AI-enhanced network security threat intelligence analysis method and system
Through AI-enhanced network security threat intelligence analysis methods, semantic recognition and threat analysis models are used to extract and match network threat characteristics, solving the problem that traditional methods are difficult to capture unknown attack behavior, improving threat detection accuracy and response speed, and enhancing the network Security defense capability.
Patent Information
- Application Number
- CN202411194032.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-28
- Publication Date
- 2024-12-06
AI Technical Summary
Existing cybersecurity threat analysis methods are difficult to accurately capture the essential characteristics of unknown attack behavior, resulting in potential security threats being easily overlooked or misjudged, and it is difficult to detect and analyze new or mutated cyber attacks in a timely manner.
Using AI-enhanced network security threat intelligence analysis method, by obtaining threat intelligence data and network traffic data in real time, using pre-trained semantic recognition model to extract semantic text features, and input them into the threat analysis model for correlation feature extraction, combining importance Weighting and similarity matching, and perform preset processing strategies.
It improves the accuracy and response speed of threat detection, enhances network security defense capabilities, can more effectively capture multi-dimensional data characteristics and in-depth analysis, optimizes resource utilization, and significantly improves the detection and response capabilities of new or mutated network attacks.
Smart Images

Figure CN119094192A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of information processing, and in particular to an AI-enhanced network security threat intelligence analysis method and system. Background Art
[0002] With the rapid development of network technology, network security issues have become increasingly prominent, and various network attack methods have emerged in an endless stream, bringing great challenges to network security. Network traffic data is a key threat intelligence carrier that records network behavior, which may hide a variety of potential attack behaviors. Therefore, how to effectively detect attack behaviors from these threat intelligence has become a crucial part of improving network security protection capabilities.
[0003] Most existing network security threat analysis methods focus on identifying and responding to known threats, or extracting and matching corresponding response rules based on existing threats to respond to unknown threats through simple statistical features. However, in today's complex and changing network environment, these methods often find it difficult to accurately capture the essential characteristics of unknown attack behaviors, resulting in many potential security threats being easily ignored or misjudged, and it is difficult to timely and effectively discover and analyze new or mutated network attacks, thus bringing serious hidden dangers to network security. Summary of the invention
[0004] In order to solve the above problems, the present application provides an AI-enhanced network security threat intelligence analysis method and system.
[0005] The above-mentioned invention objective of the present application is achieved through the following technical solutions:
[0006] An AI-enhanced network security threat intelligence analysis method comprises the following steps:
[0007] Obtain threat intelligence data and network traffic data based on network attacks in real time, and identify threat patterns corresponding to threat intelligence data and abnormal patterns corresponding to network traffic data;
[0008] Input threat intelligence data and network traffic data into the pre-trained semantic recognition model to extract semantic text features;
[0009] Pack the network traffic data with abnormal patterns and threat intelligence data of the same time series and their corresponding semantic text features to obtain an abnormal data set;
[0010] Input the abnormal data set into the pre-trained threat analysis model to extract relevant features based on threat intelligence data, network traffic data and semantic text features;
[0011] Identify the association type of the association features output by the threat analysis model, and weight the importance of the association features and the threat pattern based on the association type and perform similarity matching;
[0012] The preset corresponding processing strategy is executed based on the importance weighting and similarity matching results.
[0013] By adopting the above technical solution, threat intelligence data and network traffic data based on network attacks are obtained from the network environment in real time, and the corresponding threat patterns and abnormal patterns are respectively identified, and the threat intelligence data and network traffic data are input into a pre-trained semantic recognition model, so that the semantic recognition model extracts its semantic text features to reflect the semantic content and contextual relationship of the data, and the network traffic data, threat intelligence data and their corresponding semantic text features with the same time sequence and containing abnormal patterns are packaged to form an abnormal data set containing multi-dimensional information, and the abnormal data set is input into the pre-trained threat analysis model, so that the threat analysis model extracts related features based on the threat intelligence data, network traffic data and semantic text features, and identifies these data. According to the potential connections and rules between the features, when the associated features output by the threat analysis model are received, their association types are identified, and based on the association type, the associated features are similarly matched with the threat pattern, and weighted processing is performed based on the similarity matching results and semantic text features, so that the preset corresponding processing strategy is executed according to the importance weighting and similarity matching results; the present application obtains threat intelligence data and network traffic data in the network environment in real time, and combines pre-trained semantic recognition models and threat analysis models to achieve association analysis and deep mining of multi-dimensional data features, which solves the problem that traditional network security analysis methods are difficult to accurately capture the essential characteristics of unknown attack behaviors, and has the effect of improving threat detection accuracy, enhancing threat response speed, and improving network security defense capabilities.
