Analysis device, analysis method, and analysis program

The analysis server clusters alerts based on communication features to clearly define clusters and determine same-type alerts, reducing the workload of determining overdetection.

JP7582469B2Active Publication Date: 2024-11-13NIPPON TELEGRAPH & TELEPHONE CORP
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
JP2023526786
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Filing Date
2021-06-10
Publication Date
2024-11-13
Estimated Expiration
2041-06-10

Smart Images

  • Figure 0007582469000001
    Figure 0007582469000001
  • Figure 0007582469000002
    Figure 0007582469000002
  • Figure 0007582469000003
    Figure 0007582469000003
Patent Text Reader

Abstract

An analysis server (10) uses a model representing features of normal communication as a basis to accumulate alerts of communication determined not to be normal, in a storage unit (12). Then, the analysis server (10) performs clustering by targeting, among the accumulated alerts, alerts from which alerts having a category variable different from that of communication data used for learning of the model is excluded, and then using feature amounts of the communication included in the alerts. Subsequently, the analysis server (10) determines whether each of clusters generated by the clustering is constituted by the same type of alerts or not. Then, the analysis server (10) outputs the result of the clustering and the determination result whether or not respective clusters are constituted by the same type of alerts.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to an analysis device, an analysis method, and an analysis program for analyzing an alert indicating an abnormality in communication. [Background technology]

[0002] In conventional technology, when detecting communication anomalies based on a model that indicates the characteristics of normal communication, if new communication is added to the communication being detected, the added communication will be detected as an anomaly (overdetection) even if it is normal. As a result, there is a possibility that a large number of similar alerts will be generated. Here, the operator judges whether an alert is an overdetection or not for each terminal device and alert. Therefore, when a large number of alerts are generated, the operator's work becomes enormous.

[0003] In response to this, several technologies have been proposed to reduce the workload of operators.For example, there is a technology in which, when a detection system presents an alert, alerts that indicate the same category variables, such as the type of threat, terminal device ID, protocol, and port number, are presented together (see Non-Patent Document 1).In addition, there is a technology in which a detection system filters alerts using category variables, such as an IP address, port number, and protocol number, indicated by the alert, and the communication start date and time as keys (see Patent Documents 1 and 2). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2020-005184 A [Patent Document 2] JP 2020-135655 A [Non-patent literature]

[0005] [Non-Patent Document 1] NOZOMI Guardian, [Retrieved May 25, 2021], Internet <URL:https: / / www.exclusive-networks.com / fr / wp-content / uploads / sites / 17 / 2020 / 12 / FR-VR-Nozomi-Networks-Guardian.pdf> Summary of the Invention [Problem to be solved by the invention]

[0006] However, none of the above technologies can group alerts that are from communications with the same characteristics, such as the communication destination, but different data volumes, etc. Furthermore, even if the above technologies are used to group alerts based on data volumes, the characteristics of each group become unclear when the number of feature items to be grouped increases. Therefore, even if alerts are grouped using conventional technologies, it takes a huge amount of work to determine whether an alert is due to an overdetection.

[0007] Therefore, an object of the present invention is to solve the above-mentioned problems and reduce the work required to determine whether or not an alert is due to an overdetection. [Means for solving the problem]

[0008] In order to solve the above-mentioned problems, the present invention is characterized by comprising a storage unit that accumulates alerts of communication determined to be abnormal based on a model indicating characteristics of normal communication; a clustering unit that performs clustering of the alerts using communication characteristics contained in the accumulated alerts; a determination unit that determines, for each cluster generated by the clustering, whether or not the cluster is composed of alerts of the same type; and a result output unit that outputs the results of the clustering and a determination result as to whether or not each of the clusters is composed of alerts of the same type. Effect of the Invention

[0009] According to the present invention, it is possible to reduce the work required to determine whether or not an alert is due to an overdetection. [Brief description of the drawings]

[0010] [Figure 1] FIG. 1 is a diagram for explaining an example configuration and an overview of an analysis system. [Diagram 2] FIG. 2 is a diagram showing examples of items that the analysis server uses for clustering alerts. [Diagram 3] FIG. 3 is a diagram illustrating an example of clustering by the analysis server. [Figure 4] FIG. 4 is a flowchart illustrating an example of a processing procedure of the analysis server. [Diagram 5] FIG. 5 is a flowchart illustrating an example of a processing procedure of the analysis server. [Figure 6] FIG. 6 is a diagram showing an example of how an operator responds to an alert using the results of alert clustering by the analysis server. [Figure 7] FIG. 7 shows an example of the results of alert clustering by the analysis server and the judgment results for alerts of the same type. [Figure 8] FIG. 8 is a diagram illustrating an example of the configuration of a computer that executes the analysis program. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, a mode (embodiment) for carrying out the present invention will be described with reference to the drawings. The present invention is not limited to the embodiment described below.

