Extraction method, extraction device, and extraction program

The extraction method addresses the lack of IOC priority determination by analyzing analyst behavior history to prioritize IOC investigations, enhancing efficiency and reducing analyst workload through automated high-priority IOC analysis.

JP7578191B2Active Publication Date: 2024-11-06NIPPON TELEGRAPH & TELEPHONE CORP
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
JP2023520662
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Filing Date
2021-05-12
Publication Date
2024-11-06
Estimated Expiration
2041-05-12

AI Technical Summary

Technical Problem

Conventional techniques fail to provide characteristic information necessary for determining the priority of Indicator of Compromise (IOC) investigation, leading to inefficiencies in analyst workload and alert fatigue.

Method used

An extraction method that includes acquiring a history of analyst behavior and creating characteristic information based on this behavior history to prioritize IOC investigation, using a determination device with a feature information extraction unit, labeling unit, learning unit, and prediction unit to predict IOC priority.

Benefits of technology

Enables effective prioritization of IOC investigations, reducing analyst workload and preventing repetitive manual evaluations by automating the analysis of high-priority IOCs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007578191000006
    Figure 0007578191000006
  • Figure 0007578191000007
    Figure 0007578191000007
  • Figure 0007578191000008
    Figure 0007578191000008
Patent Text Reader

Abstract

A feature information extraction unit (21) acquires a history of behavior performed by an analysis person, regarding an investigation of IOC which is included in information relating to cyber security. The feature information extraction unit (21) creates feature information of the IOC on the basis of the information obtained from the acquired history of behavior.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to an extraction method, an extraction device, and an extraction program. [Background technology]

[0002] To ensure cybersecurity, companies and organizations have introduced systems for security management and threat detection. A Security Operation Center (SOC) is an organization that operates such systems. SOC analysts monitor and analyze the large volumes of logs and alerts output from the systems, and take any necessary action.

[0003] On the other hand, according to references 1 and 2 below, analysts who process the large number of alerts generated every day can develop a condition known as alert fatigue, which can lead to analyst burnout. Reference 1: SC Sundaramurthy, AG Bardas, J. Case, X. Ou, M. Wesch, J. McHugh, and SR Rajagopalan, “A human capital model for mitigating security analyst burnout,” Proc. SOUPS, 2015. Reference 2: Ponemon Institute, “Improving the Effectiveness of the Security Operations Center,” 2019.

[0004] To solve the above problems, it is necessary to achieve better automation and reduce the workload of analysts. In fact, according to Reference 3, many SOC managers consider the insufficient level of automation of SOC components to be the most important issue in current SOC organizations. Reference 3: FB Kokulu, A. Soneji, T. Bao, Y. Shoshitaishvili, Z. Zhao, A. Doupe, and G.-J. Ahn, “Matched and Mismatched SOCs: A Qualitative Study on Security Operations Center Issues,” Proc. ACM CCS, 2019.

[0005] In response to this, a technology has been proposed that distinguishes between truly malicious alerts and non-malicious alerts that are false positives by estimating the anomaly score and maliciousness score of each security alert from past alerts (for example, see Non-Patent Documents 1 to 5).

[0006] In addition, there is known a technique that supports the analyst's subsequent process by extracting the most relevant information for each security alert (see, for example, Non-Patent Documents 6 to 8). [Prior art documents] [Non-patent literature]

[0007] [Non-Patent Document 1] WU Hassan, S. Guo, D. Li, Z. Chen, K. Jee, Z. Li, and A. Bates, “NoDoze: Combatting Threat Alert Fatigue with Automated Provenance Triage,” Proc. NDSS, 2019. [Non-Patent Document 2] A. Oprea, Z. Li, R. Norris, and K. Bowers, “MADE: Security Analytics for Enterprise Threat Detection,” Proc. ACSAC, 2018. [Non-Patent Document 3] KA Roundy, A. Tamersoy, M. Spertus, M. Hart, D. Kats, M. Dell'Amico, and R. Scott, “Smoke Detector: Cross-Product Intrusion Detection With Weak Indicators,” Proc. ACSAC, 2017.