[0014] In a preferred example, the present application can be further configured as follows: the step of acquiring threat intelligence data and network traffic data based on network attacks in real time, and identifying threat patterns corresponding to the threat intelligence data and abnormal patterns corresponding to the network traffic data, comprises the following steps:
[0015] Obtain raw threat data and raw traffic data based on network attacks in real time;
[0016] The original threat data and original traffic data are cleaned and interference data is removed to obtain threat intelligence data based on network attacks and network traffic data.
[0017] By adopting the above-mentioned technical solution, the original threat data and original traffic data based on network attacks are obtained in real time, and they are preprocessed respectively, including data cleaning and interference data removal, so as to obtain purer and more accurate threat intelligence data and network traffic data; this application realizes the preliminary screening and purification of data by real-time acquisition and preprocessing of original network attack threat data and traffic data, solves the noise, redundancy and interference problems that may exist in the original data, and provides high-quality data input for subsequent threat intelligence analysis and network traffic anomaly detection, thereby improving the accuracy and efficiency of the entire analysis process.
[0018] In a preferred example, the present application can be further configured as follows: the step of packaging network traffic data with abnormal patterns, threat intelligence data and their corresponding semantic text features of the same time sequence to obtain an abnormal data set includes the following steps:
[0019] Perform time series analysis on threat intelligence data and network traffic data respectively to determine timestamps and durations of threat intelligence data and network traffic data;
[0020] Based on network traffic data with abnormal patterns, match threat intelligence data that has a timestamp close to its timestamp within a preset range and a duration whose overlap rate with its duration is greater than a preset value;
[0021] The network traffic data with abnormal patterns, the threat intelligence data matched with it, and the corresponding semantic text features are packaged into an abnormal data set.
[0022] By adopting the above technical solution, the threat intelligence data and network traffic data are analyzed in time series, so as to determine the timestamp and duration of each data, so as to understand its appearance time and continuous impact in the network environment. Based on the network traffic data with abnormal patterns, the threat intelligence data with a similar timestamp within a preset range and a duration overlap rate greater than a preset value are matched, and the network traffic data with abnormal patterns and the threat intelligence data matched thereto and their corresponding semantic text features are packaged to form an abnormal data set containing multi-dimensional information for subsequent deep threat analysis; this application realizes the effective integration of threat intelligence data, network traffic data and semantic text features through time series analysis and data matching technology, and solves the problems of weak data correlation and insufficient analysis depth in traditional methods. At the same time, by real-time packaging to form abnormal data sets, resource utilization is optimized and the response speed to new or variant network attacks is improved, which has the effect of significantly improving the accuracy of threat detection and network security defense capabilities.
[0023] In a preferred example, the present application can be further configured as follows: the threat analysis model includes a matching layer and an analysis layer, and the step of inputting the abnormal data set into the pre-trained threat analysis model to extract associated features based on threat intelligence data, network traffic data and semantic text features includes the steps of:
[0024] The matching layer matches the preset corresponding association extraction rules based on the threat patterns corresponding to the threat intelligence data in the abnormal data set and the abnormal patterns corresponding to the network traffic data;
[0025] The analysis layer extracts correlation features from threat intelligence data, network traffic data, and semantic text features based on matching correlation extraction rules.
[0026] By adopting the above technical solution, after receiving the abnormal data set, the matching layer matches the preset corresponding association extraction rules based on the threat pattern corresponding to its threat intelligence data and the abnormal pattern corresponding to the network traffic data. The analysis layer performs in-depth association feature extraction on the threat intelligence data, network traffic data and semantic text features based on the association extraction pair.
