WO2017163352A1 - 異常検出装置、異常検出システム、及び、異常検出方法 - Google Patents
異常検出装置、異常検出システム、及び、異常検出方法 Download PDFInfo
- Publication number
- WO2017163352A1 WO2017163352A1 PCT/JP2016/059330 JP2016059330W WO2017163352A1 WO 2017163352 A1 WO2017163352 A1 WO 2017163352A1 JP 2016059330 W JP2016059330 W JP 2016059330W WO 2017163352 A1 WO2017163352 A1 WO 2017163352A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- flow
- data
- correlation coefficient
- time
- flows
- Prior art date
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
- H04L43/0823—Errors, e.g. transmission errors
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/06—Management of faults, events, alarms or notifications
- H04L41/0631—Management of faults, events, alarms or notifications using root cause analysis; using analysis of correlation between notifications, alarms or events based on decision criteria, e.g. hierarchy, tree or time analysis
- H04L41/064—Management of faults, events, alarms or notifications using root cause analysis; using analysis of correlation between notifications, alarms or events based on decision criteria, e.g. hierarchy, tree or time analysis involving time analysis
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/40—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using virtualisation of network functions or resources, e.g. SDN or NFV entities
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/16—Threshold monitoring
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/20—Arrangements for monitoring or testing data switching networks the monitoring system or the monitored elements being virtualised, abstracted or software-defined entities, e.g. SDN or NFV
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/02—Standardisation; Integration
- H04L41/0213—Standardised network management protocols, e.g. simple network management protocol [SNMP]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/08—Configuration management of networks or network elements
- H04L41/0893—Assignment of logical groups to network elements
Definitions
- the present invention relates to data abnormality detection.
- cloud systems In recent years, with the progress of cloud computing systems (hereinafter referred to as “cloud systems”) and virtual machines, so-called silent failures such as failures due to application performance degradation and failures due to source code bugs included in application version updates Detection is sought.
- a performance item or a managed device is used as an element, and first time series information indicating a time series change of performance information related to at least a first element, and time series change of performance information related to a second element.
- a correlation model generating unit that derives a correlation function with the second performance series information shown, generates a correlation model based on the correlation function, and obtains the correlation model for a combination between the elements, and each correlation between the elements
- An operation management apparatus including a model search unit that sequentially searches for a model to determine an optimal correlation model, and predicts performance information of the second element from performance information of the first element based on the determined correlation model Is disclosed.
- an object of the present invention is to reduce the processing load of correlation analysis in data abnormality detection.
- the processor Classify multiple data flows based on the similarity of time-series changes in the amount of data in the data flow, For at least two data flows belonging to the same classification, calculate a correlation coefficient at normal time and a correlation coefficient at a certain timing, When the difference between the correlation coefficient at normal time and the correlation coefficient at the certain timing is larger than a predetermined threshold, it is determined that at least one of the at least two data flows is abnormal.
- the figure which shows the structural example of the data center which concerns on a present Example The figure which shows the structural example of a network apparatus.
- the “xxx table” or the “xxx list” can be called “xxx information”.
- the processing is described using “program” as a subject.
- the program is executed by a processor (for example, a CPU (Central Processing Unit)), so that a predetermined processing can be appropriately performed as a storage resource (for example, a memory).
- the processing subject may be a processor and an apparatus having the processor. Part or all of the processing performed by the processor may be performed by a hardware circuit.
- the computer program may be installed from a program source.
- the program source may be a program distribution server or a storage medium (for example, a portable storage medium).
- the system performs a correlation analysis on a time-series change in the amount of communication of a data flow (hereinafter may be simply referred to as “flow”), and the correlation coefficient of the analysis result is a correlation between normal time (normal time).
- flow normal time
- the system can detect, for example, an application system that exhibits unusual behavior.
- the system uses the same flow group as the first process for classifying flows having similar characteristics (periodic characteristics, frequency component characteristics, etc.) of time-series changes in flow traffic into the same flow group.
- a second process for performing a correlation analysis between the flows to which the flows belong is executed. Thereby, in the correlation analysis for detecting an abnormal flow, the number of flow combinations can be reduced. That is, the amount of calculation of correlation analysis can be reduced, and the time required for the correlation analysis process can be shortened.
- the system according to the present embodiment calculates an appropriate window size (contrast time) of two data flows to be subjected to correlation analysis based on the flow traffic. Since the amount of data communication flowing through the cloud system is enormous, it is often sampled and measured. When the amount of data communication is calculated from the number of packets sampled, a flow with a relatively small amount of data communication compared to other flows is hardly sampled. In this case, it is conceivable to increase the data communication amount measurement time (discretization width). However, if the flow discretization width is increased, it becomes difficult to detect instantaneous abnormalities. Therefore, in this embodiment, an appropriate discretization width (flow discretization width) for each flow is calculated based on the data communication amount.
- the flow discretization width is increased when the data communication amount is small, and the flow discretization width is decreased when the data communication amount is large. As a result, it is possible to detect both an instantaneous abnormality of a flow having a relatively large data communication amount and an abnormality over a long time of a flow having a relatively small data communication amount.
- a discretization width common to flows belonging to the flow group (flow group discretization width) is calculated. If the flow discretization width of each flow is different, a process for matching the discretization widths of at least two flows to be subjected to correlation analysis is required. That is, calculation processing for matching the flow discretization width is required for each combination of flows to be subjected to correlation analysis.
- a common flow group discretization width is set for each flow belonging to a flow group. As a result, the calculation process for matching the flow discretization width for each combination of flows can be omitted, and the processing time required for the correlation analysis can be shortened.
- the administrator is notified of information related to the flow in which an abnormality has been detected.
- the information related to the flow is, for example, information such as a 5-tuple and / or virtual network ID (VLAN tag, etc., the same applies hereinafter) of the flow.
- the administrator can specify the function, device, and the like that showed a behavior different from normal from the notified flow information.
- the flow is uniquely determined by the destination MAC address, source MAC address, destination IP address, source IP address, L4 port number, and virtual network ID included in the packet header of data communication.
- Data communication determined by Alternatively, the flow may be data communication uniquely determined by a destination IP address, a source IP address, an L4 port number, and a virtual network ID.