[0012] [overview] First, with reference to FIG. 1, an outline of the operation of an analysis system 1 including an analysis server (analysis device) 10 will be described.

[0013] The analysis system 1 includes a sensing device 2 and an analysis server 10. The sensing device 2 observes communications of terminal devices (for example, IoT devices).

[0014] For example, the sensing device 2 observes the communication of the terminal device by mirroring the NW (network) device to which the terminal device is connected. Then, the sensing device 2 generates information (processed information) by processing the observation result of the communication of the terminal device, generates a feature of the communication from the processed information, and transmits it to the analysis server 10.

[0015] The characteristics of the above communication are information indicating, for example, the IP address of the terminal device performing the communication, the source IP address of the session, the destination IP address of the session, the destination port number of the session, the protocol number, the session duration, the total number of forward bytes, the total number of forward packets, the total number of reverse bytes, the total number of reverse packets, etc.

[0016] The analysis server 10 determines whether the communication is normal or not based on the characteristics of the communication transmitted from the sensing device 2 and a model indicating the characteristics of normal communication, and outputs an alert if it determines that the communication is abnormal.

[0017] For example, the analysis server 10 calculates an anomaly score that indicates the deviation between the feature of the communication transmitted from the sensing device 2 and a model that indicates the feature of normal communication, and outputs an alert if the anomaly score exceeds a predetermined threshold. The operator of the analysis system 1 then checks the content of the alert, and if the alert is an overdetection, provides feedback to that effect. The feedback here is, for example, an operation in which the operator registers in the analysis server 10 that the alert is an overdetection, thereby preventing future alerts from being generated for the same type of feature in the analysis server 10.

[0018] If a large number of alerts occur, the administrator's task of checking the above-mentioned items will become enormous. Therefore, in order to enable the administrator to efficiently perform the above-mentioned checking tasks (to perform checking tasks for similar types of alerts collectively), the analysis server 10 performs clustering of the alerts so that the characteristics of each cluster become clear.

[0019] For example, when clustering alerts, the analysis server 10 excludes from the clustering targets alerts of communications not included in the communication data used to train the above model (e.g., communications by new terminal devices). The analysis server 10 also clusters the alerts using feature items that tend to reveal communication characteristics (e.g., session duration, total number of bytes in the forward direction, total number of packets in the forward direction, total number of bytes in the reverse direction, total number of packets in the reverse direction, etc.). Furthermore, when clustering alerts, the analysis server 10 calculates the difference between the values ​​used to train the model for each item and for each terminal device indicated in the alert, and performs clustering using normalized values.

[0020] In this way, the analysis server 10 can perform clustering of alerts in such a way that the characteristics of each cluster become clearly apparent.

[0021] The analysis server 10 also outputs the result of the determination as to whether each cluster is composed of the same type of alerts. This allows the operator to identify clusters that are composed of the same type of alerts and therefore require only one overdetection determination. As a result, the work required to determine whether an alert is due to an overdetection can be reduced.

[0022] Furthermore, the analysis server 10 does not use payload information that depends on a specific payload when clustering alerts, and therefore can also handle clustering of alerts from encrypted communications.

[0023] [Configuration example] Next, a configuration example of the analysis server 10 will be described with reference to Fig. 1. The analysis server 10 includes a storage unit 12, a feature amount receiving unit 131, a learning unit 132, an analysis unit (accumulation unit) 133, a screen display processing unit 134, a clustering unit 135, a determination unit 136, and a result output unit 137.

[0024] The storage unit 12 stores information referenced by the analysis server 10 when it executes various processes, and the results of the execution of the various processes.

[0025] For example, the memory unit 12 stores communication features acquired from the sensing device 2, a model showing the characteristics of normal communication learned by the learning unit 132, communication tuples (initial tuples) of the features used to learn the model, analysis results of the communication by the analysis unit 133 (e.g., the above-mentioned anomaly score), a group of alerts of communication determined to be abnormal based on the above-mentioned model, etc.