[0027] In a preferred example, the present application can be further configured as follows: the step of identifying the association type of the association feature output by the threat analysis model, and weighting the importance of the association feature and the threat pattern based on the association type and performing similarity matching, comprises the following steps:
[0028] Based on the association type of the association feature output by the threat analysis model, the association feature is represented as an association feature vector, wherein the association feature vector includes a plurality of association feature elements;
[0029] Extracting a threat feature corresponding to the threat pattern, and expressing the threat feature as a threat pattern vector, wherein the threat pattern vector includes a plurality of threat pattern elements;
[0030] Calculate the similarity between the associated feature vector and the threat pattern vector. The calculation formula is:
[0031] Among them, σ is the similarity, A i is the i-th associated characteristic element of the associated characteristic vector, B i is the ith threat pattern element of the threat pattern vector, and n is the dimension of the vector;
[0032] A weight is assigned to each associated feature element and threat pattern element based on the semantic text features, and the similarity between the weighted associated feature vector and the threat pattern vector is calculated.
[0033] By adopting the above technical solution, according to the association type of the association feature output by the threat analysis model, the association feature is converted into an association feature vector, the association feature vector contains several association feature elements, the threat feature corresponding to the threat pattern is extracted, and the threat feature is converted into a threat pattern vector, the threat pattern vector contains several threat pattern elements, the similarity between the association feature vector and the threat pattern vector is calculated based on a specific similarity calculation formula, a weight is assigned to each association special element and threat pattern element based on semantic text features, and the similarity between the weighted association feature vector and the threat pattern vector is calculated, so as to obtain a more accurate similarity matching result; the present application realizes the effective matching between the association feature output by the threat analysis model and the threat pattern through the representation of the association feature vector and the threat pattern vector, and the calculation of the similarity and weighted similarity, and the weight is assigned by combining the association type of the association feature with the semantic text feature, thereby improving the accuracy and pertinence of the matching, solving the problems of incomplete matching and low accuracy in traditional methods, and having the effect of significantly improving the accuracy of threat detection, enhancing network security defense capabilities and optimizing resource utilization.
[0034] In a preferred example, the present application can be further configured as follows: the step of assigning a weight to each associated feature element and threat pattern element based on the semantic text feature, and calculating the similarity between the weighted associated feature vector and the threat pattern vector, comprises the following steps:
[0035] Calculate the similarity between the weighted correlation feature vector and the threat pattern vector. The calculation formula is:
[0036] Among them, σ 加权 is the weighted similarity, A i is the i-th associated characteristic element of the associated characteristic vector, B i is the ith threat pattern element of the threat pattern vector, n is the dimension of the vector, α A,i is the weight of the i-th associated feature element, β B,i is the weight of the i-th threat pattern element.
[0037] By adopting the above technical solution, the similarity between the weighted associated feature vector and the threat pattern vector is calculated based on a specific similarity calculation formula. By considering the semantic text features, weights are assigned to the associated feature elements and the threat pattern elements, and the weighted similarity is calculated. This not only improves the matching accuracy, but also makes the matching process more in line with the actual context and semantic relationship.
[0038] The second object of the invention is achieved by the following technical solutions:
[0039] An AI-enhanced cybersecurity threat intelligence analysis system, comprising:
[0040] A data processing module is used to obtain threat intelligence data based on network attacks and network traffic data in real time, and to identify threat patterns corresponding to the threat intelligence data and abnormal patterns corresponding to the network traffic data;
[0041] A feature extraction module is used to input threat intelligence data and network traffic data into a pre-trained semantic recognition model to extract semantic text features;
[0042] The data packaging module is used to package the network traffic data with abnormal patterns and threat intelligence data of the same time sequence and their corresponding semantic text features to obtain an abnormal data set;
[0043] The correlation extraction module is used to input the abnormal data set into the pre-trained threat analysis model to extract correlation features based on threat intelligence data, network traffic data and semantic text features;
[0044] A similarity matching module is used to identify the association type of the association feature output by the threat analysis model, and to weight the importance of the association feature and the threat pattern and perform similarity matching based on the association type;
[0045] The execution module is used to execute the preset corresponding processing strategy based on the importance weighting and similarity matching results.