- the flow may be data communication uniquely determined by the destination IP address, the source IP address, and the virtual network ID.
- FIG. 1 shows a configuration example of a data center according to the present embodiment.
- the data center includes a management system 10, an analysis system 100, a control network 21, a plurality of network devices 30, and a plurality of computers 50.
- the plurality of network devices 30 and the plurality of computers 50 may constitute a data network 3 connected by a communication network.
- the data network 3 may be connected to the control network 21.
- the management system 10 and / or the network device 30 may be virtually implemented.
- the network device 30 may be virtually mounted on the computer 50. Details of the network device 30 and the analysis system 100 will be described later with reference to FIGS.
- the management system 10 is a system used by an administrator to manage the data network 3 constituting the customer system.
- the management system 10 is connected to the analysis system 100 via a predetermined network 20.
- the management system 10 may present various information transmitted from the analysis system 100 to the administrator.
- the management system 10 notifies the administrator of the abnormal flow information transmitted from the analysis system 10.
- the administrator may analyze the abnormality that has occurred in the customer system based on the notified abnormality flow information.
- the administrator may register the content of the abnormality that has occurred in the customer system when the abnormal flow is detected, in the analysis system 100 via the GUI of the management system 10.
- the administrator may be able to refer to the correspondence relationship between the abnormality of the customer system that has occurred in the past and the information of the abnormal flow notified at that time via the management system 10.
- the data network 3 may be logically separated for each customer system.
- one data network 3 may be one customer system.
- the customer system may be an application system for each customer configured by at least one application.
- one customer system may be configured for each company using the data center.
- An example of the protocol of the data network 3 is native IP communication.
- the control network 21 is a network that connects the network device 30 and the analysis system 100. Data of each data network 3 may be collected by the analysis system 100 via the control network 21.
- the computer 50 has computational resources such as a CPU, memory, and storage, and executes applications in the customer system.
- the application is, for example, a program such as a WEB server, an application server, or a DB (Database) server.
- the application may be implemented in a VM (Virtual Machine).
- FIG. 2 shows a configuration example of the network device 30.
- the network device 30 is a communication device realized by, for example, a router or a switch.
- the network device 30 may include a switch 31, a switch management unit 32, a flow statistics management unit 33, a transfer unit 34, a port 35, and a management port 36 as functions.
- the switch 31 may be an Ethernet (registered trademark) fabric switch that transfers a communication packet received from the port 35 to an output destination port that matches the header information of the communication packet.
- Ethernet registered trademark
- the switch management unit 32 manages the switch 31.
- the switch management unit 32 may process a data reference request or a setting request transmitted from the management terminal.
- Examples of the protocol exchanged with the management terminal include SNMP (Simple Network Management Protocol) and sFlow.
- the flow statistics management unit 33 counts the communication amount or the number of communication packets for each flow of communication packets received by the network device 30.
- the flow statistics management unit 33 may correspond to the sFlow protocol.
- the transfer unit 34 transmits the value (measurement value) counted by the flow statistics management unit 33 to the analysis system 100.
- the port 35 is a physical port for transmitting and receiving communication packets to and from the computer 50.
- the management port 36 is a physical port for transmitting / receiving data to / from a management terminal, for example.
- the management port 36 is a physical port for transmitting the measurement value of the flow statistics management unit 33 to the analysis system 100.
- FIG. 3 shows a configuration example of the analysis system 100.
- the analysis system 100 is a system for analyzing a data flow (data communication amount) in the data network 3.
- the analysis system 100 may be configured by a computer including a CPU 150, a communication I / F 130, an input I / F 140, a memory 110, a storage 120, and the like.
- the memory 110 is, for example, a DRAM (Dynamic Random Access Memory), an FeRAM (Ferroelectric Random Access Memory), an MRAM (Magnetic Resistive Random Access Memory), or the like.
- the storage 120 is, for example, an SSD (solid state drive), an HDD (Hard Disk Drive), or the like.
- the input I / F 140 is used for notifying the administrator of the detected abnormal flow via the operation screen of the management system 10 connected to the analysis system 100 or for receiving an input of a failure content from the administrator. (North band) interface.
- the communication I / F 103 is a (south band) interface for receiving a measurement result from the network device 30.
- the memory 110 may store a flow group generation unit 111, a correlation calculation unit 112, and an abnormality detection unit 113 as functions. These functions may be realized by a program held in the storage 120 being read into the memory 110 and executed by the CPU 150. The program may be stored in the storage 120 in advance, or may be installed from the outside via a predetermined network or a portable storage medium. Note that these functions 111, 112, and 113 may be collectively referred to as a flow analysis unit.
- a flow information table 121 In the storage 120, a flow information table 121, a flow characteristic table 122, a flow group information table 123, a correlation information table 124, an abnormality information table 125, a traffic volume table 126, and a post-discretization traffic volume table 127 are stored as data. May be.
- each table will be described.
- the following table is an example, and each table may be normalized as a plurality of tables, or may be combined with other tables.
- FIG. 4 shows a configuration example of the flow information table 121.
- the flow information table 121 manages information related to flows (referred to as “flow information”).
- the flow information table 121 includes, as data items, a flow ID 200, a destination IP address 201, a source IP address 202, a destination MAC address 203, a source MAC address 204, a destination port number 205, a source port number 206, a transformer.
- the port layer 207, the network layer 208, and the virtual network ID 209 may be included.
- the flow ID 200 is a value for uniquely identifying a flow that flows through the data network 3.
- the flow ID 200 may be given by the flow statistics management unit 33 of the network device 30.
- the destination IP address 201 indicates the IP address of the destination of the flow with the flow ID 200.
- the source IP address 202 indicates the IP address of the source of the flow with the flow ID 200.
- the destination MAC address 203 indicates the destination MAC address of the flow with the flow ID 200.
- the sender MAC address 204 indicates the MAC address of the sender of the flow with the flow ID 200.
- the destination port number 205 indicates the destination port number of the flow with the flow ID 200.
- the transmission source port number 206 indicates the port number of the transmission source of the flow with the flow ID 200.
- the transport layer 207 indicates the type of transport layer (TCP, UDP, etc.) of the flow with the flow ID 200.
