DGA domain name detection system and method based on behavior scenarios

Through the DGA domain name detection system based on behavioral scenarios, combined with whitelist automatic learning and multi-dimensional behavior correlation analysis, the problem of high false alarm rate in DGA detection is solved, and higher detection accuracy and efficiency are achieved.

CN115333766BActive Publication Date: 2024-11-19NAT IND INFORMATION SECURITY DEV RES CENT
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210730414.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-06-24
Publication Date
2024-11-19
Estimated Expiration
2042-06-24

AI Technical Summary

Technical Problem

When detecting domain names generated by DGA algorithms, the false alarm rate is high and it is impossible to effectively avoid the problem that normal domain names are falsely reported as malicious domain names, resulting in low detection accuracy and efficiency.

Method used

Adopting a DGA domain name detection system based on behavior scenarios, combining whitelist automatic learning and machine learning through data acquisition and analysis modules, DGA request behavior detection modules, DGA request process behavior monitoring modules, DGA request result behavior monitoring modules and behavior correlation analysis engines Algorithm, perform multi-dimensional behavior correlation analysis to reduce false positive rate.

Benefits of technology

It effectively reduces the false alarm rate, improves the accuracy and efficiency of DGA detection, reduces the workload of security analysts, and improves the reliability of detection results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115333766B_ABST
    Figure CN115333766B_ABST
Patent Text Reader

Abstract

The present application provides a DGA domain name detection system and method based on behavior scenarios, wherein the detection system includes a data collection and analysis module, a DGA request behavior detection module, a DGA request process behavior monitoring module, a DGA request result behavior monitoring module, an associated device behavior monitoring module, and a behavior association analysis engine; the data collection and analysis module is used to collect network traffic logs and DNS logs, and parse the collected logs; the DGA request behavior detection module is used to detect the domain name obtained by log parsing to determine the DGA malicious domain name; the behavior association analysis engine is used to conduct a comprehensive analysis based on the analysis results of the DGA request process behavior monitoring module and the DGA request result behavior monitoring module, so as to generate a DGA alarm based on the analysis results. The present application can reduce the false alarm rate, effectively reduce the workload of manual analysis by security analysts, and improve analysis efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application belongs to the field of information security, and specifically relates to a DGA domain name detection system and method based on behavior scenarios. Background Art

[0002] In the field of information security, DGA (Domain Generation Algorithm) detection has always been a technical problem. Due to the continuous improvement of the DGA algorithm, traditional feature-based detection methods are basically unable to accurately detect domain names generated by the DGA algorithm. However, the discovery of DGA is the most direct and effective evidence of a compromised host, so people have been continuously studying ways to improve the accuracy of DGA detection.

[0003] In recent years, with the development of AI technology, people have begun to use AI technology to learn and model the domain name rules generated by the DGA algorithm, which has greatly improved the detection rate of DGA. However, since the accuracy of the model trained by the AI ​​algorithm depends largely on the training data set, when the normal domain name (i.e., white sample) is not in the training data set, it is easy to generate a large number of false positives. Moreover, since the number of normal domain names is usually much larger than the number of malicious domain names (i.e., black samples), even if the false positive rate is relatively low, the actual number of false positives is still huge.

[0004] The current mainstream method is to optimize the DGA algorithm through various methods or use more complex algorithm models to enhance the accuracy of detection. However, the actual effect is very limited, because no matter how optimized the algorithm is, it cannot avoid the appearance of a large number of white sample data that have not appeared in the training data set in the actual domain name data, and the problem of inaccurate models and a large number of false positives caused by the amount of white sample data being much higher than the amount of black sample data. Other methods use threat intelligence filtering and clustering algorithm detection results to strengthen the evidence of malicious domain names. These methods still cannot fundamentally solve the problem of misreporting a large number of normal domain names as malicious domain names. Summary of the invention

[0005] In order to overcome the problems existing in the related art at least to a certain extent, the present application provides a DGA domain name detection system and method based on behavior scenarios.

[0006] According to a first aspect of an embodiment of the present application, the present application provides a DGA domain name detection system based on a behavior scenario, which includes a data collection and parsing module, a DGA request behavior detection module, a DGA request process behavior monitoring module, a DGA request result behavior monitoring module, an associated device behavior monitoring module, and a behavior association analysis engine;

[0007] The data collection and analysis module is used to collect network traffic logs and DNS logs, and analyze the collected logs;

[0008] The DGA request behavior detection module is used to detect the domain name obtained by log parsing to determine the DGA malicious domain name;

[0009] The DGA request process behavior monitoring module is used to determine whether the DGA request process behavior is abnormal, the DGA request result behavior monitoring module is used to determine whether the DGA request result behavior is abnormal, and the associated device behavior monitoring module is used to obtain relevant clues from the associated device when the local judgment clues are missing;

[0010] The behavior correlation analysis engine is used to perform comprehensive analysis based on the abnormal behavior analysis results of the DGA request process and the abnormal behavior analysis results of the DGA request result, so as to generate a DGA alarm based on the analysis results.