[0046] By adopting the above technical solution, a data processing module is used to obtain threat intelligence data and network traffic data based on network attacks in real time, and identify the threat mode corresponding to the threat intelligence data and the abnormal mode corresponding to the network traffic data; a feature extraction module is used to input the threat intelligence data and the network traffic data into a pre-trained semantic recognition model to extract semantic text features; a data packaging module is used to package the network traffic data, threat intelligence data and their corresponding semantic text features with abnormal modes of the same time series to obtain an abnormal data set; an association extraction module is used to input the abnormal data set into a pre-trained threat analysis model to extract association features based on threat intelligence data, network traffic data and semantic text features; a similarity matching module is used to identify the association type of the association feature output by the threat analysis model, and weight the importance and perform similarity matching on the association feature and the threat mode based on the association type; and an execution module is used to execute a preset corresponding processing strategy based on the importance weighting and similarity matching results.
[0047] In a preferred example, the present application can be further configured as follows: the data processing module includes:
[0048] The data acquisition submodule is used to obtain the original threat data and original traffic data based on network attacks in real time;
[0049] The preprocessing submodule is used to clean the original threat data and original traffic data and remove interference data, so as to obtain threat intelligence data based on network attacks and network traffic data.
[0050] By adopting the above technical solution, the data acquisition submodule is used to obtain the original threat data and original traffic data based on network attacks in real time; the preprocessing submodule is used to clean the original threat data and original traffic data and remove interference data, so as to obtain threat intelligence data based on network attacks and network traffic data.
[0051] In summary, the present application includes at least one of the following beneficial technical effects:
[0052] 1. This application obtains threat intelligence data and network traffic data in the network environment in real time, and combines pre-trained semantic recognition models and threat analysis models to achieve correlation analysis and deep mining of multi-dimensional data features, solving the problem that traditional network security analysis methods are difficult to accurately capture the essential characteristics of unknown attack behaviors, and has the effect of improving threat detection accuracy, enhancing threat response speed, and improving network security defense capabilities;
[0053] 2. By acquiring and preprocessing the original network attack threat data and traffic data in real time, the preliminary screening and purification of the data is achieved, and the noise, redundancy and interference problems that may exist in the original data are solved, providing high-quality data input for subsequent threat intelligence analysis and network traffic anomaly detection, thereby improving the accuracy and efficiency of the entire analysis process;
[0054] 3. Through time series analysis and data matching technology, the effective integration of threat intelligence data, network traffic data and semantic text features is achieved, solving the problems of weak data correlation and insufficient analysis depth in traditional methods. At the same time, by packaging abnormal data sets in real time, resource utilization is optimized and the response speed to new or variant network attacks is improved, which has the effect of significantly improving threat detection accuracy and network security defense capabilities;
[0055] 4. Through the representation of associated feature vectors and threat pattern vectors, as well as the calculation of similarity and weighted similarity, effective matching between the associated features output by the threat analysis model and the threat pattern is achieved. The weights are assigned based on the association type of the associated features combined with the semantic text features, thereby improving the accuracy and pertinence of the matching, solving the problems of incomplete matching and low accuracy in traditional methods, and significantly improving the accuracy of threat detection, enhancing network security defense capabilities, and optimizing resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS
[0056] Figure 1 This is a flow chart of an embodiment of an AI-enhanced network security threat intelligence analysis method of the present application;
[0057] Figure 2 This is a flowchart for implementing step S30 in an embodiment of an AI-enhanced network security threat intelligence analysis method of the present application;
[0058] Figure 3 This is an implementation flowchart of step S50 in an embodiment of an AI-enhanced network security threat intelligence analysis method of the present application. DETAILED DESCRIPTION
[0059] The following is combined with Figure 1-3 This application is described in further detail.