- the network layer 208 indicates the type of the network layer of the flow with the flow ID 200 (IPv4, IPv6, ICMP (Internet Control Message Protocol), etc.).
- the virtual network ID 209 indicates the ID of the virtual network to which the flow with the flow ID 200 belongs.
- one record of the flow information table 121 may be information found from one IP packet. That is, a plurality of entries having the same flow ID 200 may exist in the flow information table 121.
- FIG. 5 shows a configuration example of the flow characteristic table 122.
- the flow characteristic table 122 manages information (referred to as “flow characteristics”) relating to characteristics of time-series changes in flow traffic.
- the flow characteristic table 122 may include a flow ID 300, a measurement time 310, a communication amount average 320, a communication amount standard deviation 330, a flow group ID 340, a flow discretization width 350, and a frequency component 360 as data items.
- the flow ID 300 is the same as the flow ID 200 in FIG.
- the measurement time 310 indicates the measurement time of the flow communication amount of the flow ID 300.
- the traffic average 320 indicates the average per unit time of the flow traffic of the flow ID 300.
- the communication traffic average 320 may be calculated from the flow communication traffic measured within the measurement time 310.
- the traffic standard deviation 330 indicates the standard deviation per unit time of the flow traffic of the flow ID 300.
- the standard deviation 330 of the flow communication amount may be calculated from the flow communication amount measured within the measurement time 310.
- the flow group ID 340 is a number for uniquely identifying a flow group.
- the flows with the flow ID 300 having the same flow group ID 340 belong to the same flow group.
- the flow group into which the flow ID 300 is classified may be determined based on the measurement time 310, the traffic average 320, and the traffic standard deviation 330. Details of the classification method will be described later.
- the flow discretization width 350 indicates the discretization width (time) of the flow with the flow ID 300.
- the flow discretization width 350 is used when calculating a correlation coefficient between flows.
- the initial value of the flow discretization width 350 may be set by the administrator. Details of the calculation method of the flow discretization width 350 will be described later.
- the frequency component 360 indicates a frequency component of a time-series change in the flow communication amount of the flow ID 300.
- the frequency component 360 may store a frequency band including a frequency component equal to or higher than a predetermined threshold. A method for calculating the frequency component 360 will be described later.
- FIG. 6 shows a configuration example of the flow group information table 123.
- the flow group information table 123 manages information related to the flow group.
- the flow group information table 123 may include a flow group ID 400, a flow group discretization width 410, and a window size 420 as data items.
- the flow group ID 400 is the same as the flow group ID 340 in FIG.
- the flow group discretization width 410 indicates the discretization width for the flow group with the flow group ID 400.
- the window size 420 indicates the window size for the flow group with the flow group ID 400.
- the common flow group discretization width 410 and window size 420 are applied to all the flows of the flow ID 300 belonging to the flow group ID 400.
- the window size (contrast time) that is a correlation coefficient calculation target may be calculated as a predetermined multiple of the flow group discretization width 410.
- the window size 420 (associated with the flow group ID in the flow group information table 123 ( (Contrast time) may be used. That is, according to the present embodiment, it is not necessary to match the discretization width every time the correlation coefficient is calculated.
- FIG. 7 shows a configuration example of the correlation information table 124.
- the correlation information table 124 manages information related to the result of correlation analysis.
- the correlation information table 124 includes, as data items, a flow ID 500, a flow ID 501, a correlation coefficient 502, a correlation coefficient calculation count 503, a correlation coefficient average 504, a correlation coefficient standard deviation 505, and a correlation coefficient change time 506. May be included.
- the flow ID 500 and the flow ID 501 are the same as the flow ID 200 in FIG.
- the correlation coefficient 502 indicates a correlation coefficient between the flow with the flow ID 500 and the flow with the flow ID 501.
- the flow ID 500 and the anti-flow ID 501 belong to the same flow group. Therefore, the correlation coefficient 502 is a value calculated using the window size 420 associated with the flow group ID 400 to which the flow ID 500 and the counter flow ID 501 belong in the flow group information table 123.
- the correlation coefficient calculation count 503 indicates the number of times the correlation coefficient 502 has been calculated.
- Correlation coefficient average 504 indicates the average of correlation coefficient 502. That is, the correlation coefficient average 504 is an average when the correlation coefficient 502 calculated this time is included in the original correlation coefficient average 504. That is, the correlation coefficient average 504 may be updated every time the correlation coefficient 502 is calculated.
- the correlation coefficient standard deviation 505 indicates the standard deviation of the correlation coefficient 502. That is, the correlation coefficient standard deviation 505 is a standard deviation when the correlation coefficient calculated this time is included in the original correlation coefficient standard deviation 505. That is, the correlation coefficient standard deviation 505 may be updated every time the correlation coefficient 502 is calculated.
- Correlation coefficient change time 506 is a time (timing) when a significant change has occurred in the correlation coefficient 502. For example, the time when the flow ID 500 or the flow ID 501 related to the correlation coefficient 502 is detected when the difference between the correlation coefficient 502 and the correlation coefficient average 504 is larger than a predetermined threshold.
- the correlation coefficient change time 506 may be blank if the correlation coefficient 502 has not changed significantly.
- FIG. 8 shows a configuration example of the abnormality information table 125.
- the abnormality information table 125 manages information regarding a flow (abnormal flow) detected as abnormal.
- the abnormality information table 125 may include a flow ID 600, an anti-flow ID 601, an abnormality content 602, an abnormality duration 603, and an abnormality improvement method 604 as data items.
- Flow ID 600 and anti-flow ID 601 are flow IDs detected as abnormal.
- the flow ID 600 and the pair flow ID 601 may be the flow ID 500 and the pair flow ID 501 in which the time is stored in the correlation coefficient change time 506 of the correlation information table 124.
- the abnormality content 602 indicates the content of the abnormality that occurred in the customer system and is associated with the flow ID 600 and the flow ID 601.
- the abnormal continuation time 603 indicates the time that the abnormality of the abnormality content 602 has continued in the customer system.
- the abnormality improvement method 604 indicates information on how to improve the abnormality of the abnormality content 602 in the customer system.
- the abnormality content 602 may store the content of the abnormality that occurred in the customer system at the correlation coefficient change time 506 corresponding to the flow ID 600 and the flow ID 601 in the correlation information table 124.