Non-licensed Document 4

Non-licensed Document 5

Non-licensed Document 6

Non-licensed Document 7

Non-licensed Document 8

[0008] However, the conventional techniques have a problem in that they may not provide useful feature information for determining the priority of IOC investigations.

[0009] For example, the technology described in the above-mentioned prior art documents employs feature information necessary to determine whether an IOC is abnormal or malignant, but whether an IOC is abnormal or malignant is independent of whether the IOC requires detailed investigation by an analyst. [Means for solving the problem]

[0010] In order to solve the above-mentioned problems and achieve the objective, the extraction method is an extraction method executed by an extraction device, and is characterized in that it includes an acquisition process for acquiring a history of actions taken by an analyst in investigating IOCs (Indicators of Compromise) contained in information related to cybersecurity, and a creation process for creating characteristic information of the IOCs based on information obtained from the history of actions acquired by the acquisition process. Effect of the Invention

[0011] According to the present invention, it is possible to obtain characteristic information that is useful for determining the priority of an IOC investigation. [Brief description of the drawings]

[0012] [Figure 1] FIG. 1 is a diagram illustrating a security system. [Diagram 2]FIG. 2 is a diagram showing an example of an alert monitor screen. [Diagram 3] FIG. 3 is a diagram showing an example of an IOC checker screen. [Figure 4] FIG. 4 is a diagram illustrating an example of the configuration of the determination device according to the first embodiment. [Diagram 5] FIG. 5 is a diagram showing an example of a request period. [Figure 6] FIG. 6 is a flowchart showing the flow of the learning process. [Figure 7] FIG. 7 is a flowchart showing the flow of the process for extracting feature information. [Figure 8] FIG. 8 is a flowchart showing the flow of the prediction process. [Figure 9] FIG. 9 is a diagram illustrating an example of a computer that executes a determination program. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] Hereinafter, an embodiment of an extraction method, an extraction device, and an extraction program according to the present application will be described in detail with reference to the drawings. Note that the present invention is not limited to the embodiment described below. Note that in this embodiment, a determination device functions as an extraction device.

[0014] [Configuration of the first embodiment] First, a security system including a determination device according to the first embodiment will be described with reference to Fig. 1. Fig. 1 is a diagram for explaining the security system.

[0015] In the security system 1, an analysis engine performs automatic analysis or a human analyst performs analysis based on predetermined information generated in the security priorities of a client organization.

[0016] Security priorities include, for example, an Intrusion Prevention System (IPS), a proxy, a sandbox, and a Unified Threat Management (UTM).

[0017] The SOC analyzes security information obtained from security appliances in real time, including security logs and alerts.

[0018] 1, the SOC is used as an outsourced SOC provided by a large-scale Managed Security Service Provider (MSSP). However, this embodiment can also be applied to an in-house SOC.

[0019] Although there are organizational differences between outsourced SOCs and in-house SOCs, the overall workflow is similar. For this reason, the effects of this embodiment are likely to be achieved if the in-house SOC is of an organization large enough to fully utilize economies of scale.

[0020] The following describes the flow of processing in the security system 1. As shown in Fig. 1, first, the security appliance of the client organization transmits an alert and a security log to the analysis engine 10 of the SOC (step S1).

[0021] An example of how to process an alert will be described below. The security system 1 can also process a security log in the same way as it processes an alert.

[0022] The analysis engine 10 performs an automatic analysis (step S2). The analysis engine 10 responds to the alert by performing an analysis based on known malicious characteristics and / or predefined rules and blacklists.

[0023] The analysis engine 10 may perform analysis using a function called SOAR (Security Orchestration, Automation, and Response).

[0024] The analysis engine 10 transmits an alert that satisfies a predetermined condition to the decision device 20, the alert monitor 30, or the IOC checker 40 (step S3).

[0025] At this time, the alert monitor 30 displays information about the alert, as shown in Fig. 2. Fig. 2 is a diagram showing an example of the screen of the alert monitor.

[0026] For example, the alert monitor 30 displays the date of the event that caused the alert, the customer name, the device that sent the alert, the name of the alert, and an overview of the situation that triggered the alert.

[0027] Furthermore, as shown in Fig. 3, the IOC checker 40 displays information about an IOC (Indicator of Compromise) included in the alert. Fig. 3 is a diagram showing an example of the screen of the IOC checker.