[0011] In the above-mentioned DGA domain name detection system based on behavior scenarios, the DGA request behavior detection module includes a domain name whitelist and a DGA model detection engine;

[0012] The domain name whitelist includes a learning module and a filtering module. The learning module is used to collect domain names that are accessed daily through DNS logs when the network is in a healthy state and save them as a whitelist; the filtering module is used to filter out domain names that exist on the whitelist in the DNS log to be detected according to the current whitelist;

[0013] The DGA model detection engine includes a model training module and a model reasoning module. The model training module obtains whitelist samples from a domain name whitelist, retrains the DGA detection model and deploys the trained DGA detection model to the DGA model detection engine; the model reasoning module is used to send the domain name filtered by the filtering module to the trained DGA detection model for analysis.

[0014] In the above-mentioned DGA domain name detection system based on behavior scenarios, the DGA request process behavior monitoring module includes an NXDomain baseline monitoring module and a Fast-Flux feature monitoring module;

[0015] The NXDomain baseline monitoring module determines whether the number of NXDomains returned by the DNS request exceeds the baseline range by checking the return result of the DNS request on the asset corresponding to the alarm reported by the DGA model detection engine;

[0016] The Fast-Flux feature monitoring module provides baseline monitoring based on multiple unfamiliar domain name requests within a preset time period and baseline monitoring of the same domain name requests returning different IPs.

[0017] In the above-mentioned DGA domain name detection system based on behavioral scenarios, the DGA request result behavior monitoring module includes a suspicious external connection monitoring module and a threat intelligence module; the suspicious external connection monitoring module is used to judge the geographical location, service port and time of the IP returned by the current asset connection DGA request, and analyze whether it is a suspicious behavior; the threat intelligence module is used to perform threat intelligence collision on the IP successfully returned by the DGA request to detect whether it is a known malicious IP.

[0018] According to a second aspect of the embodiment of the present application, the present application also provides a DGA domain name detection method based on a behavior scenario, which includes the following steps:

[0019] Detect DGA request behavior to obtain the request domain name;

[0020] Use machine learning algorithms to detect domain names to identify DGA malicious domain names;

[0021] Conduct behavioral correlation analysis on DGA malicious domain names to determine whether the DGA request process behavior and DGA request result behavior are abnormal;

[0022] A comprehensive analysis is performed based on the abnormal behavior analysis results of the DGA request process and the abnormal behavior analysis results of the DGA request result to generate a DGA alarm based on the analysis results.

[0023] In the above-mentioned DGA domain name detection method based on behavior scenarios, the specific process of detecting the DGA request behavior to obtain the requested domain name is:

[0024] When the network is in a healthy state, the domain names that are accessed daily are collected through DNS logs and saved as a whitelist.

[0025] Based on the current whitelist, the request domain name in the DNS log to be detected is filtered. If the request domain name exists in the whitelist, it is filtered out. Otherwise, the request domain name is detected using a machine learning algorithm.

[0026] Furthermore, the specific process of performing behavioral association analysis on DGA malicious domain names is as follows:

[0027] Conduct analysis and judgment on abnormal behavior during the DGA request process;

[0028] Conduct abnormal analysis on DGA request results.

[0029] Furthermore, the process of analyzing and judging the abnormal behavior of the DGA request process is as follows:

[0030] Set baseline detection conditions, including NXDomain baseline, unfamiliar domain name access baseline, and the baseline of the number of IP addresses returned by requests for the same domain name;

[0031] The NXDomain baseline anomaly detection is used to maintain the NXDomain number of each local device on the current day. The Gaussian distribution model of the NXDomain number of all devices in the network on the current day in the historical data is used as the baseline threshold. The NXDomain number of the local device corresponding to the current suspected DGA request on the current day is compared with the historical baseline threshold. If the NXDomain number of the local device corresponding to the current suspected DGA request on the current day exceeds the historical baseline threshold, the NXDomain number is determined to be abnormal.

[0032] The unfamiliar domain name access baseline anomaly detection is used to maintain the number of unfamiliar non-whitelist domain name accesses of each local device on the same day. The Gaussian distribution model of the number of unfamiliar domain name accesses of all devices in the network in one day in the historical data is used as the baseline threshold. The number of unfamiliar domain name accesses counted on the same day by the local device corresponding to the current suspected DGA request is compared with the historical baseline threshold. If the number of unfamiliar domain name accesses counted on the same day by the local device corresponding to the current suspected DGA request exceeds the historical baseline threshold, the number of unfamiliar domain name accesses is determined to be abnormal.