- the abnormality content 602, the abnormality duration 603, and / or the abnormality improvement method 604 may be input by the administrator.
- the analysis system 100 presents the correlation coefficient change time 506 to the administrator via the management system 10, and the abnormality content that occurred in the customer system at the correlation coefficient change time and the abnormality continue to the administrator. And / or an improvement method for the abnormality may be input.
- FIG. 9 shows a configuration example of the traffic table 126.
- the communication amount table 126 manages the data communication amount at each time of each flow.
- the communication amount table 126 may include a flow ID 700, a time 701, and a communication amount 702 as data items.
- the flow ID 700 is the same as the flow ID 200 in FIG.
- the time 701 is the time when the traffic amount 702 of the flow with the flow ID 700 is measured.
- the time 701 may be a time when the analysis system 100 receives information on the traffic from the network device 30 or may be a time when the network device 30 measures the traffic.
- the communication amount 702 is the communication amount at the time 701 of the flow with the flow ID 700.
- the communication amount 702 may be a value actually measured by the network device 30 or a value calculated from sampled data (packets).
- the data items of the post-discretization traffic table 127 may be the same as the traffic table 126 of FIG. Therefore, the drawing of the post-discretization communication amount table 127 is omitted.
- FIG. 10 is a sequence chart showing an example of flow group generation processing.
- the flow group generation process may be executed when the analysis system 100 is introduced, periodically, when a new application is deployed or configured, or when a predetermined event occurs.
- FIG. 10 shows an example of processing in which the network device 30 measures the communication amount of data transmitted from the computer 50-1 to the computer 50-2, and the analysis system 100 generates a flow group based on the measurement result.
- the computer 50-1 transmits to the network device 30 data having the destination 5050 as the computer 50-2.
- the data may be an IP packet.
- Step 1010 The network device 30 transfers the data transmitted from the source computer 50-1 to the destination computer 50-2.
- the network device 30 measures the flow communication amount of the transfer data, and transmits the flow information and the measurement result to the analysis system 100.
- the flow information may be information included in the header of transfer data (IP packet) (that is, a value corresponding to a data item in the flow information table 120).
- the flow measurement result may be statistical information based on sampling (for example, measurement time 310, traffic average 320, traffic standard deviation 330).
- the network device 30 may execute the processing of step 1020 for each data transfer, periodically, or whenever the number of data transfers reaches a predetermined number.
- the flow ID may be given by the network device 30 or may be given by the analysis system 100.
- the network device 30 may transmit the flow measurement result to the analysis system 100 according to the sFlow protocol.
- Step 2010 The analysis system 100 executes a flow group generation process. Next, the process will be described.
- FIG. 11 is a flowchart showing an example of the flow group generation process. This process corresponds to the process of step 2010 in FIG.
- the flow group generation unit 111 calculates the traffic of each flow.
- the flow group generation unit 111 may execute the following processes (A1) to (A4) for each flow ID.
- the flow group generation unit 111 counts the entries having the same flow ID 200 from the flow information table 121.
- the flow group generation unit 111 calculates the number of packets of the flow with the flow ID based on the number of entries.
- the number of packets may be calculated as “sampling rate in network device 30 ⁇ number of entries”.
- the sampling rate may be initially set in the network device 30 and the analysis system 100.
- the flow group generation unit 111 calculates the communication amount of the flow ID based on the number of packets, the average packet length, and the measurement time.
- the communication amount may be calculated as “number of packets ⁇ average packet length / measurement time”.
- the average packet length and the measurement time may be initially set in the network device 30 and the analysis system 100, or may be measured by the network device 30.
- the flow group generation unit 111 stores the flow ID, the time when the measurement result is received in step 1020, and the calculated communication amount in association with each other in the communication amount table 126.
- the time when the measurement result is received may be the time when the network device 30 receives the data.
- the flow group generation unit 111 calculates a traffic average 320 and a traffic standard deviation 330 for each flow.
- the flow group generation unit 111 may execute the following processes (B1) to (B2) for each flow ID.
- the flow group generation unit 111 extracts the entries having the same flow ID 700 from the communication amount table 126. Then, the flow group extraction unit 111 identifies the oldest time and the latest time from the time 701 of the extracted entry.
- the flow group generation unit 111 stores the time from the oldest time to the latest time in the measurement time 310 corresponding to the flow ID specified in (B1) of the flow characteristic table 122.
- the flow group extraction unit 111 calculates the average calculated from the communication amount 702 extracted in the above (B1) to the communication amount average 320 and the communication amount standard deviation 330 corresponding to the flow ID specified in the above (B1) in the flow characteristic table 122. And the standard deviation is stored (overwritten).
- Step 5020 The flow group generation unit 111 calculates the flow discretization width 350 and the frequency component 360 of each flow.
- a method for calculating the flow discretization width 350 and the frequency component 306 will be described.
- an appropriate sampling time (statistic reliability is equal to or greater than a predetermined value) for each flow is calculated based on the amount of traffic of each flow.
- This sampling time is called “flow discretization width”.
- the flow discretization width 350 may be calculated as “analyzable communication amount / communication amount average”. This analyzable communication amount may be a predetermined value.
- This traffic average may be the traffic average 320 associated with the flow ID in the flow characteristic table 122.
- a method for determining whether or not the flow is abnormal can be considered as follows. That is, for all combinations of measured flows, correlation coefficients related to time-series changes in the traffic amount at normal time (normal time) are calculated. Then, correlation coefficients are calculated for all combinations, and if the difference between the calculated correlation coefficient and the correlation coefficient at normal time is larger than a predetermined value, the flow related to the combination is determined to be abnormal.
- the frequency component is an index used when classifying each flow into a flow group.
- the characteristics of the time-series change of the traffic volume of the flow are represented by (C1) non-stationary and regular flow characteristics (hereinafter referred to as “flow characteristics with high periodicity”), (C2) stationary And (C3) unsteady and irregular flow characteristics (hereinafter referred to as “flow characteristics with low periodicity”). If the flow characteristics are similar, the correlation is likely to be high. Conversely, if the flow characteristics are not similar, the correlation is likely to be low. Since the period characteristic can be expressed as a frequency characteristic, “high periodicity” can be expressed as “a specific frequency component is strong”.