[0028] For example, IOCs include domain names, IP addresses, URLs, file hash values, etc.

[0029] As shown in FIG. 3, for example, the IOC checker 40 displays the investigation status (Status) at the SOC, the most recent SOC decision regarding the malignancy of the IOC (SOC Last Decision), the latest threat intelligence result for the IOC (Detection in TI), and the like.

[0030] For example, an analyst may use tools dedicated to IOC evaluation, such as an alert monitor 30 and an IOC checker 40, to triage (evaluate) IOCs for alerts that could not be processed by the analysis engine 10.

[0031] SOC analysts process a large number of alerts in their daily SOC workflow, so the determining device 20 determines high-priority IOCs and notifies the analysts of them, which can prevent multiple analysts in the SOC from manually evaluating the same IOCs.

[0032] Furthermore, according to the determination device 20, it is possible to give priority to analyzing IOCs with high priorities, thereby improving the efficiency of the analyst's workload.

[0033] The determining device 20 predicts the priority of the IOC by learning the model or by using the model (step S4). Then, the determining device 20 determines an IOC with a high priority based on the prediction result and notifies the determined IOC (step S5).

[0034] For example, the determining device 20 notifies an analyst of the determined IOC via the IOC checker 40 .

[0035] The analyst performs an analysis based on the notified priority (step S6). The analyst may also search a threat intelligence service (e.g., VirusTotal (https: / / www.virustotal.com / )) during the analysis (step S7).

[0036] Some threat intelligence services provide a score on the level or severity of a threat, but by definition such a score does not necessarily determine an analyst's next action.

[0037] For example, an IOC associated with an attack that exploits a vulnerability that has already been patched may have a high maliciousness score, but does not pose an immediate threat to protecting a customer organization.

[0038] As such, alert analysis in a SOC is not simple, making it difficult to fully automate alert analysis and sometimes requiring judgment by analysts.

[0039] Therefore, the determination of high-priority IOCs by the determination device 20 can be said to be useful for securing time for analysts to make decisions and reducing the investigation workload of each IOC.

[0040] The analyst ultimately determines whether the alert being analyzed and the IOCs contained in the alert are malicious or non-malicious, and further determines whether a report to the customer is necessary. If a report to the customer is necessary, the analyst reports to the system administrator or other person in the client organization (step S8).

[0041] For example, once an analyst has completed evaluation of an IOC, the triggering conditions for an alert in analysis engine 10 may be changed based on the results.

[0042] For example, if an IOC is identified as obviously malicious during an analyst evaluation, that IOC can be used in the analysis engine 10 as a custom blacklist or custom signature.

[0043] In this case, logs containing the same IOCs can be automatically made available to other SOC customers, and if the evaluation identifies false positives or IOCs with a low threat level, the SIEM logic that triggers the alert can be modified to prevent the same false positive alert from occurring again, reducing analyst workload.

[0044] Hereinafter, the process in which the determining device 20 determines an IOC with a high priority will be described in detail together with the configuration of the determining device 20.

[0045] Fig. 4 is a diagram illustrating an example of the configuration of a determining device according to the first embodiment. As illustrated in Fig. 4, the determining device 20 includes a feature information extracting unit 21, a label assigning unit 22, a learning unit 23, a predicting unit 24, and model information 25.

[0046] The determination device 20 performs model learning processing using a machine learning technique and prediction processing using the learned model.

[0047] In the learning process, a feature information extraction unit 21, a label assignment unit 22, and a learning unit 23 are used. In the prediction process, a feature information extraction unit 21 and a prediction unit 24 are used.

[0048] The characteristic information extraction unit 21 extracts characteristic information from an IOC included in information related to cybersecurity. For example, the information related to cybersecurity is an alert obtained from the analysis engine 10.

[0049] The characteristic information extraction unit 21 extracts, from an IOC included in a past alert obtained from the analysis engine 10, information that characterizes the characteristics of the IOC (hereinafter, referred to as characteristic information).

[0050] The characteristic information may be a domain name, an IP address, a URL, a file hash value, etc., included in the IOC.

[0051] For example, the characteristic information extraction unit 21 extracts characteristic information from alerts that occurred within a certain number of days that has been determined in advance.