[0033] Preset an IP address number threshold, compare the number of IP addresses returned by the suspected DGA domain name on that day with the preset IP address number threshold, if the number of IP addresses returned by the suspected DGA domain name on that day exceeds the IP address number threshold, it is determined that an abnormal baseline of the number of IP addresses returned by the same domain name request has occurred;

[0034] If at least two baseline detection conditions are abnormal, the DGA request process behavior is determined to be abnormal.

[0035] Furthermore, the process of determining whether the DGA request result is abnormal is as follows:

[0036] Blacklists constructed using suspicious IP address source countries, suspicious common ports, and suspicious access times;

[0037] Determine whether the data related to the external connection of the local device using the IP address obtained by the suspected DGA request matches the data in the blacklist. If so, a behavior anomaly is generated. At least two behavior anomalies are determined to be an external connection behavior anomaly.

[0038] When local analysis clues are missing, relevant clues are obtained from associated devices to replace local analysis clues, including requesting other devices with the same domain name that is currently suspected of DGA to see whether the DGA request process behavior is abnormal, whether there is any external connection behavior abnormality, and if so, the behavior of the associated device can be used to replace the corresponding behavior of the local device.

[0039] Furthermore, the process of performing comprehensive judgment based on the abnormal judgment result of the DGA request process behavior and the abnormal judgment result of the DGA request result behavior to generate a DGA alarm based on the judgment result is:

[0040] If the IP address collision match of the threat intelligence is successful, a DGA high-risk alarm will be directly generated;

[0041] Determine whether the DGA request process behavior of the current device is abnormal or whether there is a suspicious external connection anomaly. If there is an undetected anomaly, determine whether the associated device behavior anomaly can replace the undetected anomaly. If so, replace it.

[0042] If the DGA request process behavior is abnormal and the conditions of suspicious external connection abnormality exist, a DGA high-risk alarm is generated;

[0043] If only the DGA request process behavior abnormality occurs, a DGA medium-risk alarm is generated;

[0044] If only DGA suspicious external connection abnormal behavior occurs, a DGA low-risk alarm is generated.

[0045] According to the above specific implementation methods of the present application, it can be known that at least the following beneficial effects are achieved: the DGA domain name detection system based on behavioral scenarios provided by the present application filters a large number of local normal domain name requests by means of automatic learning of whitelists, thereby reducing the normal sample data input to the DGA detection model for judgment and reducing false alarms; through behavioral scenario association analysis, a comprehensive judgment is made from multiple clues in multiple dimensions such as DGA request process behavior and DGA request result behavior, and the accuracy of the current algorithm analysis results is verified with relevant clues, which can achieve better false alarm elimination effects than simply optimizing the algorithm; by supplementing the local device behavior with the associated device behavior, the problem that some local devices cannot expose relevant clues due to the failure of malicious programs to be further executed due to operating environment problems can be effectively avoided; in the process of research and judgment, all clues before and after the DGA behavior occur are analyzed, which can effectively reduce the workload of manual analysis by security analysts and improve analysis efficiency.

[0046] It should be understood that the above general description and the following specific embodiments are only exemplary and illustrative and do not limit the scope of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The attached drawings below are part of the specification of the present application, which illustrate embodiments of the present application. The attached drawings together with the description of the specification are used to explain the principles of the present application.

[0048] Figure 1 A structural diagram of a DGA domain name detection system based on behavior scenarios provided for a specific implementation of the present application.

[0049] Figure 2 A flowchart of a DGA domain name detection method based on behavioral scenarios is provided for a specific implementation of the present application.

[0050] Description of Reference Numerals

[0051] 1. Data collection and analysis module;

[0052] 2. DGA request behavior detection module; 21. Domain name whitelist; 22. DGA model detection engine;

[0053] 3. DGA request process behavior monitoring module; 31. NXDomain baseline monitoring module; 32. Fast-Flux feature monitoring module;

[0054] 4. DGA request result behavior monitoring module; 41. Suspicious external connection monitoring module; 42. Threat intelligence module;

[0055] 5. Associated device behavior monitoring module; 6. Behavior association analysis engine. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clearly understood, the spirit of the contents disclosed in the present application will be clearly explained with the accompanying drawings and detailed descriptions below. After understanding the embodiments of the contents of the present application, any technician in the relevant technical field can change and modify them according to the techniques taught by the contents of the present application without departing from the spirit and scope of the contents of the present application.

[0057] The exemplary embodiments and descriptions of the present application are used to explain the present application, but are not intended to limit the present application. In addition, elements / components with the same or similar reference numerals used in the drawings and embodiments are used to represent the same or similar parts.