[0060] In one embodiment, if Figure 1 As shown, the present application discloses an AI-enhanced network security threat intelligence analysis method, which specifically includes the following steps:
[0061] S10: Acquire threat intelligence data based on network attacks and network traffic data in real time, and identify threat patterns corresponding to the threat intelligence data and abnormal patterns corresponding to the network traffic data;
[0062] In this embodiment, threat intelligence data refers to information data related to network security threats, and its source channels include open source intelligence, hacker forums, malware samples, etc.; network traffic data refers to data packets and traffic information during network transmission, including source address, destination address, transmission protocol, data packet size, etc., which are used to monitor and analyze network activities to detect abnormal or potential attack behaviors; threat patterns are specific attack patterns or behavior characteristics identified in threat intelligence data, which represent a known attack method or malicious behavior, and are used to match and detect network traffic data; abnormal patterns are patterns or characteristics detected in network traffic data that are inconsistent with normal network behaviors, which represent the existence of potential network attacks or malicious activities;
[0063] Specifically, it obtains threat intelligence data based on network attacks and network traffic data from the network environment in real time, and identifies the corresponding threat patterns and abnormal patterns respectively;
[0064] S20: Input the threat intelligence data and network traffic data into the pre-trained semantic recognition model to extract semantic text features;
[0065] In this embodiment, the semantic recognition model is a pre-trained machine learning model for extracting semantic features from text data, that is, for understanding the meaning and context of text, and is specifically used to extract useful text information from threat intelligence data and network traffic data;
[0066] Specifically, the threat intelligence data and network traffic data are input into a pre-trained semantic recognition model, so that the semantic recognition model extracts its semantic text features to reflect the semantic content and contextual relationship of the data;
[0067] S30: Packing network traffic data with abnormal patterns and threat intelligence data of the same time series and their corresponding semantic text features to obtain an abnormal data set;
[0068] In this embodiment, the abnormal data set is a data set formed by packaging network traffic data and threat intelligence data with abnormal patterns in the same time series and semantic text features corresponding to the network traffic data, which is used for subsequent threat analysis and processing;
[0069] Specifically, the network traffic data, threat intelligence data and their corresponding semantic text features with the same time sequence and containing abnormal patterns are packaged to form an abnormal data set containing multi-dimensional information;
[0070] S40: Input the abnormal data set into the pre-trained threat analysis model to extract related features based on threat intelligence data, network traffic data and semantic text features;
[0071] In this embodiment, the threat analysis model is a machine learning model used to analyze the correlation features between threat intelligence data, network traffic data, and semantic text features in an abnormal data set; the correlation features are features extracted by the threat analysis model that may represent a certain attack behavior or malicious threat;
[0072] Specifically, the abnormal data set is input into a pre-trained threat analysis model, so that the threat analysis model extracts related features based on threat intelligence data, network traffic data, and semantic text features, and identifies potential connections and patterns between these data features;
[0073] S50: Identify the association type of the association feature output by the threat analysis model, and perform importance weighting and similarity matching on the association feature and the threat pattern based on the association type;
[0074] In this embodiment, the association type is different categories or types of association features, including different types of attack techniques, malicious behavior types, and network activity pattern types; importance weighting is to assign weights to association features and threat patterns according to the importance of semantic text features; similarity matching includes comparing the association features before weighting with the threat pattern, and the association features after weighting with the threat pattern, so as to calculate their similarity;
[0075] Specifically, when receiving the correlation feature output by the threat analysis model, identifying its correlation type, and performing similarity matching between the correlation feature and the threat pattern based on the correlation type, and performing weighted processing based on the similarity matching result and the semantic text feature;
[0076] S60: Execute a preset corresponding processing strategy based on the importance weighting and similarity matching results.
[0077] In this embodiment, the preset corresponding processing strategy is a preset countermeasure for the identified network threat based on the importance weighting and similarity matching results, including isolating the attacked system and preventing the spread of malicious traffic, etc.;
[0078] Specifically, according to the importance weighting and similarity matching results, a preset rule strategy for processing the corresponding information is executed.
[0079] In one embodiment, step S10 comprises the steps of:
[0080] S11: Obtain original threat data and original traffic data based on network attacks in real time;
[0081] S12: Clean the original threat data and original traffic data and remove interference data to obtain threat intelligence data based on network attacks and network traffic data.