- (C1) to (C3) will be described.
- the steady flow characteristic can also be expressed as a highly periodic flow characteristic having a very large period and a very small amplitude. Since the amplitude is very small, it is highly possible that the phase shift will not significantly affect the correlation coefficient. Therefore, the correlation coefficient between the steady flow characteristics increases as the frequency components are similar.
- C3 There is a high possibility that characteristic frequency components (periods) and phases do not exist in the flow characteristics with low periodicity. For example, data transmitted and received by an application system triggered by an event such as an access from a user is likely to have low periodic flow characteristics. However, for example, in the WEB three-layer model, data transmitted from the WEB server to the application server and data transmitted from the application server to the DB server are linked (sent at the same timing). Probability is high. In this way, data transmitted at the same timing triggered by the same event is close to the behavior of a pulse wave, so there is no periodicity, but there is a possibility of having high frequency components in almost the same frequency band. high.
- Each flow may be classified as (C1) to (C3) described above, but may be classified under slightly looser conditions. For example, classification may be performed using only frequency components of flow characteristics. Compared with the classification method described above, this classification method increases the possibility that a combination of flows with low correlation exists in the same classification (false positive), and increases the load of correlation coefficient calculation processing. On the contrary, it is less likely that there is no highly correlated flow combination (false negative) in the same classification.
- the flow group generation unit 111 extracts the entries having the same flow ID 700 from the communication amount table 126.
- the flow group generation unit 111 divides the extracted times 701 of the plurality of entries at intervals of the flow discretization width 350 corresponding to the flow ID. Then, the total (or average) of the communication amount 702 of each divided entry is calculated. For example, when the flow discretization width 350 is “1 minute”, the times 701 of the plurality of extracted entries are divided at 1-minute intervals. Then, the total (or average) of the divided one-minute traffic is calculated. Thereby, time-series data of the traffic volume recalculated with the flow discretization width 350 (hereinafter referred to as “the discretized flow traffic volume”) is generated.
- the flow group generation unit 111 performs frequency analysis on the post-discretization flow traffic calculated in (D2), and calculates a frequency component.
- the flow group generation unit 111 stores the flow ID, the time corresponding to the flow discretization width, and the post-discretization flow traffic volume in the post-discretization traffic volume table 127 (not shown).
- the flow group generation unit 111 stores (overwrites) the flow discretization width and frequency component calculated above in the flow discretization width 350 and the frequency component 360 corresponding to the flow ID 300 in the flow characteristic table 122. .
- the flow group generation unit 111 performs the processes (D1) to (D5) for all the flow IDs.
- the flow group generation unit 111 identifies a frequency band having a large frequency component 360 of each flow from the flow characteristic table 122.
- the flow group extraction unit 111 may specify the frequency band to which the upper N (N is a positive integer) frequency components belong.
- the flow group extraction part 111 may specify the frequency band to which the frequency component more than a predetermined threshold belongs.
- the flow group extraction unit 111 classifies each flow into each flow group based on the specified frequency band. For example, the flow group generation unit 111 classifies each flow into two flow groups based on whether the specified frequency band belongs to a larger or smaller one than a predetermined threshold. Also good. For example, the flow group generation unit 111 may classify each flow into a plurality of flow groups based on which of the plurality of different sections the specified frequency band belongs to. For example, the flow group extraction unit 111 may classify each flow into a plurality of flow groups by a known clustering method such as the K-MEANS method using the specified frequency band as an attribute.
- a known clustering method such as the K-MEANS method using the specified frequency band as an attribute.
- the flow group generation unit 111 assigns a common flow group ID to the flow group ID 340 corresponding to the flow ID 300 classified into the same flow group in the flow group extraction unit 122.
- Step 5030 The flow group generation unit 111 calculates a flow group discretization width and a window size for each flow group. This is because in order to calculate the correlation coefficient for the combination of flows, the discretization widths of those flows need to match. Therefore, in this embodiment, a flow group discretization width is set for each flow group as follows.
- the flow group generation unit 111 extracts entries having the same flow group ID from the flow characteristic table 122.
- the flow group generation unit 111 specifies the maximum flow discretization width among the extracted entries.
- the flow group generation unit 111 calculates a window size by multiplying the specified maximum flow discretization width (flow group discretization width) by a predetermined value.
- This predetermined value may be one or more values set in advance.
- the flow group generation unit 111 sets the maximum flow discretization width calculated in (E3) and the flow group discretization width 410 and the window size 420 corresponding to the flow group ID 300 in the flow group information table 123, respectively. Store (overwrite) the window size.
- the flow group generation unit 111 may create a new entry when the flow group ID (E1) does not exist in the flow group information table 123.
- Step 5035 The flow group generation unit 111 uses the communication amount table 126 to perform the flow group of the flow group ID 340 to which the flow ID belongs for each flow ID in the same procedure as (D1) to (D5) of step 5020.
- the time corresponding to the discretization width 410 and the discretized flow traffic are calculated and stored (overwritten) in the discretized traffic table 127 (not shown).
- flows with similar time-series changes in data volume can be classified into the same flow group.
- the flow group discretization width and window size common to the flow group can be calculated.
- FIG. 12 is a sequence chart showing an example of an abnormal flow detection process.
- the abnormal flow detection process may be executed at any time.
- FIG. 12 is an example of processing in which the network device 30 measures the amount of data transmitted from the computer 50-1 to the computer 50-2, and the analysis system 100 detects an abnormal flow based on the measurement result.
- step 2000 to step 2020 is the same as each process from step 1000 to step 1020 in FIG. Therefore, the description is omitted here.
- Step 2030 The analysis system 100 executes an abnormal flow detection process. Details of this processing will be described later (see FIG. 13).
- Step 2040 When the analysis system 100 detects an abnormal flow, the analysis system 100 transmits information related to the abnormal flow (data items of the flow information table 121, correlation coefficient change time 506 of the correlation information table 124, etc.) to the management system 10. .
- Step 2050 The administrator inputs, via the management system 10, the details of an abnormality that occurred in the customer system when the notified abnormal flow occurred.
- the management system 10 transmits the input abnormality content and the like to the analysis system 100.