[0058] The terms “first”, “second”, etc. used in this document do not particularly refer to an order or sequence, nor are they used to limit this application. They are only used to distinguish elements or operations described with the same technical terms.

[0059] The words “include,” “including,” “have,” “contain,” etc. used in this document are open-ended terms, meaning including but not limited to.

[0060] As used herein, "and / or" includes any or all combinations of the items described.

[0061] As used herein, “plurality” includes “two” and “more than two”; as used herein, “plurality groups” includes “two groups” and “more than two groups”.

[0062] Certain terms used to describe the present application are discussed below or elsewhere in this specification to provide additional guidance to those skilled in the art regarding the description of the present application.

[0063] like Figure 1 As shown, the DGA domain name detection system based on behavioral scenarios provided in the embodiment of the present application includes a data collection and analysis module 1, a DGA request behavior detection module 2, a DGA request process behavior monitoring module 3, a DGA request result behavior monitoring module 4, an associated device behavior monitoring module 5 and a behavior association analysis engine 6.

[0064] The data collection and analysis module 1 is used to collect network traffic logs and DNS (Domain Name System) logs, and analyze the collected logs.

[0065] The DGA request behavior detection module 2 includes a domain name whitelist 21 and a DGA model detection engine 22 .

[0066] The domain name whitelist 21 includes a learning module and a filtering module. The learning module is used to collect domain names that are accessed daily through DNS logs when the network is in a healthy state, save these domain names as a whitelist, and execute periodically to ensure the timeliness of the whitelist.

[0067] The filtering module is used to filter out the domain names on the white list in the DNS log to be detected according to the current white list, so as to greatly reduce the white samples input into the DGA model detection engine 22.

[0068] The DGA model detection engine 22 includes a model training module and a model reasoning module. The model training module obtains whitelist samples from the domain name whitelist 21, retrains the DGA detection model and deploys the trained DGA detection model to the DGA model detection engine 22.

[0069] The model reasoning module is used to send the domain name filtered by the filtering module into the trained DGA detection model for analysis, and the generated alarm will be sent to the DGA request process behavior monitoring module 3 as the behavior chain judgment entry condition for further analysis.

[0070] The DGA request process behavior monitoring module 3 provides a statistical perspective for reviewing the behavior characteristics of the request and response process after multiple executions of the DGA algorithm over a longer period of time, which includes the NXDomain baseline monitoring module 31 and the Fast-Flux feature monitoring module 32.

[0071] Among them, the NXDomain baseline monitoring module 31 provides an audit method for determining whether there are multiple DNS requests for non-existent domain names from a statistical dimension, and determines whether the number of NXDomains returned by the DNS request exceeds the baseline range by checking the return results of the DNS request on the assets corresponding to the alarm reported by the DGA model detection engine 22.

[0072] The Fast-Flux feature monitoring module 32 provides a baseline monitoring based on multiple unfamiliar domain name requests within a preset time period and a baseline monitoring based on the same domain name request returning different IP addresses.

[0073] The DGA request result behavior monitoring module 4 is used to monitor the behavior after the DGA request successfully obtains the malicious host IP, and includes a suspicious external connection monitoring module 41 and a threat intelligence module 42. Among them, the suspicious external connection monitoring module 41 is used to judge the geographical location, service port and time of the IP returned by the current asset connection DGA request, and analyze whether it is a suspicious behavior. The threat intelligence module 42 is used to perform threat intelligence collision on the IP successfully returned by the DGA request to detect whether it is a known malicious IP.

[0074] The associated device behavior monitoring module 5 provides DGA behavior detection and judgment results of other hosts in the same network based on the same suspected DGA request domain name, which is used to reduce false alarms of several baseline monitoring and conditional judgments. For example, if other request nodes of a suspected DGA domain name do not generate NXDomain baseline violations and do not have Fast-Flux features, then it is possible that the current asset host is not executing the DGA algorithm, but a configuration error has occurred.

[0075] The behavior correlation analysis engine 6 provides the function of unified correlation analysis of the whole process behavior of the DGA algorithm execution and the behavior verification of the associated equipment, and provides conditional logic operations on the analysis results of multiple behaviors. The entry condition for executing the behavior correlation analysis is the output alarm of the DGA model detection engine 22. The DGA request process behavior monitoring module 3, the DGA request result behavior monitoring module 4 and the associated equipment behavior monitoring module 5 serve as positive or negative judgment conditions for this alarm to determine whether to issue an alarm or determine it as a false alarm.

[0076] The behavior scenario-based DGA domain name detection system provided in the embodiment of the present application performs multi-dimensional noise reduction, false alarm removal and evidence enhancement on the DGA detection results based on the entire process of attack behavior in which the DGA participates, which can effectively improve the accuracy of DGA detection and improve the efficiency of security analysis.