[0052] Here, a detailed description will be given of a method for extracting feature information by the feature information extraction unit 21. The feature information extraction unit 21 functions as an extraction device having an acquisition unit and a creation unit.

[0053] The acquisition unit acquires a history of actions taken by an analyst in relation to an investigation of an IOC contained in cybersecurity information. The creation unit creates IOC characteristic information based on information obtained from the history of actions acquired by the acquisition unit.

[0054] The characteristic information extraction unit 21 creates characteristic information based on the history of actions, such as which analyst in which shift investigated each IOC, when and how much, for the entire SOC.

[0055] For example, the characteristic information extraction unit 21 observes requests made by an analyst to investigate each IOC, and creates characteristic information from the observation information.

[0056] As shown in step S7 of FIG. 1, in the SOC workflow, an analyst may send a request to a threat intelligence service to search for IOCs contained in an alert from a client organization.

[0057] Therefore, the characteristic information extraction unit 21 can obtain information on requests to the threat intelligence service as the behavioral history of the analyst.

[0058] The characteristic information extraction unit 21 does not need to acquire the detailed behavioral history of each analyst individually. Also, the characteristic information extraction unit 21 can acquire the behavioral history without changing the daily SOC workflow. Also, in most SOCs, the behavioral history described above is easily available.

[0059] In this embodiment, the feature information extracting unit 21 extracts a total of 80 pieces of feature information consisting of three major items and eight minor items.

[0060] The feature information extraction unit 21 extracts feature information using five different time windows (for example, 1 day, 3 days, 7 days, 14 days, and 30 days).

[0061] By using multiple different time windows in this manner, the characteristic information extraction unit 21 can distinguish between IOCs contained in alerts observed in bursts over a short period of time and IOCs of alerts observed over a longer period of time.

[0062] In addition, the characteristic information extraction unit 21 can further divide the most recent week into 1-day, 3-day, and 7-day periods to obtain characteristic information suitable for predicting priority in real time with an emphasis on the most recent information.

[0063] The feature information of each item will be explained. First, the feature information of item 1 is feature information based on the timing of the analyst's request. For example, item 1 includes 55 pieces of feature information. Hereinafter, feature information of an item expressed as item XY is considered to be feature information included in item X.

[0064] The characteristic information extraction unit 21 creates characteristic information included in item 1 based on information regarding the number of actions and the time intervals at which the actions were performed. For example, the number of actions and the time intervals at which the actions were performed are the number of requests to a threat intelligence service and the request intervals.

[0065] (Item 1-1) The characteristic information extraction unit 21 counts the number of request queries to the threat intelligence service used by the SOC for each of five time windows (e.g., 1 day, 3 days, 7 days, 14 days, and 30 days) as characteristic information. In this way, the characteristic information extraction unit 21 obtains, for example, five pieces of characteristic information.

[0066] The reason for using the characteristic information in item 1-1 is that the IOCs of suspicious candidates that SOC analysts manually investigated each have different characteristics.

[0067] For example, there may be IOCs included in alerts that are observed simultaneously by many customer organizations in the short term, and conversely, there may be IOCs that are observed by multiple customer organizations over the long term.

[0068] (Item 1-2) The characteristic information extraction unit 21 calculates statistics such as the average, minimum, maximum, standard deviation, and variance of the number of requests in the item 1-1 as characteristic information. In this way, the characteristic information extraction unit 21 obtains, for example, 25 pieces of characteristic information.

[0069] In this manner, the feature information extraction unit 21 creates feature information based on information obtained from the behavior history and statistics calculated from the information.

[0070] The characteristic information of item 1-2 makes it possible to grasp the tendency of how multiple analysts at the SOC base investigated the IOC during each time window.

[0071] (Item 1-3) The characteristic information extraction unit 21 calculates statistics such as the average, minimum, maximum, standard deviation, and variance of the time interval between requests as characteristic information, thereby obtaining, for example, 25 pieces of characteristic information.

[0072] The characteristic information in items 1-3 helps distinguish between IOC investigation requests used in targeted attacks against a specific customer company and IOC investigation requests used in non-targeted attacks that are conducted indiscriminately against multiple companies.

