (step S105), and the setting device 100 ends the process.

[0056] 5. Effects of the Embodiment As described above, the setting device 100 according to the present embodiment includes the generation unit 121, the extraction unit 122, the calculation unit 123, and the allocation unit 124. The generation unit 121 generates a graph relating to devices constituting a network and the connections between the devices. The extraction unit 122 extracts clusters from the graph based on the graph generated by the generation unit 121.

[0057] The calculation unit 123 calculates the calculation cost of each cluster extracted by the extraction unit 122. The assignment unit 124 assigns each cluster to one of a plurality of learning models for detecting communication anomalies in a network, based on the calculation cost calculated by the calculation unit 123.

[0058] As a result, the setting device 100 divides the monitoring range of a large-scale network into multiple clusters and assigns a cluster to each anomaly detection model so that the calculation cost of each cluster is uniform, thereby achieving the effect of being able to appropriately set the monitoring range to be assigned to each model.

[0059] Furthermore, by setting an appropriate monitoring range for each anomaly detection model 200 using the setting device 100, it is expected that the anomaly detection performance of the anomaly detection model 200 will improve and calculation resources will be optimized.

[0060] In addition, the calculation unit 123 of the setting device 100 further calculates the similarity between the clusters extracted by the extraction unit 122, and the assignment unit 124 assigns each cluster to one of multiple learning models that detect communication anomalies in the network based on the calculation cost calculated by the calculation unit 123 and the similarity between the clusters.

[0061] As a result, the setting device 100 not only makes the computational cost uniform for each cluster, but also assigns clusters so that the similarity between clusters assigned to the same anomaly detection model 200 is high, thereby achieving the effect of being able to more appropriately set the monitoring range for each model.

[0062] Furthermore, the generator 121 of the setting device 100 generates graph data in which devices are nodes and connections between devices are edges. This allows the setting device 100 to appropriately generate graph data to which graph clustering can be applied for devices and connections between devices that exist in the network to be monitored.

[0063] Furthermore, the extraction unit 122 of the setting device 100 extracts clusters from the graph so that modularity, which shows a high value when nodes in an extracted cluster are connected to each other and are loosely connected to nodes in other clusters, is maximized. This allows the setting device 100 to appropriately extract a cluster structure in which nodes in a cluster are tightly connected to each other from the graph data.

[0064] Furthermore, the calculation unit 123 of the setting device 100 calculates the calculation cost using the processing features per predetermined time output by the devices included in the cluster. As a result, the setting device 100 calculates the number of processing features of the devices in each cluster as the calculation cost, which has the effect of enabling allocation processing to be performed so that the number of processing features of the clusters allocated to each model is equal.

[0065] [6. System Configuration, etc.] Of the processes described in the above embodiments, some of the processes described as being performed automatically can also be performed manually. Alternatively, all or some of the processes described as being performed manually can be performed automatically using known methods. In addition, the information including the processing procedures, specific names, various data, and parameters shown in the above documents and drawings can be changed as desired unless otherwise specified. For example, the various information shown in each drawing is not limited to the information shown in the drawings.

[0066] Furthermore, the components of each device shown in the figure are conceptual functional units and do not necessarily have to be physically configured as shown. In other words, the specific form of distribution and integration of each device is not limited to that shown in the figure, and all or part of the devices can be functionally or physically distributed and integrated in any unit depending on various loads, usage conditions, etc. Furthermore, all or any part of the processing functions performed by each device can be realized by a CPU and a program analyzed and executed by the CPU, or can be realized as hardware using wired logic.

[0067] 2 may be held in a storage server or the like, rather than being held by the setting device 100. In this case, the setting device 100 acquires various pieces of information by accessing the storage server.

[0068] 6 is a diagram showing an example of a hardware configuration. The setting device 100 according to the embodiment described above is realized by a computer 1000 having a configuration as shown in FIG.

[0069] 6 is a diagram showing an example of a computer that executes a setting program. The computer 1000 includes, for example, a memory 1010 and a CPU 1020. The computer 1000 also includes a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. These components are connected by a bus 1080.