[0077] According to the DGA domain name detection system based on behavior scenarios provided by the present application, the present application also provides a DGA domain name detection method based on behavior scenarios.

[0078] like Figure 2 As shown, the DGA domain name detection method based on behavior scenarios provided in the embodiment of the present application includes the following steps:

[0079] S1. Detect DGA request behavior. The specific process is as follows:

[0080] S11. When the network is in a healthy state, the domain names that are accessed daily are collected through DNS logs and these domain names are saved as a whitelist.

[0081] S12. Based on the current whitelist, filter the request domain name in the DNS log to be detected. If the request domain name exists in the whitelist, filter it out without further processing; otherwise, go to step S2 and use a machine learning algorithm to detect the request domain name.

[0082] S2. Use a machine learning algorithm to detect the domain name. If the domain name is detected as a normal domain name, no further processing is performed; if the domain name is detected as a DGA malicious domain name, proceed to step S3 for behavior association analysis.

[0083] S3. Conduct behavioral correlation analysis on domain names. The specific process is as follows:

[0084] S31. Conducting analysis and judgment on abnormal behavior during the DGA request process;

[0085] Set baseline detection conditions, including the NXDomain baseline, unfamiliar domain name access baseline, and the baseline for the number of IP addresses returned by requests for the same domain name.

[0086] NXDomain baseline anomaly detection is used to maintain the NXDomain number of each local device on the day. The μ+3σ of the Gaussian distribution model of the NXDomain number of all devices in the network on the day in the historical data is used as the baseline threshold. The NXDomain number of the local device corresponding to the current suspected DGA request on the day is compared with the historical baseline threshold. If the NXDomain number of the local device corresponding to the current suspected DGA request on the day exceeds the historical baseline threshold, the NXDomain number is determined to be abnormal.

[0087] The unfamiliar domain name access baseline anomaly detection is used to maintain the number of unfamiliar non-whitelist domain name accesses of each local device on the same day. The μ+3σ of the Gaussian distribution model of the number of unfamiliar domain name accesses of all devices in the network in one day in the historical data is used as the baseline threshold. The number of unfamiliar domain name accesses counted on the same day by the local device corresponding to the current suspected DGA request is compared with the historical baseline threshold. If the number of unfamiliar domain name accesses counted on the same day by the local device corresponding to the current suspected DGA request exceeds the historical baseline threshold, the number of unfamiliar domain name accesses is determined to be abnormal.

[0088] A preset IP address number threshold is used to compare the number of IP addresses returned by the suspected DGA domain name on that day with the preset IP address number threshold. If the number of IP addresses returned by the suspected DGA domain name on that day exceeds the IP address number threshold, it is determined that an abnormal baseline of the number of IP addresses returned by the same domain name request has occurred.

[0089] After making the above judgment, if at least two baseline detection conditions are abnormal, it is determined that the DGA request process behavior is abnormal.

[0090] S32, conducting DGA request result behavior abnormality analysis;

[0091] A blacklist constructed using suspicious IP address source countries, suspicious common ports, and suspicious access times.

[0092] It is determined whether the data related to the external connection of the local device using the IP address obtained by the suspected DGA request matches the data in the blacklist. If so, a behavior anomaly is generated. At least two behavior anomalies are determined to be external connection behavior anomalies.

[0093] When local analysis clues are missing, relevant clues are obtained from associated devices to replace local analysis clues, including requesting other devices with the same domain name that is currently suspected of DGA to see whether the DGA request process behavior is abnormal, whether there is any external connection behavior abnormality, and if so, the behavior of the associated device can be used to replace the corresponding behavior of the local device.

[0094] S4. Comprehensively judge based on the abnormal judgment results of the DGA request process behavior and the abnormal judgment results of the DGA request result behavior. The specific process is as follows:

[0095] If the IP address collision match of the threat intelligence is successful, a DGA high-risk alarm will be directly generated.

[0096] Determine whether the DGA request process behavior of the current device is abnormal or whether there is a suspicious external connection anomaly. If there is an undetected anomaly, determine whether the associated device behavior anomaly can replace the undetected anomaly. If so, replace it.

[0097] If the DGA request process behaves abnormally and the conditions of suspicious external connection abnormality exist, a DGA high-risk alarm is generated.

[0098] If only the DGA request process behavior abnormality occurs, a DGA medium-risk alarm is generated.

[0099] If only DGA suspicious external connection abnormal behavior occurs, a DGA low-risk alarm is generated.

[0100] The DGA domain name detection method based on behavioral scenarios provided in the present application filters a large number of local normal domain name requests by means of automatic whitelist learning, thereby reducing the normal sample data input to the DGA detection model for judgment, and can reduce false positives.