[0073] For example, IOCs used in targeted attacks reach only a few companies and their employees, and as a result are relatively rarely investigated by SOC analysts, whereas IOCs used in non-targeted attacks are widely distributed regardless of company or employee, and therefore are investigated by multiple analysts at the same SOC in a short period of time.

[0074] The feature information of item 2 is feature information based on the period of the analyst's request. The feature information included in item 2 is, for example, 15 pieces of feature information.

[0075] The feature information extraction unit 21 creates feature information contained in item 2 based on information about the time that has elapsed since the time point when an action was performed within a predetermined time window. Each period used in item 2 is as shown in Fig. 5. Fig. 5 is a diagram showing an example of a period of a request.

[0076] (Item 2-1) The characteristic information extraction unit 21 calculates the number of days elapsed from the first investigation date for each time window as characteristic information, and thereby obtains, for example, five pieces of characteristic information.

[0077] The characteristic information in item 2-1 makes it possible to distinguish IOCs that analysts have been investigating for an early period from those that have been investigated more recently.

[0078] (Item 2-2) The characteristic information extraction unit 21 calculates the number of days elapsed since the last investigation date for each time window as characteristic information, and thereby obtains, for example, five pieces of characteristic information.

[0079] The characteristic information in item 2-2 makes it possible to distinguish whether an IOC has been continuously investigated by analysts until recently.

[0080] (Item 2-3) The characteristic information extraction unit 21 calculates the number of days elapsed from the first investigation date to the last investigation date by the analyst for each time window as characteristic information. In this way, the characteristic information extraction unit 21 obtains, for example, five pieces of characteristic information.

[0081] The characteristic information in item 2-3 makes it possible to distinguish IOCs that have been investigated by analysts for a long period of time from those that have only been investigated for a short period of time.

[0082] The feature information of item 3 is feature information based on the analyst's shift. The feature information included in item 3 is, for example, 10 pieces of feature information.

[0083] The characteristic information extraction unit 21 creates characteristic information for item 3 based on the date and time when the action was performed and information related to the analyst's work style.

[0084] (Item 3-1) The characteristic information extraction unit 21 calculates the percentage of days on which analysts made requests to the threat intelligence service for each time window that were weekdays as characteristic information. For example, a weekday is Monday to Friday in local time at the location where the SOC is based. In this way, the characteristic information extraction unit 21 obtains, for example, five pieces of characteristic information.

[0085] On weekdays, SOC analysts often investigate IOCs contained in alerts recently observed from client organizations. On the other hand, on weekends, analysts tend to investigate IOCs that are different from usual, or to conduct active investigations of threats that cannot be done on weekdays because the absolute number of alerts is smaller. The characteristic information in item 3-1 is characteristic information that takes such tendencies into account.

[0086] (Item 3-2) The characteristic information extraction unit 21 calculates, for each time window, the percentage of the time when the analyst made a request to the threat intelligence service during the day shift as characteristic information. For example, an analyst's shift may include a day shift (e.g., 8:00 to 16:00) and a night shift (e.g., 16:00 to 8:00 the next day) to respond 24 hours a day, 365 days a year. As a result, the characteristic information extraction unit 21 obtains, for example, five pieces of characteristic information.

[0087] Similar to the relationship between weekdays and weekends mentioned above, analysts may have different survey trends for day shifts and night shifts. The feature information in item 3-2 is feature information that takes such trends into account.

[0088] The label assignment unit 22 assigns a label to each IOC according to the actual amount of work required to deal with the associated alert.

[0089] Here, the label is assumed to be binary data indicating whether the priority is high or low. For example, the labeling unit 22 assigns a label indicating high priority to an IOC that has consumed a lot of analyst work in the past, and assigns a label indicating low priority to other IOCs.

[0090] In the conventional techniques (for example, the techniques described in Non-Patent Documents 4 to 8), a label is attached to an IOC to indicate whether it is malignant (or malicious). On the other hand, in the present embodiment, a label is attached based on the amount of work of an analyst.

[0091] The label assignment unit 22 assigns a label indicating high priority to an IOC for which the number of manual investigations that have occurred within a certain period of time for related alerts is equal to or greater than a predetermined value, and assigns a label indicating low priority to an IOC for which the number of manual investigations is less than the predetermined value.