[0026] In the following description, some of the categorical variables included in the feature amount are called communication tuples. Among these, the communication tuples of the feature amount used in learning the model are called initial tuples. The initial tuples are, for example, the IP address of the terminal device, the source IP address of the session, the destination IP address of the session, the destination port number of the session, the protocol number, etc., included in the feature amount used in learning the model, but are not limited to the above five.

[0027] The feature amount receiving unit 131 receives the feature amount of communication from the sensing device 2. Then, the feature amount receiving unit 131 accumulates the received feature amount of communication in the storage unit 12. For example, the feature amount receiving unit 131 first receives the feature amount of communication for model learning (feature amount of normal communication) and accumulates it in the storage unit 12. Thereafter, the feature amount receiving unit 131 receives the feature amount of communication to be detected and accumulates it in the storage unit 12.

[0028] The learning unit 132 learns a model that indicates the characteristics of normal communication, using the features of communication (features of normal communication) stored in the storage unit 12. After that, the learning unit 132 stores, in the storage unit 12, each parameter of the learned model and data used in learning the model (for example, initial tuples, etc.).

[0029] The analysis unit 133 analyzes the communication to be detected using the model learned by the learning unit 132. For example, the analysis unit 133 calculates an anomaly score indicating the deviation between the feature amount of the communication to be detected and the above model. Then, when the anomaly score exceeds a predetermined threshold (i.e., when there is a high possibility that the communication is not normal), the analysis unit 133 outputs an alert. The alert includes, for example, identification information of the communication that is the subject of the alert, the date and time of occurrence of the communication, etc. The output alert is accumulated in the storage unit 12.

[0030] The screen display processing unit 134 displays on the screen the alerts stored in the storage unit 12. For example, the screen display processing unit 134 displays on the screen a list of the alerts stored in the storage unit 12 based on an instruction input by the operator. This allows the operator to check on the screen what kind of alert has been output.

[0031] The clustering unit 135 performs clustering of the alerts stored in the storage unit 12 based on the feature amount of communication included in the alerts.

[0032] For example, the clustering unit 135 acquires an alert and a feature of the communication that is the subject of the alert from the storage unit 12. Then, the clustering unit 135 performs clustering of the alert based on the acquired feature of the communication. Here, the clustering unit 135 narrows down and normalizes the items of feature used for clustering among the feature of the communication that is the subject of the alert so that the feature of each cluster appears clearly.

[0033] For example, the clustering unit 135 excludes from the clustering targets alerts with communication tuples different from the communication tuples (initial tuples) of the features used in learning the model. For example, the clustering unit 135 compares the communication tuples (initial tuples) of the features used in learning the model with the communication tuples of the features of the communication included in the alert, and excludes from the clustering targets alerts with different categorical variables (combinations of the IP address of the terminal device itself, the source IP address of the session, the destination IP address of the session, the destination port number of the session, and the protocol number).

[0034] In this way, the clustering unit 135 can exclude alerts for which the reason for the alert is clear, such as communication from a new terminal device, from the targets of clustering.

[0035] Furthermore, the clustering unit 135 performs clustering by excluding information included in the alert such as the date and time of communication occurrence, the anomaly score, etc. For example, the clustering unit 135 performs clustering of the alert using at least one of the session duration, the total number of bytes in the forward direction, the total number of packets in the forward direction, the total number of bytes in the reverse direction, and the total number of packets in the reverse direction from among the features of the communication included in the alert.

[0036] Furthermore, the clustering unit 135 performs clustering of alerts for each terminal device using a normalized value of the difference between the value of the feature of the communication used to learn the model (the value at the time of initial learning) and the value of the feature of the communication included in the alert.

[0037] For example, the clustering unit 135 calculates the difference between the values ​​at the time of initial learning and the session duration, the total number of bytes in the forward direction, the total number of packets in the forward direction, the total number of bytes in the reverse direction, and the total number of packets in the reverse direction, among the features of the communication of the alert, for each terminal device.The clustering unit 135 then performs clustering of the alerts using values ​​obtained by converting the differences into a logarithmic scale (see FIG. 2).