[0101] The DGA domain name detection method based on behavioral scenarios provided in this application makes a comprehensive judgment from multiple dimensions and multiple clues such as DGA request process behavior and DGA request result behavior through behavioral scenario correlation analysis, and verifies the accuracy of the current algorithm analysis results with relevant clues, which can achieve better false alarm elimination effects than simply optimizing the algorithm.

[0102] The DGA domain name detection method based on behavioral scenarios provided in this application supplements the local device behavior by associating the device behavior, which can effectively avoid the problem that some local devices cannot expose relevant clues because malicious programs are not further executed due to operating environment problems.

[0103] The DGA domain name detection method based on behavioral scenarios provided in this application has analyzed all clues before and after the DGA behavior occurs during the research and judgment process, which can effectively reduce the workload of manual analysis by security analysts and improve analysis efficiency.

[0104] In order to more clearly understand the DGA domain name detection method for the behavior scenario provided by the present application, a specific embodiment is used for illustration below.

[0105] The implementation process of the DGA domain name detection method based on behavioral scenarios provided in this application includes an information collection and model training phase and a detection and analysis phase.

[0106] During the information collection and model training phase, the data collection and parsing module reads and parses one month of historical DNS logs stored in the database. These historical DNS logs are considered to be normal business access requests without DGA requests.

[0107] Generate a domain name whitelist library based on historical DNS logs and generate a white sample data set.

[0108] Based on the white sample data set and the DGA sample data set obtained from the Internet, a DGA detection model based on LSTM (Long Short-Term Memory) is trained, and the DGA detection model is deployed in the DGA model detection engine.

[0109] Based on the DNS response logs in the historical DNS logs, calculate the number of NXDomain response messages per local device IP address in one day, and calculate the mean and variance of the total number of NXDomain response messages.

[0110] Based on the DNS request logs in the historical DNS logs, the number of unfamiliar domain names requested by the local device IP in one day is calculated (if this domain name has only been requested once in the data sample set, it is determined to be an unfamiliar domain name), and the mean and variance of the total number of unfamiliar domain name request messages are calculated.

[0111] In the detection and analysis stage, assuming that the local device IP is IP1, the domain name of the DNS request initiated by device IP1 is Domain1, the DNS request obtains a DNS response and returns the IP corresponding to the domain name Domain1 as IP2. The DNS request and response log corresponding to the received device IP1 is obtained through the data collection and analysis module. Whitelist filtering is performed first. If the domain name is not on the whitelist, DGA model detection is performed.

[0112] The DGA detection model detects and analyzes Domain1 and determines it to be a DGA domain name. The behavior association analysis engine then performs further analysis and judgment.

[0113] The behavior correlation analysis engine starts the analysis scenario for Domain1, which is currently suspected of being a DGA. It first performs a threat intelligence collision analysis on IP3. If it returns that it is not found, it further performs other analyses in the scenario.

[0114] Perform NXDomain baseline anomaly detection and unfamiliar domain name access baseline anomaly detection on device IP1, perform baseline anomaly detection on the number of IPs returned when requesting the same domain name on Domain1, and judge the three detection results as abnormal DGA request process behavior if there are at least two abnormal results. Assume that all returns here are abnormal.

[0115] Assuming that device IP1 has no external connection to IP2, the suspicious external connection is determined to be normal.

[0116] Through Domain1, find the device IP3 that also requests the domain name, and find the abnormal external connection log of device IP3 to find the connection with IP2. If the country, port and connection time of IP2 at the time of connection are all in the abnormal list (that is, the blacklist), then the associated device abnormality is returned.

[0117] The behavior correlation analysis engine outputs the final alarm, namely, the DGA high-risk alarm, based on the set correlation model, the detection results of abnormal behavior of the DGA request process and the detection results of abnormal associated devices.

[0118] In an exemplary embodiment, the embodiment of the present application further provides a computer storage medium, which is a computer-readable storage medium, for example, a memory including a computer program, and the above-mentioned computer program can be executed by a processor in the DGA domain name detection system based on behavioral scenarios to respectively complete the steps in the above-mentioned DGA domain name detection method based on behavioral scenarios.

[0119] The computer readable storage medium may be a magnetic random access memory, a read-only memory, a programmable read-only memory, an erasable programmable read-only memory, an electrically erasable programmable read-only memory, a flash memory, a magnetic surface storage, an optical disk, or a read-only optical disk and the like.

[0120] The above description is only an illustrative specific implementation manner of the present application. Without departing from the concept and principle of the present application, any equivalent changes and modifications made by any technician in the field should fall within the scope of protection of the present application.