[0070] The memory 1010 includes a ROM (Read Only Memory) 1011 and a RAM 1012. The ROM 1011 stores a boot program such as a BIOS (Basic Input Output System). The hard disk drive interface 1030 is connected to a hard disk drive 1090. The disk drive interface 1040 is connected to a disk drive 1141. A removable storage medium such as a magnetic disk or optical disk is inserted into the disk drive 1141. The serial port interface 1050 is connected to a mouse 1110 and a keyboard 1120, for example. The video adapter 1060 is connected to a display 1130, for example.

[0071] The hard disk drive 1090 stores, for example, an OS (Operating System) 1091, an application program 1092, a program module 1093, and program data 1094. That is, the program that defines each process of the setting device 100 is implemented as a program module 1093 in which code that can be executed by the computer 1000 is written. The program module 1093 is stored, for example, in the hard disk drive 1090. For example, a program module 1093 for executing the same process as the functional configuration of the setting device 100 is stored in the hard disk drive 1090. Note that the hard disk drive 1090 may be replaced with an SSD (Solid State Drive).

[0072] Furthermore, setting data used in the processing of the above-described embodiment is stored as program data 1094, for example, in memory 1010 or hard disk drive 1090. Then, CPU 1020 reads out program module 1093 or program data 1094 stored in memory 1010 or hard disk drive 1090 into RAM 1012 as necessary and executes them.

[0073] The program module 1093 and program data 1094 may not necessarily be stored in the hard disk drive 1090, but may also be stored in a removable storage medium and read by the CPU 1020 via the disk drive 1141 or the like. Alternatively, the program module 1093 and program data 1094 may be stored in another computer connected via a network (LAN, WAN, etc.). The program module 1093 and program data 1094 may then be read by the CPU 1020 from the other computer via the network interface 1070.

[0074] 10 Intrusion detection system 100 Setting device 110 Communication unit 120 Control unit 121 Generation unit 122 Extraction unit 123 Calculation unit 124 Allocation unit 130 Storage unit 131 Graph data storage unit 132 Cluster information storage unit 133 Calculation information storage unit 200, 200-1 to 200-n Anomaly detection model

Claims

1. A setting device comprising: a generation unit that generates a graph relating to devices that constitute a network and connections between the devices; an extraction unit that extracts clusters from the graph based on the graph generated by the generation unit; a calculation unit that calculates the calculation cost of each cluster extracted by the extraction unit; and an assignment unit that assigns each cluster to one of a plurality of learning models that detect communication anomalies in the network based on the calculation cost calculated by the calculation unit.

2. The setting device described in claim 1, characterized in that the calculation unit further calculates the similarity between the clusters extracted by the extraction unit, and the assignment unit assigns each cluster to one of a plurality of learning models for detecting network communication anomalies based on the calculation cost calculated by the calculation unit and the similarity between the clusters.

3. The setting device according to claim 1, characterized in that the generation unit generates graph data in which the devices are nodes and connections between the devices are edges.

4. The setting device described in claim 1, characterized in that the extraction unit extracts the clusters from the graph so as to maximize modularity, which is a high value when the nodes in the extracted clusters are connected to each other and sparsely connected to the nodes of other clusters.

5. The setting device according to claim 1, characterized in that the calculation unit calculates the calculation cost using processing features per specified time output by the devices included in the cluster.

6. A setting method executed by a setting device, comprising: a generation step of generating a graph relating to devices constituting a network and connections between the devices; an extraction step of extracting clusters from the graph based on the graph generated by the generation step; a calculation step of calculating the calculation cost of each cluster extracted by the extraction step; and an assignment step of assigning each cluster to one of a plurality of learning models that detect communication anomalies in the network based on the calculation cost calculated by the calculation step.

7. A configuration program for causing a computer to execute the following steps: a generation step for generating a graph relating to devices that constitute a network and connections between the devices; an extraction step for extracting clusters from the graph based on the graph generated by the generation step; a calculation step for calculating the calculation cost of each cluster extracted by the extraction step; and an assignment step for assigning each cluster to one of a plurality of learning models that detect communication anomalies in the network based on the calculation cost calculated by the calculation step.

Citation Information

Patent Citations

  • Generation device, generation method, and generation program

    JP2020005234A

  • Incident effect range estimation device, incident effect range estimation method, storage medium, and system

    WO2018079439A1