- the analysis system 100 stores the transmitted abnormality content in an entry corresponding to the abnormality flow ID in the abnormality information table 125. Thereby, the abnormal flow is associated with the content of the abnormality that occurred in the customer system.
- FIG. 13 is a flowchart showing an example of an abnormal flow detection process. This process corresponds to the process of step 2030 in FIG.
- Step 6010 The correlation calculation unit 112 selects a flow group ID to be processed.
- the correlation calculation unit 112 calculates a correlation coefficient between two flow IDs (flow ID 500 and anti-flow ID 501) having the flow group ID selected in step 6010, and the correlation coefficient of the correlation information table 124 Stored in 502.
- the correlation coefficient is calculated by the following processes (F1) to (F4).
- the correlation calculation unit 112 extracts records corresponding to the flow ID 500 and the flow ID 501 from the post-discretization communication amount table 127. For example, if the traffic volume at time “i” of flow ID 500 “X” is “Xi” and the traffic volume at time “i” of flow ID “Y” is “Yi”, flow ID “X” and flow ID “ The correlation coefficient “r” with “Y” is calculated by the following equation (1).
- N (N is a positive integer) is the number of entries of the flow ID “X” (or flow ID “Y”) in the post-discretization traffic table 127. Since flows belonging to the same flow group are discretized with the same flow group discretization width, the number of entries of the flow IDs “X” and “Y” is the same “N”.
- the correlation calculation unit 112 stores the calculated correlation coefficient “r” in the correlation coefficient 502 corresponding to the flow ID “X” and the flow ID “Y” in the correlation information table 124.
- the correlation calculation unit 112 updates the correlation coefficient average value 504 and the correlation coefficient standard deviation 505 calculated in the past in the correlation information table 124 using the correlation coefficient “r” calculated this time. . Further, the correlation calculation unit 112 increments the correlation coefficient calculation count 503.
- the correlation calculation unit 112 executes the processes (F1) to (F3) for all combinations of flow IDs belonging to the flow group ID selected in step 6010.
- the correlation calculation unit 112 executes the processes (F1) to (F4) for all the flow group IDs.
- the number of correlation coefficient calculations is the number of combinations of flows belonging to the same flow group (product of the square of the number of flows belonging to the flow group and the number of flow groups).
- the number of calculations is less than the total number of flows (the number of flows squared). Therefore, according to the present embodiment, calculation resources and / or calculation time required for calculating the correlation coefficient can be reduced.
- the correlation calculation unit 112 calculates a difference between the correlation coefficient 502 and the correlation coefficient average 504 in the correlation information table 124, and identifies an entry in which the difference is larger than a predetermined threshold. Then, the correlation calculation unit 112 stores the flow ID 500 and the counter flow ID 501 of the identified entries in the flow ID 600 and the counter flow ID 601 of the abnormality information table 125. Because, when the correlation coefficient is far from the average correlation coefficient (when there is a significant change in the correlation coefficient), there is a possibility that the flow and / or pair flow related to the correlation coefficient is abnormal Because it is expensive.
- the predetermined threshold for the difference may be defined as a threshold based on the standard deviation of the correlation coefficient.
- Data network 10 Management system 21: Control network 30: Network device 50: Computer 100: Analysis system
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Environmental & Geological Engineering (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
- Debugging And Monitoring (AREA)
Abstract
Description
当該プロセッサは、
複数のデータフローを、データフローのデータ量の時系列変化の類似性に基づいて分類し、
同じ分類に属する少なくとも2つのデータフローの間について、通常時における相関係数と、或るタイミングにおける相関係数とを算出し、
通常時における相関係数と前記或るタイミングにおける相関係数との差分が所定の閾値よりも大きい場合、前記少なくとも2つのデータフローの内の少なくとも何れかが異常であると判定する。
フロー毎にフロー離散化幅を算出する理由は次の通りである。フローの通信量が非常に小さい場合、そのフローに対してサンプリングされるパケット数も少ない。したがって、その少数のサンプリングされたパケット数に基づいて上記(A3)のように通信量を算出するにあたり、サンプリングされるパケット数が少し増減するだけで、算出される通信量が大きく変動してしまう。この場合、検出される通信量の変動が、実際に通信量の増減によるものか(つまり有意な変動であるのか)、それとも、サンプリングされたパケット数がたまたま増減しただけなのか(つまり無意な変動であるのか)を判断することができない。
フローが異常か否かは、例えば次のように判定する方法が考えられる。すなわち、計測されたフローの全ての組み合わせについて、それぞれ、通常時(正常時)における通信量の時系列変化に係る相関係数を算出しておく。そして、全ての組み合わせについて相関係数を算出し、その算出した相関係数と通常時の相関係数との差分が所定よりも大きい場合、当該組み合わせに係るフローを異常と判定する。
Claims (10)
- データフローの異常を検出する異常検出装置であって、プロセッサ及びメモリを有し、
前記プロセッサは、
複数のデータフローを、データフローのデータ量の時系列変化の類似性に基づいて分類し、
同じ分類に属する少なくとも2つのデータフローの間について、通常時における相関係数と、或るタイミングにおける相関係数とを算出し、