Claims

1. A DGA domain name detection system based on behavioral scenarios, It is characterized in that It includes data collection and analysis module, DGA request behavior detection module, DGA request process behavior monitoring module, DGA request result behavior monitoring module, associated device behavior monitoring module and behavior association analysis engine; The data collection and analysis module is used to collect network traffic logs and DNS logs, and analyze the collected logs; The DGA request behavior detection module is used to detect the domain name obtained by log parsing to determine the DGA malicious domain name; The DGA request process behavior monitoring module is used to determine whether the DGA request process behavior is abnormal, the DGA request result behavior monitoring module is used to determine whether the DGA request result behavior is abnormal, and the associated device behavior monitoring module is used to obtain relevant clues from the associated device when local judgment clues are missing; The behavior correlation analysis engine is used to conduct comprehensive analysis based on the abnormal behavior analysis results of the DGA request process and the abnormal behavior analysis results of the DGA request result, so as to generate a DGA alarm based on the analysis results; The process of determining whether the DGA request process behavior is abnormal is as follows: Set baseline detection conditions, including NXDomain baseline, unfamiliar domain name access baseline, and the baseline of the number of IP addresses returned by requests for the same domain name; The NXDomain baseline anomaly detection is used to maintain the NXDomain number of each local device on the current day. The Gaussian distribution model of the NXDomain number of all devices in the network on the current day in the historical data is used as the baseline threshold. The NXDomain number of the local device corresponding to the current suspected DGA request on the current day is compared with the historical baseline threshold. If the NXDomain number of the local device corresponding to the current suspected DGA request on the current day exceeds the historical baseline threshold, the NXDomain number is determined to be abnormal. The unfamiliar domain name access baseline anomaly detection is used to maintain the number of unfamiliar non-whitelist domain name accesses of each local device on the same day. The Gaussian distribution model of the number of unfamiliar domain name accesses of all devices in the network in one day in the historical data is used as the baseline threshold. The number of unfamiliar domain name accesses counted on the same day by the local device corresponding to the current suspected DGA request is compared with the historical baseline threshold. If the number of unfamiliar domain name accesses counted on the same day by the local device corresponding to the current suspected DGA request exceeds the historical baseline threshold, the number of unfamiliar domain name accesses is determined to be abnormal. Preset an IP address number threshold, compare the number of IP addresses returned by the suspected DGA domain name on that day with the preset IP address number threshold, if the number of IP addresses returned by the suspected DGA domain name on that day exceeds the IP address number threshold, it is determined that an abnormal baseline of the number of IP addresses returned by the same domain name request has occurred; If at least two baseline detection conditions are abnormal, the DGA request process behavior is determined to be abnormal; The process of determining whether the DGA request result behavior is abnormal is as follows: Blacklists constructed using suspicious IP address source countries, suspicious common ports, and suspicious access times; Determine whether the data related to the external connection of the local device using the IP address obtained by the suspected DGA request matches the data in the blacklist. If so, a behavior anomaly is generated. At least two behavior anomalies are determined to be an external connection behavior anomaly. When local analysis clues are missing, relevant clues are obtained from associated devices to replace local analysis clues, including requesting whether other devices with the same domain name as the currently suspected DGA have abnormal DGA request process behavior, whether there are external connection behavior abnormalities, and if so, use the associated device behavior to replace the corresponding behavior of the local device.

2. The DGA domain name detection system based on behavior scenarios according to claim 1, It is characterized in that The DGA request behavior detection module includes a domain name whitelist and a DGA model detection engine; The domain name whitelist includes a learning module and a filtering module. The learning module is used to collect domain names that are accessed daily through DNS logs when the network is in a healthy state and save them as a whitelist; The filtering module is used to filter out the domain names on the whitelist in the DNS log to be detected according to the current whitelist; The DGA model detection engine includes a model training module and a model reasoning module. The model training module obtains whitelist samples from a domain name whitelist, retrains the DGA detection model and deploys the trained DGA detection model to the DGA model detection engine; the model reasoning module is used to send the domain name filtered by the filtering module to the trained DGA detection model for analysis.

3. The DGA domain name detection system based on behavior scenarios according to claim 1, It is characterized in that The DGA request process behavior monitoring module includes a NXDomain baseline monitoring module and a Fast-Flux feature monitoring module; The NXDomain baseline monitoring module determines whether the number of NXDomains returned by the DNS request exceeds the baseline range by checking the return result of the DNS request on the asset corresponding to the alarm reported by the DGA model detection engine; The Fast-Flux feature monitoring module provides baseline monitoring based on multiple unfamiliar domain name requests within a preset time period and baseline monitoring of the same domain name requests returning different IPs.