[0092] In the following description, a label indicating high priority will be denoted as "priority", and a label indicating low priority will be denoted as "non-priority".

[0093] The learning unit 23 uses learning data that combines the feature information extracted by the feature information extraction unit 21 and the labels assigned by the label assignment unit 22 to learn a model that outputs labels from the feature information of the IOCs.

[0094] The learning unit 23 creates and updates a model by supervised machine learning. The model information 25 is information including parameters for constructing a model. The learning unit 23 creates and updates the model information 25.

[0095] The learning unit 23 can employ any known supervised machine learning algorithm. In this embodiment, the learning unit 23 employs standard logistic regression.

[0096] Logistic regression is scalable and fast, making it well suited for predicting IOCs in large volumes of alerts from many customers, such as in a SOC environment.

[0097] In addition, logistic regression is known to be highly interpretable. The output of logistic regression can be interpreted as the probability that the input IOC is preferred, and can also show which features of the feature information corresponding to each IOC contribute to the result. In this way, logistic regression has the advantage of being highly interpretable.

[0098] Here, it is assumed that the learning unit 23 particularly uses logistic regression with L1 regularization.

[0099] First, when a vector x representing the feature information extracted by the feature information extraction unit 21 is given, the learning unit 23 models the conditional probability y of the label shown in equation (1) as in equation (2).

[0100]

number

[0101]

number

[0102] where θ is the parameter of the logistic regression model, σ is the sigmoid function, and all features of x are normalized to the range [0,1].

[0103] The learning unit 23 uses a set of n labeled learning data pieces shown in equation (3) to find a parameter θ for minimizing the objective function of equation (4) that introduces a hyperparameter λ that determines the degree of regularization.

[0104]

number

[0105]

number

[0106] In equation (4), the L1 regularization part λ||θ||1 adds a penalty to the objective function, and has the effect of identifying and reducing feature information that does not contribute significantly.

[0107] This reduction in features not only helps prevent overfitting, which is the process of fitting more than necessary to the training data, but also reduces memory usage and makes the results presented to SOC analysts more concise and easier to interpret.

[0108] The prediction unit 24 predicts a label from the feature information of the IOC using the model trained by the training unit 23.

[0109] The prediction unit 24 utilizes the model learned by the learning unit 23 to input feature information corresponding to IOCs contained in new alerts that occur in real time, and predicts which IOCs will consume a lot of analyst work in the future.

[0110] For example, the prediction unit 24 makes a prediction using a logistic regression model constructed based on the model information 25.

[0111] For example, the prediction unit 24 may predict the probability that an analyst will manually analyze a target IOC K or more times within P days (where P is an integer).

[0112] The prediction unit 24 uses the parameter θ determined by the learning unit 23 to calculate the probability p that the vector x of the feature information corresponding to the IOC is “priority”, and defines the predicted label ^y (^ directly above y) using equation (5).

[0113]

number

[0114] Based on the labels predicted by the prediction unit 24, the determination device 20 outputs IOCs that are likely to lead to repeated investigation by the SOC analyst, i.e., IOCs predicted to have “priority” labels, in order of highest probability p, and presents them to the analyst.

[0115] At this time, the analyst can use the information presented by the decision device 20 to prioritize the investigation targets and efficiently perform triage and detailed analysis.

[0116] SOC analysts are expected to determine and document, to the extent possible, what actions should be taken against the IOCs.

[0117] According to this embodiment, an analyst can investigate IOCs with high priority and reflect the results in the analysis engine 10. This enables the analysis engine 10 to automatically process alerts containing the same IOC, thereby preventing analysts from manually investigating the IOCs every time and reducing the workload of the SOC as a whole.

[0118] For example, an analyst can investigate an IOC that has been determined to have a high priority, and based on the results, have the analysis engine 10 automatically analyze the IOC. This prevents the IOC from being handed over to other analysts, thereby reducing the workload.

[0119] The determination device 20 periodically re-executes the learning process offline (for example, once a day) to update the model information 25. The determination device 20 performs the learning process using data for a predetermined period around the time point at which the feature information is extracted as shown in Fig. 5. For example, the determination device 20 performs the learning process using data for F+L days, which is a combination of F days (corresponding to a time window) up to the time point at which the feature information is extracted and L days from the time point at which the feature information is extracted (where F and L are integers).