前記通常時における相関係数と前記或るタイミングにおける相関係数との差分が所定の閾値よりも大きい場合、前記少なくとも2つのデータフローの内の少なくとも何れかが異常であると判定する
異常検出装置。 - 前記データフローとは、発信元から着信先へ通信ネットワークを介して流れるデータの流れである
請求項1に記載の異常検出装置。 - 前記プロセッサは、データ量の時系列変化の周波数成分の特性が類似するデータフローを、同じ分類に属させる
請求項2に記載の異常検出装置。 - 前記周波数成分の特性が類似するとは、所定の閾値以上の周波数成分を含む周波数帯域の少なくとも一部が重複することである
請求項3に記載の異常検出装置。 - データフローのデータ量の時系列変化に対して相関係数の算出対象の範囲として設定される対比時間は、同じ分類に属するデータフローにおいて共通である
請求項2に記載の異常検出装置。 - 前記対比時間は、前記同じ分類に属するデータフローのデータ量の時系列変化に対して共通に設定される離散化幅の倍数として算出される
請求項5に記載の異常検出装置。 - 前記共通に設定される離散化幅は、当該同じ分類に属するデータフロー毎にデータ量の時系列変化に基づいて算出した離散化幅のうち、最長の離散化幅である
請求項6に記載の異常検出装置。 - 前記プロセッサは、データフローが異常であると判定した場合、当該異常を検出したタイミングと、当該データフローの発信元及び着信先の情報とを通知し、当該タイミングにおいて発生した障害内容の入力を受け付ける
請求項1に記載の異常検出装置。 - データフローの異常を検出する異常検出システムであって、分析装置及びネットワーク装置を有し、
前記分析装置は、
ネットワーク装置から複数のデータフローのデータ量の時系列変化の情報を収集し、
それら収集した複数のデータフローを、データフローのデータ量の時系列変化の類似性に基づいて分類し、
同じ分類に属する少なくとも2つのデータフローの間について、通常時における相関係数と、或るタイミングにおける相関係数とを算出し、
前記通常時における相関係数と前記或るタイミングにおける相関係数との差分が所定の閾値よりも大きい場合、前記少なくとも2つのデータフローの内の少なくとも何れかが異常であると判定する
異常検出システム。 - データフローの異常を検出する計算機装置による異常検出方法であって、
複数のデータフローを、データフローのデータ量の時系列変化の類似性に基づいて分類し、
同じ分類に属する少なくとも2つのデータフローの間について、通常時における相関係数と、或るタイミングにおける相関係数とを算出し、
前記通常時における相関係数と前記或るタイミングにおける相関係数との差分が所定の閾値よりも大きい場合、前記少なくとも2つのデータフローの内の少なくとも何れかが異常であると判定する
異常検出方法。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2018506692A JP6535809B2 (ja) | 2016-03-24 | 2016-03-24 | 異常検出装置、異常検出システム、及び、異常検出方法 |
PCT/JP2016/059330 WO2017163352A1 (ja) | 2016-03-24 | 2016-03-24 | 異常検出装置、異常検出システム、及び、異常検出方法 |
US15/758,739 US10673721B2 (en) | 2016-03-24 | 2016-03-24 | Anomaly detection apparatus, anomaly detection system, and anomaly detection method using correlation coefficients |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/JP2016/059330 WO2017163352A1 (ja) | 2016-03-24 | 2016-03-24 | 異常検出装置、異常検出システム、及び、異常検出方法 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2017163352A1 true WO2017163352A1 (ja) | 2017-09-28 |
Family
ID=59900218
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2016/059330 WO2017163352A1 (ja) | 2016-03-24 | 2016-03-24 | 異常検出装置、異常検出システム、及び、異常検出方法 |
Country Status (3)
Country | Link |
---|---|
US (1) | US10673721B2 (ja) |
JP (1) | JP6535809B2 (ja) |
WO (1) | WO2017163352A1 (ja) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108362957A (zh) * | 2017-12-19 | 2018-08-03 | 东软集团股份有限公司 | 设备故障诊断方法、装置、储存介质和电子设备 |
JP2019152656A (ja) * | 2018-02-28 | 2019-09-12 | 株式会社デンソー | 電池監視システム |
JP2019152657A (ja) * | 2018-02-28 | 2019-09-12 | 株式会社デンソー | 電池監視システム |
JP2020501476A (ja) * | 2016-12-29 | 2020-01-16 | エヌエスフォーカス インフォメーション テクノロジー カンパニー,リミテッドNsfocus Information Technology Co.,Ltd | ネットワークにおけるトラフィックの異常を検出するための方法および装置 |
CN111817923A (zh) * | 2020-07-28 | 2020-10-23 | 城云科技(中国)有限公司 | 交换机端口流量突变的预警分析方法和装置 |
CN112445842A (zh) * | 2020-11-20 | 2021-03-05 | 北京思特奇信息技术股份有限公司 | 一种基于时间序列数据的异常值检测方法和系统 |
WO2023079757A1 (ja) * | 2021-11-08 | 2023-05-11 | 日本電信電話株式会社 | 分析装置、分析方法及び分析プログラム |
Families Citing this family (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10742672B2 (en) * | 2017-04-03 | 2020-08-11 | Level 3 Communication, Llc | Comparing metrics from different data flows to detect flaws in network data collection for anomaly detection |
US11592488B2 (en) * | 2018-02-28 | 2023-02-28 | Denso Corporation | Battery monitoring system |
US10742567B2 (en) * | 2018-12-13 | 2020-08-11 | Accenture Global Solutions Limited | Prescriptive analytics based storage class placement stack for cloud computing |
EP3989492B1 (en) * | 2019-07-23 | 2024-01-03 | Nippon Telegraph And Telephone Corporation | Abnormality detection device, abnormality detection method, and abnormality detection program |
US11651249B2 (en) | 2019-10-22 | 2023-05-16 | EMC IP Holding Company LLC | Determining similarity between time series using machine learning techniques |
US11175838B2 (en) | 2019-10-22 | 2021-11-16 | EMC IP Holding Company LLC | Automatic identification of resources in contention in storage systems using machine learning techniques |
US11175829B2 (en) | 2019-10-22 | 2021-11-16 | EMC IP Holding Company LLC | Automatic identification of workloads contributing to behavioral changes in storage systems using machine learning techniques |
US11062173B2 (en) | 2019-10-22 | 2021-07-13 | EMC IP Holding Company LLC | Automatic identification of workloads contributing to system performance degradation using machine learning techniques |
US11995562B2 (en) | 2020-12-03 | 2024-05-28 | International Business Machines Corporation | Integrating documentation knowledge with log mining for system diagnosis |
US11599404B2 (en) | 2020-12-03 | 2023-03-07 | International Business Machines Corporation | Correlation-based multi-source problem diagnosis |
US11474892B2 (en) | 2020-12-03 | 2022-10-18 | International Business Machines Corporation | Graph-based log sequence anomaly detection and problem diagnosis |
US11403326B2 (en) | 2020-12-03 | 2022-08-02 | International Business Machines Corporation | Message-based event grouping for a computing operation |
US11797538B2 (en) | 2020-12-03 | 2023-10-24 | International Business Machines Corporation | Message correlation extraction for mainframe operation |