4. The DGA domain name detection system based on behavior scenarios according to claim 1, It is characterized in that The DGA request result behavior monitoring module includes a suspicious external connection monitoring module and a threat intelligence module; the suspicious external connection monitoring module is used to judge the geographic location, service port and time of the IP returned by the current asset connection DGA request, and analyze whether it is a suspicious behavior; the threat intelligence module is used to perform threat intelligence collision on the IP successfully returned by the DGA request to detect whether it is a known malicious IP.

5. A DGA domain name detection method based on behavioral scenarios, It is characterized in that The following steps are involved: Detect DGA request behavior to obtain the request domain name; Use machine learning algorithms to detect domain names to identify DGA malicious domain names; Conduct behavioral correlation analysis on DGA malicious domain names to determine whether the DGA request process behavior and DGA request result behavior are abnormal; Perform comprehensive analysis based on the abnormal analysis results of the DGA request process behavior and the abnormal analysis results of the DGA request result behavior, so as to generate a DGA alarm based on the analysis results; The specific process of behavioral association analysis of DGA malicious domain names is as follows: Conduct analysis and judgment on abnormal behavior during the DGA request process; Conduct abnormal analysis on DGA request results; The process of determining whether the DGA request process behavior is abnormal is as follows: Set baseline detection conditions, including NXDomain baseline, unfamiliar domain name access baseline, and the baseline of the number of IP addresses returned by requests for the same domain name; The NXDomain baseline anomaly detection is used to maintain the NXDomain number of each local device on the current day. The Gaussian distribution model of the NXDomain number of all devices in the network on the current day in the historical data is used as the baseline threshold. The NXDomain number of the local device corresponding to the current suspected DGA request on the current day is compared with the historical baseline threshold. If the NXDomain number of the local device corresponding to the current suspected DGA request on the current day exceeds the historical baseline threshold, the NXDomain number is determined to be abnormal. The unfamiliar domain name access baseline anomaly detection is used to maintain the number of unfamiliar non-whitelist domain name accesses of each local device on the same day. The Gaussian distribution model of the number of unfamiliar domain name accesses of all devices in the network in one day in the historical data is used as the baseline threshold. The number of unfamiliar domain name accesses counted on the same day by the local device corresponding to the current suspected DGA request is compared with the historical baseline threshold. If the number of unfamiliar domain name accesses counted on the same day by the local device corresponding to the current suspected DGA request exceeds the historical baseline threshold, the number of unfamiliar domain name accesses is determined to be abnormal. Preset an IP address number threshold, compare the number of IP addresses returned by the suspected DGA domain name on that day with the preset IP address number threshold, if the number of IP addresses returned by the suspected DGA domain name on that day exceeds the IP address number threshold, it is determined that an abnormal baseline of the number of IP addresses returned by the same domain name request has occurred; If at least two baseline detection conditions are abnormal, the DGA request process behavior is determined to be abnormal; The process of determining whether the DGA request result behavior is abnormal is as follows: Blacklists constructed using suspicious IP address source countries, suspicious common ports, and suspicious access times; Determine whether the data related to the external connection of the local device using the IP address obtained by the suspected DGA request matches the data in the blacklist. If so, a behavior anomaly is generated. At least two behavior anomalies are determined to be an external connection behavior anomaly. When local analysis clues are missing, relevant clues are obtained from associated devices to replace local analysis clues, including requesting whether other devices with the same domain name as the currently suspected DGA have abnormal DGA request process behavior, whether there are external connection behavior abnormalities, and if so, use the associated device behavior to replace the corresponding behavior of the local device.

6. The DGA domain name detection method based on behavior scenarios according to claim 5, It is characterized in that The specific process of detecting the DGA request behavior to obtain the requested domain name is as follows: When the network is in a healthy state, the domain names that are accessed daily are collected through DNS logs and saved as a whitelist. Based on the current whitelist, the request domain name in the DNS log to be detected is filtered. If the request domain name exists in the whitelist, it is filtered out. Otherwise, the request domain name is detected using a machine learning algorithm.

7. The DGA domain name detection method based on behavior scenarios according to claim 5, It is characterized in that The process of performing comprehensive judgment based on the abnormal judgment result of the DGA request process behavior and the abnormal judgment result of the DGA request result behavior to generate a DGA alarm based on the judgment result is as follows: If the IP address collision match of the threat intelligence is successful, a DGA high-risk alarm will be directly generated; Determine whether the DGA request process behavior of the current device is abnormal or whether there is a suspicious external connection anomaly. If there is an undetected anomaly, determine whether the associated device behavior anomaly can replace the undetected anomaly. If so, replace it. If the DGA request process behavior is abnormal and the conditions of suspicious external connection abnormality exist, a DGA high-risk alarm is generated; If only the DGA request process behavior abnormality occurs, a DGA medium-risk alarm is generated; If only DGA suspicious external connection abnormal behavior occurs, a DGA low-risk alarm is generated.