[0038] An example of the clustering result by the clustering unit 135 is shown by reference numeral 301 in Fig. 3. Here, the clustering result (cluster 0 to cluster 7) is shown, in which the clustering unit 135 clusters 165 alerts related to five terminal devices (IoT-A, IoT-B, IoT-C, IoT-D, IoT-E) based on five feature quantities, namely, session duration, total number of bytes in the forward direction, total number of packets in the forward direction, total number of bytes in the reverse direction, and total number of packets in the reverse direction. Reference numeral 301 shows the state of the above clustering result as viewed from the direction along which the session duration and the total number of bytes in the reverse direction are set as axes.

[0039] The clustering algorithm used by the clustering unit 135 is, for example, Birch, k-means, etc. The user of the analysis server 10 can set which clustering algorithm the clustering unit 135 uses.

[0040] Returning to the explanation of Fig. 1, the determination unit 136 determines whether or not each cluster generated by the clustering unit 135 is made up of alerts of the same type.

[0041] For example, the determination unit 136 determines whether each cluster is composed of alerts of the same type based on the ratio of the number of alerts in each cluster to all alerts, the number of terminal devices that are the subject of alerts in each cluster, the density of each cluster, etc.

[0042] For example, the determination unit 136 calculates the value of (the number of alerts constituting the cluster / the total number of alerts) for each cluster, and determines that a cluster for which this value is equal to or greater than a predetermined threshold value (e.g., 70%) is a cluster composed of alerts of the same type.

[0043] Also, for example, the determination unit 136 calculates, for each cluster, how many terminal devices that are the subject of the alert are included in that cluster, and determines that a cluster in which the calculated number is equal to or less than a predetermined threshold (for example, 1 device) is composed of the same type of alert.

[0044] Also, for example, the determination unit 136 creates a scatter plot of the clusters, calculates the density of the clusters using the area of ​​the clusters on the scatter plot and the number of alerts included in the clusters, and determines that a cluster with the calculated density equal to or greater than a predetermined threshold is a cluster composed of alerts of the same type. For example, the determination unit 136 calculates (the number of alerts constituting the cluster / the area of ​​the cluster) for each cluster, and determines that a cluster with the calculated value equal to or greater than a predetermined threshold (for example, 1000) is a cluster composed of alerts of the same type.

[0045] For example, the following case will be described in which the determination unit 136 determines, for the clusters (cluster 1 to cluster 7) indicated by reference numeral 301 in FIG. 3, that a cluster that satisfies any one of the following conditions (1) to (3) is a cluster composed of alerts of the same type, based on the information indicated by reference numeral 302.

[0046] (1) The ratio of the number of alerts in the cluster to the total number of alerts is 70% or more. (2) The number of terminal devices that were the subject of the alert in the cluster was 1 (3) The density of the cluster is 1,000 or more.

[0047] The information indicated by the reference numeral 302 is information indicating the number of alerts, area, and density of each terminal device (IoT-A, IoT-B, IoT-C, IoT-D, IoT-E) for each cluster (cluster 0 to cluster 7) indicated by the reference numeral 301.

[0048] Referring to the information indicated by the reference numeral 302, there is no cluster in which the ratio of the number of alerts in the cluster to the total alerts is 70% or more. However, cluster 0 (cluster 0) is a cluster containing only IoT-C alerts, and its density is also 1000 or more. Cluster 3 (cluster 3) is also a cluster containing only IoT-E alerts, and its density is also 1000 or more. Therefore, the determination unit 136 determines that cluster 0 and cluster 3 are clusters composed of alerts of the same type.

[0049] Returning to the explanation of Fig. 1, the result output unit 137 outputs the clustering result obtained by the clustering unit 135 and the determination result obtained by the determination unit 136 as to whether or not each cluster is composed of alerts of the same type.

[0050] For example, the result output unit 137 outputs the clustering result indicated by reference numeral 301 in FIG. 3 and information indicating that, among the clusters (cluster 1 to cluster 7) indicated by reference numeral 301, cluster 0 and cluster 3 are clusters composed of alerts of the same type.

[0051] When clustering alerts, such an analysis server 10 can generate clusters in which the characteristics of each cluster are clearly expressed. Furthermore, the analysis server 10 outputs the determination result as to whether each cluster is composed of the same type of alert. This allows the operator to identify clusters for which a single overdetection determination is sufficient. As a result, the work required to determine whether an alert is the result of an overdetection can be reduced.

[0052] [Example of processing procedure] Next, an example of the processing procedure of the analysis server 10 will be described with reference to Figures 4 and 5. First, the processing by the analysis server 10 to accumulate alerts will be described with reference to Figure 4, and then the processing by the analysis server 10 to cluster the accumulated alerts will be described with reference to Figure 5.