[0120] On the other hand, when the determining device 20 processes an IOC included in an alert from a client organization in real time, that is, when performing a prediction process, it extracts feature information for the IOC using data for the past F days.

[0121] Then, the determining device 20 calculates, from the extracted characteristic information, a probability p that the analyst will perform K or more manual inspections within P days in the future.

[0122] The determination device 20 repeats the above prediction process for each IOC received in real time. As a result, a list of IOCs that should be investigated by the analyst with priority is displayed on the screen of the IOC checker 40 as shown in Figure 3 and is continuously updated.

[0123] [Processing of the first embodiment] 6 is a flowchart showing the flow of the learning process. As shown in FIG 6, first, the determining device 20 accepts an input of a past alert (step S101).

[0124] Next, the determining device 20 extracts feature information from the IOC included in the input alert (step S102). Next, the determining device 20 assigns a correct answer label regarding the priority based on the amount of work of the analyst for each IOC (step S103).

[0125] Then, the determining device 20 uses the correct label to learn a model that outputs a label related to the priority from the feature information (step S104).

[0126] 7 is a flowchart showing the flow of processing for extracting feature information, which corresponds to step S102 in FIG.

[0127] First, as shown in FIG. 7, the determining device 20 acquires the behavior history of the analyst (step S102a).

[0128] Next, the determination device 20 creates characteristic information based on the timing of the analyst's request (item 1) (step S102b). Then, the determination device 20 creates characteristic information based on the period of the analyst's request (item 2) (step S102c). Furthermore, the determination device 20 creates characteristic information based on the analyst's shift (item 3) (step S102d).

[0129] Fig. 8 is a flowchart showing the flow of the prediction process. As shown in Fig. 8, first, the determining device 20 receives an input of the most recent alert (step S201).

[0130] Next, the determining device 20 extracts feature information from the IOCs included in the input alert (step S202). Then, the determining device 20 extracts a correct label based on the amount of work of the analyst for each IOC (step S203).

[0131] Then, the determining device 20 inputs the feature information into the trained model to predict a label related to the priority (step S204).

[0132] The decision unit 20 can notify a SOC analyst of high priority IOCs based on the predicted labels.

[0133] [Advantages of the first embodiment] As described above, the characteristic information extraction unit 21 acquires the history of actions taken by an analyst regarding the investigation of IOCs included in cybersecurity information. The characteristic information extraction unit 21 creates IOC characteristic information based on information obtained from the acquired history of actions.

[0134] This provides useful characteristic information for prioritizing IOC investigations.

[0135] The characteristic information extraction unit 21 creates characteristic information based on information regarding the number of times a behavior occurred and the time intervals at which the behavior occurred.

[0136] This allows the characteristic information extraction unit 21 to reflect the tendency of an analyst's IOC investigation in the characteristic information.

[0137] The feature information extraction unit 21 creates feature information based on information about the time that has elapsed since a behavior was performed within a predetermined time window.

[0138] This allows the characteristic information extraction unit 21 to reflect in the characteristic information whether the IOC has been investigated for a long period of time or only for a short period of time.

[0139] The characteristic information extraction unit 21 creates characteristic information based on the date and time when the action was performed and information related to the analyst's work style.

[0140] This allows the characteristic information extraction unit 21 to reflect the tendency of the survey content according to the analyst's work shift in the characteristic information.

[0141] The feature information extraction unit 21 creates feature information based on information obtained from the behavior history and statistics calculated from the information.

[0142] This allows the feature information extraction unit 21 to obtain more feature information from limited information.

[0143] [System configuration, etc.] In addition, each component of each device 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 or 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 any part by a CPU (Central Processing Unit) and a program analyzed and executed by the CPU, or can be realized as hardware by wired logic. Note that the program may be executed not only by the CPU but also by other processors such as a GPU.

[0144] Furthermore, among the processes described in this embodiment, 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.

[0145] [program] As an embodiment, the determination device 20 can be implemented by installing a determination program that executes the above-mentioned determination process as package software or online software on a desired computer. For example, the above-mentioned determination program can be executed by an information processing device, thereby making the information processing device function as the determination device 20. The information processing device here includes desktop or 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 slate terminals such as PDA (Personal Digital Assistant), etc.