US11243835B1 (en) | 2020-12-03 | 2022-02-08 | International Business Machines Corporation | Message-based problem diagnosis and root cause analysis |
US11513930B2 (en) | 2020-12-03 | 2022-11-29 | International Business Machines Corporation | Log-based status modeling and problem diagnosis for distributed applications |
CN114679409A (zh) * | 2020-12-09 | 2022-06-28 | 华为技术有限公司 | 一种流表发送方法及相关装置 |
CN112752113B (zh) * | 2020-12-28 | 2022-08-02 | 上海哔哩哔哩科技有限公司 | 直播服务器异常因素的确定方法及装置 |
US11625237B2 (en) | 2021-06-03 | 2023-04-11 | International Business Machines Corporation | Autonomous contextual software support anomaly detection and playbook automation |
US11895005B1 (en) * | 2022-12-02 | 2024-02-06 | Arista Networks, Inc. | Network devices with hardware accelerated table updates |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH11177549A (ja) * | 1997-12-09 | 1999-07-02 | Fujitsu Ltd | トラフィック監視装置及びトラフィック監視方法 |
JP2006115129A (ja) * | 2004-10-13 | 2006-04-27 | Tohoku Univ | ネットワーク異常検出システム |
JP4112584B2 (ja) * | 2005-12-19 | 2008-07-02 | 日本電信電話株式会社 | 異常トラヒック検出方法及び装置 |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110213869A1 (en) * | 2000-09-25 | 2011-09-01 | Yevgeny Korsunsky | Processing data flows with a data flow processor |
JP4872945B2 (ja) | 2008-02-25 | 2012-02-08 | 日本電気株式会社 | 運用管理装置、運用管理システム、情報処理方法、及び運用管理プログラム |
-
2016
- 2016-03-24 JP JP2018506692A patent/JP6535809B2/ja active Active
- 2016-03-24 WO PCT/JP2016/059330 patent/WO2017163352A1/ja active Application Filing
- 2016-03-24 US US15/758,739 patent/US10673721B2/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH11177549A (ja) * | 1997-12-09 | 1999-07-02 | Fujitsu Ltd | トラフィック監視装置及びトラフィック監視方法 |
JP2006115129A (ja) * | 2004-10-13 | 2006-04-27 | Tohoku Univ | ネットワーク異常検出システム |
JP4112584B2 (ja) * | 2005-12-19 | 2008-07-02 | 日本電信電話株式会社 | 異常トラヒック検出方法及び装置 |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2020501476A (ja) * | 2016-12-29 | 2020-01-16 | エヌエスフォーカス インフォメーション テクノロジー カンパニー,リミテッドNsfocus Information Technology Co.,Ltd | ネットワークにおけるトラフィックの異常を検出するための方法および装置 |
CN108362957A (zh) * | 2017-12-19 | 2018-08-03 | 东软集团股份有限公司 | 设备故障诊断方法、装置、储存介质和电子设备 |
CN108362957B (zh) * | 2017-12-19 | 2020-09-11 | 东软集团股份有限公司 | 设备故障诊断方法、装置、储存介质和电子设备 |
JP2019152656A (ja) * | 2018-02-28 | 2019-09-12 | 株式会社デンソー | 電池監視システム |
JP2019152657A (ja) * | 2018-02-28 | 2019-09-12 | 株式会社デンソー | 電池監視システム |
JP7225896B2 (ja) | 2018-02-28 | 2023-02-21 | 株式会社デンソー | 電池監視システム |
JP7225897B2 (ja) | 2018-02-28 | 2023-02-21 | 株式会社デンソー | 電池監視システム |
CN111817923A (zh) * | 2020-07-28 | 2020-10-23 | 城云科技(中国)有限公司 | 交换机端口流量突变的预警分析方法和装置 |
CN111817923B (zh) * | 2020-07-28 | 2021-09-14 | 城云科技(中国)有限公司 | 交换机端口流量突变的预警分析方法和装置 |
CN112445842A (zh) * | 2020-11-20 | 2021-03-05 | 北京思特奇信息技术股份有限公司 | 一种基于时间序列数据的异常值检测方法和系统 |
WO2023079757A1 (ja) * | 2021-11-08 | 2023-05-11 | 日本電信電話株式会社 | 分析装置、分析方法及び分析プログラム |
Also Published As
Publication number | Publication date |
---|---|
JP6535809B2 (ja) | 2019-06-26 |
US20180270132A1 (en) | 2018-09-20 |
JPWO2017163352A1 (ja) | 2018-10-18 |
US10673721B2 (en) | 2020-06-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2017163352A1 (ja) | 異常検出装置、異常検出システム、及び、異常検出方法 | |
US11528283B2 (en) | System for monitoring and managing datacenters | |
US9386028B2 (en) | System and method for malware detection using multidimensional feature clustering | |
US10027694B1 (en) | Detecting denial of service attacks on communication networks | |
US9787556B2 (en) | Apparatus, system, and method for enhanced monitoring, searching, and visualization of network data | |
US20190007292A1 (en) | Apparatus and method for monitoring network performance of virtualized resources | |
KR20150105436A (ko) | 향상된 스트리밍 방법과 네트워크 메타데이터를 처리하기 위한 시스템 | |
US20190007285A1 (en) | Apparatus and Method for Defining Baseline Network Behavior and Producing Analytics and Alerts Therefrom | |
JP2015173406A (ja) | 分析システム、分析装置、及び分析プログラム | |
AU2016393663A1 (en) | Method and system for compression and optimization of in-line and in-transit information security data | |
US11451443B2 (en) | Communications network performance | |
JP6317685B2 (ja) | 通信監視システム、通信監視方法およびプログラム | |
US20230261957A1 (en) | Framework for Anomaly Detection with Dynamic Model Selection | |
EP3092737B1 (en) | Systems for enhanced monitoring, searching, and visualization of network data | |
Johnson et al. | Network anomaly detection using autonomous system flow aggregates | |
JP2019213029A (ja) | 感染拡大攻撃検知システム及び方法、並びに、プログラム | |
WO2015105684A1 (en) | Apparatus, system, and method for enhanced monitoring and interception of network data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 15758739 Country of ref document: US |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2018506692 Country of ref document: JP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 16895388 Country of ref document: EP Kind code of ref document: A1 |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 16895388 Country of ref document: EP Kind code of ref document: A1 |