[0053] First, the feature receiving unit 131 of the analysis server 10 accumulates the features of communication (features of normal communication) transmitted from the sensing device 2 in the storage unit 12 (S1 in FIG. 4). Next, the learning unit 132 uses the features of communication accumulated in S1 to learn (initial learning) a model indicating the characteristics of normal communication (S2).

[0054] When the learning of the model in S2 is completed, the feature receiving unit 131 accumulates the feature of the communication to be detected in the storage unit 12 (S3). Then, the analysis unit 133 calculates an anomaly score indicating a deviation between the feature of the communication to be detected and the model learned in S2 (S4).

[0055] Here, if the analysis unit 133 determines that the anomaly score calculated in S4 exceeds a predetermined threshold (Yes in S5), it generates an alert for the communication and stores it in the storage unit 12 (S6). Then, if processing has not yet been performed for any of the communications to be detected (No in S7), the analysis server 10 performs processing from S4 onward for the communications to be detected that have not yet been processed. On the other hand, if processing has been performed for all of the communications to be detected (Yes in S7), the analysis server 10 ends processing.

[0056] In addition, if the analysis unit 133 determines that the anomaly score calculated in S4 is equal to or less than a predetermined threshold value (No in S5) and processing has not yet been performed on any of the communications to be detected (No in S7), it performs processing from S4 onwards on the communications to be detected that have not yet been processed.

[0057] Next, the process of clustering accumulated alerts performed by the analysis server 10 will be described with reference to FIG.

[0058] First, the clustering unit 135 of the analysis server 10 excludes from the clustering targets, among the alerts stored in the memory unit 12, alerts whose communication tuples differ from the communication tuples (initial tuples) of the features used to train the model (S11 in FIG. 5).

[0059] For example, the clustering unit 135 compares the communication tuples (initial tuples) used to train the model with the communication tuples of the communication features included in the alerts, and excludes from the clustering any alerts whose categorical variables (combinations of the IP address of the terminal device, the source IP address of the session, the destination IP address of the session, the destination port number of the session, and the protocol number) differ from the initial tuple.

[0060] After S11, the clustering unit 135 clusters the alerts based on the features of the communication included in the alerts, namely, the session duration, the total number of bytes in the forward direction, the total number of packets in the forward direction, the total number of bytes in the reverse direction, and the total number of packets in the reverse direction (S12).

[0061] After S12, the determination unit 136 determines whether or not each of the clusters generated by the clustering in S12 is made up of alerts of the same type (S13).

[0062] For example, the determination unit 136 determines whether each cluster is composed of alerts of the same type based on the ratio of the number of alerts in each cluster to all alerts, the number of terminal devices that are the subject of alerts in each cluster, the density of each cluster, etc.

[0063] After S13, the result output unit 137 outputs the clustering results of S12 and the determination results of whether or not each cluster in S13 is composed of alerts of the same type (S14).

[0064] [Example of how operators respond to alerts] Next, with reference to FIG. 6, an example of how an operator responds to an alert will be described, using the results of the alert clustering output by the analysis server 10 and the determination result as to whether each cluster is composed of the same type of alert.

[0065] For example, after an alert occurs, the analysis server 10 performs clustering of the alerts and determines whether each cluster is made up of alerts of the same type (S101: clustering).

[0066] Then, the operator checks the alerts using the result of the clustering of the alerts and the result of the determination as to whether each cluster is composed of the same type of alert. For example, the operator performs the alert checking process shown in S102 to S108 for each alert or for each cluster until there are no unchecked alerts.

[0067] First, the operator checks whether the communication pattern that is the subject of the alert is a new communication pattern that was not present when the model was learned (S102). The operator also checks whether the communication parameters that are the subject of the alert are new communication parameters of a learned communication pattern (S103). Based on the above check result, the operator makes an inquiry to the department in charge of the terminal device that is the subject of the alert (S104). Then, if the inquiry shows that there is an abnormality in the terminal device (Yes in S105), the operator takes measures according to the abnormality (S106). Then, the alert confirmation process ends. After that, the analysis server 10 continues monitoring.

[0068] On the other hand, if the inquiry shows that there is no abnormality in the terminal device (No in S105), the operator judges whether the alert is due to an overdetection (S107). If the operator judges that the alert is due to an overdetection (Yes in S107), the operator performs overdetection feedback (S108). At this time, the operator performs overdetection feedback collectively for clusters of the same type of alert.