[0146] The decision device 20 may also be implemented as a decision server device that provides a service related to the above decision process to a client, the client being a terminal device used by a user. For example, the decision server device is implemented as a server device that provides a decision service that receives a security alert as an input and outputs an IOC with a high priority. In this case, the decision server device may be implemented as a Web server or as a cloud that provides a service related to the above decision process by outsourcing.

[0147] 9 is a diagram showing an example of a computer that executes a decision 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.

[0148] 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.

[0149] 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, a program that defines each process of the determining device 20 is implemented as a program module 1093 in which a computer-executable code is written. The program module 1093 is stored, for example, in the hard disk drive 1090. For example, a program module 1093 for executing the same process as the functional configuration in the determining device 20 is stored in the hard disk drive 1090. The hard disk drive 1090 may be replaced by an SSD (Solid State Drive).

[0150] Furthermore, setting data used in the processing of the above-described embodiment is stored as program data 1094, for example, in 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 the processing of the above-described embodiment.

[0151] 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]

[0152] 1. Security System 10. Analysis Engine 20 Determination device 21 Feature information extraction unit 22 Labeling section 23 Learning Department 24 Prediction Department 25 Model Information 30 Alert Monitor 40 IOC Checker

Claims

1. A method of extraction performed by an extraction device, comprising: An acquisition step of acquiring a history of actions taken by an analyst in relation to an investigation of an indicator of compromise (IOC) included in cybersecurity information; a creating step of creating IOC feature information based on information obtained from the behavior history acquired by the acquiring step, the information being related to the number of times the behavior was performed and the time intervals at which the behavior was performed; An extraction method comprising the steps of:

2. An extraction method performed by an extraction device, comprising: An acquisition step of acquiring a history of actions taken by an analyst in relation to an investigation of an indicator of compromise (IOC) included in cybersecurity information; a creating step of creating IOC feature information based on information obtained from the history of the behavior acquired by the acquiring step, the information being related to the elapsed time from the time the behavior was performed within a predetermined time window; An extraction method comprising the steps of:

3. An extraction method performed by an extraction device, comprising: An acquisition step of acquiring a history of actions taken by an analyst in relation to an investigation of an indicator of compromise (IOC) included in cybersecurity information; a creating step of creating IOC feature information based on information obtained from the behavior history acquired in the acquiring step, the information being the date and time when the behavior was performed and information on the analyst's work style; An extraction method comprising the steps of:

4. 4. The extraction method according to claim 1, wherein the creating step creates the feature information based on information obtained from the behavior history and statistics calculated from the information.

5. An acquisition unit that acquires a history of actions taken by an analyst in relation to an investigation of an indicator of compromise (IOC) included in cybersecurity information; a creating unit that creates IOC feature information based on information obtained from the behavior history acquired by the acquiring unit, the information being related to the number of times the behavior occurred and the time intervals at which the behavior occurred; An extraction apparatus comprising:

6. An acquisition unit that acquires a history of actions taken by an analyst in relation to an investigation of an indicator of compromise (IOC) contained in cybersecurity information; a creation unit that creates IOC feature information based on information obtained from the behavior history acquired by the acquisition unit, the information being related to an elapsed time from a time point when the behavior was performed within a predetermined time window; and An extraction apparatus comprising:

7. An acquisition unit that acquires a history of actions taken by an analyst in relation to an investigation of an indicator of compromise (IOC) contained in cybersecurity information; a creation unit that creates IOC feature information based on information obtained from the behavior history acquired by the acquisition unit, the information being information on the date and time when the behavior was performed and the work style of the analyst; An extraction apparatus comprising:

8. An extraction program that causes a computer to function as an extraction device described in any one of claims 5 to 7.

Citation Information

Patent Citations

  • Method and apparatus for managing security in computer networks

    JP2018521430A

  • Network threat indicator extraction and response

    JP2020035424A

  • Email analysis system, method for controlling email analysis system, and control program for email analysis system

    WO2016117132A1

  • Analysis device, log analysis method, and recording medium

    WO2018235252A1