[0069] On the other hand, if the operator determines that the alert is not due to an overdetection (No in S107), no action is taken and the alert confirmation process ends. Then, the analysis server 10 continues to monitor the communications of the terminal device.

[0070] In this way, the analysis server 10 outputs the results of clustering the alerts and the determination result of whether each cluster is composed of the same type of alert, so that the operator can determine whether an alert is an overdetection on a cluster-by-cluster basis for clusters of the same type of alert and provide feedback. As a result, the work required to determine whether an alert is an overdetection can be reduced.

[0071] [Clustering experiment results] Next, the results of an experiment on clustering of alerts by the analysis server 10 will be described with reference to FIG. 7. Here, the analysis server 10 assigns two types of loads (10 0.2 Double, 10 0.4 times), 4 types of loads with large positive differentials (10 1 Double, 10 1.2 Double, 10 1.4 Double, 10 1.6 times), 3 types of loads with small negative differentials (10 -0.2 Double, 10 -0.4 Double, 10 -0.6 The analysis server 10 applied 10 load patterns (changes to communication) of loads of 10 times (times) and loads of differential communication tuples (differences due to categorical variables) and generated alerts. In addition, the analysis server 10 used Birch as the clustering algorithm for alerts, and set the maximum number of clusters to 10.

[0072] Based on the above conditions, the analysis server 10 clustered the alerts related to the above 10 patterns of communication, and the result is shown in Fig. 7. As shown in Fig. 7, out of the 828 alerts excluding the alerts due to the difference communication tuples, the number of alerts included in the largest cluster with an alert ratio of more than 70% was 643.

[0073] Here, the analysis server 10 judges the largest cluster with an alert ratio exceeding 70% as a cluster composed of the same type of alerts and outputs the judgment result, allowing the operator to collectively judge whether the cluster is an overdetection. As a result, the above 643 alerts are compressed to 9 (one per load), so the number of alerts judged as overdetection by the operator is reduced from 939 to 305 (a 67.5% reduction).

[0074] [System configuration, etc.] In addition, each component of each unit shown in the figure is a functional concept, and does not necessarily have to be physically configured as shown in the figure. 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 a part of it can be functionally or physically distributed and integrated in any unit according to various loads, usage conditions, etc. Furthermore, each processing function performed by each device can be realized in whole or in part by a CPU and a program executed by the CPU, or can be realized as hardware using wired logic.

[0075] Furthermore, among the processes described in the above embodiments, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically by a known method. In addition, the information including the processing procedures, control procedures, specific names, various data and parameters shown in the above documents and drawings can be changed arbitrarily unless otherwise specified.

[0076] [program] The above-mentioned analysis server 10 can be implemented by installing a program as package software or online software on a desired computer. For example, the above-mentioned program can be executed by an information processing device, causing the information processing device to function as the analysis server 10. The information processing device here includes desktop and notebook personal computers. In addition, the information processing device also includes mobile communication terminals such as smartphones, mobile phones, and PHS (Personal Handyphone System), and further terminals such as PDAs (Personal Digital Assistants).

[0077] The analysis server 10 can also be implemented as a server device that provides services related to the above-mentioned processes to a client, the client being a terminal device used by the user. In this case, the server device may be implemented as a Web server, or may be implemented as a cloud that provides services related to the above-mentioned processes by outsourcing.

[0078] 8 is a diagram showing an example of a computer that executes an analysis 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.

[0079] The memory 1010 includes a read only memory (ROM) 1011 and a random access memory (RAM) 1012. The ROM 1011 stores a boot program such as a basic input output system (BIOS). 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 1100. A removable storage medium such as a magnetic disk or an optical disk is inserted into the disk drive 1100. 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.

[0080] The hard disk drive 1090 stores, for example, an OS 1091, an application program 1092, a program module 1093, and program data 1094. That is, the programs that define each process executed by the above-mentioned analysis server 10 are implemented as program modules 1093 in which computer-executable codes are written. The program modules 1093 are stored, for example, in the hard disk drive 1090. For example, the program modules 1093 for executing processes similar to those of the functional configuration of the analysis server 10 are stored in the hard disk drive 1090. The hard disk drive 1090 may be replaced by an SSD (Solid State Drive).

[0081] Data used in the processing of the above-described embodiment is stored as program data 1094 in, for example, the memory 1010 or the hard disk drive 1090. Then, the CPU 1020 reads out the program module 1093 or the program data 1094 stored in the memory 1010 or the hard disk drive 1090 into the RAM 1012 as necessary and executes them.

[0082] Note that the program module 1093 and the program data 1094 are not limited to being stored in the hard disk drive 1090, but may be stored in, for example, a removable storage medium and read by the CPU 1020 via the disk drive 1100 or the like. Alternatively, the program module 1093 and the program data 1094 may be stored in another computer connected via a network (such as a local area network (LAN) or wide area network (WAN)). The program module 1093 and the program data 1094 may then be read by the CPU 1020 from the other computer via the network interface 1070. [Explanation of symbols]

[0083] 1. Analysis system 2. Sensing equipment 10 Analysis Server 12 Storage section 131 Feature Reception Unit 132 Learning Department 133 Analysis Department (Storage Department) 134 Screen display processing unit 135 Clustering Department 136 Judgment section 137 Result output section

Claims

1. a storage unit that stores an alert of a communication that is determined to be abnormal based on a model that indicates characteristics of normal communication; a clustering unit that performs clustering of the alerts using feature amounts of communications included in the accumulated alerts; a determination unit that determines, for each cluster generated by the clustering, whether or not the cluster is composed of alerts of the same type; a result output unit that outputs the clustering result and a determination result as to whether or not each of the clusters is composed of alerts of the same type; An analytical device comprising:

2. The clustering unit includes: A feature amount of the communication used for learning the model is compared with a feature amount of the communication included in the alert, and an alert having a different combination of an IP address of a terminal device of the communication, a source IP address of the session, a destination IP address of the session, a destination port number of the session, and a protocol number is excluded from the targets of the clustering, and the clustering is performed. The analysis device according to claim 1 .

3. The clustering unit includes: Clustering the alerts using at least one of the features of communication included in the alerts, namely, a session duration, a total number of bytes in the forward direction, a total number of packets in the forward direction, a total number of bytes in the reverse direction, and a total number of packets in the reverse direction. The analysis device according to claim 1 .

4. The clustering unit includes: Calculating the difference between the value of the feature of the communication used in learning the model and the value of the feature of the communication included in the alert for each terminal device, and clustering the alerts using the value obtained by converting the difference in the feature into a logarithmic scale. The analysis device according to claim 1 .

5. The determination unit is For each of the generated clusters, a cluster in which the ratio of the number of alerts constituting the cluster to the number of all alerts is equal to or greater than a predetermined threshold is determined to be a cluster constituted by alerts of the same type.

2. The analysis device according to claim 1 .

6. The determination unit is For each of the generated clusters, the number of terminal devices that are the subject of the alert is calculated, and a cluster in which the calculated number is equal to or less than a predetermined threshold is determined to be a cluster composed of alerts of the same type.

2. The analysis device according to claim 1 .

7. The determination unit is 2. The analysis device according to claim 1, further comprising: a scatter plot of the generated clusters; a density of the clusters is calculated using an area of ​​the clusters on the scatter plot and the number of alerts contained in the clusters; and a cluster with the calculated density equal to or greater than a predetermined threshold is determined to be a cluster composed of alerts of the same type.

8. An analytical method performed by an analytical device, comprising: accumulating alerts for communications that are determined to be abnormal based on a model that indicates characteristics of normal communications; A step of performing clustering of the alerts using features of communications included in the accumulated alerts; determining, for each cluster generated by the clustering, whether or not the cluster is composed of alerts of the same type; outputting the clustering result and a determination result as to whether each of the clusters is composed of alerts of the same type; An analytical method comprising the steps of:

9. accumulating alerts for communications that are determined to be abnormal based on a model that indicates characteristics of normal communications; A step of performing clustering of the alerts using features of communications included in the accumulated alerts; determining, for each cluster generated by the clustering, whether or not the cluster is composed of alerts of the same type; outputting the clustering result and a determination result as to whether each of the clusters is composed of alerts of the same type; An analysis program for running the above on a computer.

Citation Information

Patent Citations

  • Abnormal behavior detection method and system for Internet-of-Things equipment

    CN112600792A

  • FR/17/2020/12/

  • Security processing method and server

    JP2017111796A

  • Identification of Troubleshooting Options for Network Failures

    JP2017509262A

  • Communication system and communication method

    JP2020005184A