[0082] In this embodiment, the original threat data is the unprocessed original data related to network attacks directly obtained from different source channels; the original traffic data is the unprocessed traffic data directly captured from the network; data cleaning is a data preprocessing technology used to remove noise, errors, duplications or inconsistent data in the original data, including removing invalid network traffic records, correcting erroneous threat intelligence information, etc.; interference data removal is to remove data that may interfere with analysis or cause false positives from the original data, including removing abnormal data caused by network fluctuations, equipment failures or malicious interference;
[0083] Specifically, the original threat data and original traffic data based on network attacks are obtained in real time, and pre-processed respectively, including data cleaning and interference data removal, so as to obtain purer and more accurate threat intelligence data and network traffic data.
[0084] In one embodiment, if Figure 2 As shown, step S30 includes the steps of:
[0085] S31: performing time series analysis on the threat intelligence data and the network traffic data respectively to determine timestamps and durations of the threat intelligence data and the network traffic data;
[0086] S32: Based on the network traffic data with the abnormal pattern, match the threat intelligence data that has a timestamp close to the timestamp within a preset range and a duration whose coincidence rate with the duration is greater than a preset value;
[0087] S33: Pack the network traffic data with abnormal patterns, the threat intelligence data matched therewith, and the corresponding semantic text features into an abnormal data set.
[0088] In this embodiment, time series analysis is to analyze the data in chronological order, specifically to determine the timestamp and duration of threat intelligence data and network traffic data, so as to understand the development process and duration of attack events; the timestamp is the specific time point when the data or event occurs, specifically in the threat intelligence data and network traffic data, the timestamp is used to identify the moment when the data or event occurs, so as to perform time series analysis and matching; the duration is the length of time that the threat intelligence data and network traffic data last;
[0089] Specifically, time series analysis is performed on threat intelligence data and network traffic data respectively to determine the timestamp and duration of each data in order to understand its appearance time and continuous impact in the network environment. Based on the network traffic data with abnormal patterns, the threat intelligence data with a timestamp close to the preset range and a duration overlap rate greater than the preset value is matched. The network traffic data with abnormal patterns and the threat intelligence data matched to it and their corresponding semantic text features are packaged to form an abnormal data set containing multi-dimensional information for subsequent in-depth threat analysis.
[0090] In one embodiment, the threat analysis model includes a matching layer and an analysis layer, and step S40 includes the steps of:
[0091] S41: The matching layer matches the preset corresponding association extraction rules based on the threat pattern corresponding to the threat intelligence data in the abnormal data set and the abnormal pattern corresponding to the network traffic data;
[0092] S42: The analysis layer extracts correlation features from threat intelligence data, network traffic data, and semantic text features based on matching correlation extraction rules.
[0093] In this embodiment, the matching layer is a layer of the threat analysis model used to match the associated extraction rules based on the threat pattern and the abnormal pattern; the associated extraction rules are preset rules for guiding how to extract key features from the threat intelligence data, network traffic data and semantic text features, which perform feature extraction based on known attack patterns, behavioral characteristics or network activity patterns; the analysis layer is a layer responsible for performing data analysis, specifically, performing associated feature extraction on the threat intelligence data, network traffic data and semantic text features based on the associated extraction rules;
[0094] Specifically, after receiving the abnormal data set, the matching layer matches the preset corresponding association extraction rules based on the threat pattern corresponding to its threat intelligence data and the abnormal pattern corresponding to the network traffic data. The analysis layer performs in-depth association feature extraction on the threat intelligence data, network traffic data and semantic text features based on the association extraction pair.
[0095] In one embodiment, if Figure 3 As shown, step S50 includes the steps of:
[0096] S51: Based on the association type of the association feature output by the threat analysis model, the association feature is represented as an association feature vector, where the association feature vector includes a plurality of association feature elements;
[0097] S52: extracting a threat feature corresponding to the threat pattern, and expressing the threat feature as a threat pattern vector, wherein the threat pattern vector includes a plurality of threat pattern elements;
[0098] S53: Calculate the similarity between the associated feature vector and the threat pattern vector. The calculation formula is:
[0099] Among them, σ is the similarity, A i is the i-th associated characteristic element of the associated characteristic vector, B i is the ith threat pattern element of the threat pattern vector, and n is the dimension of the vector;
[0100] S54: assigning a weight to each associated feature element and threat pattern element based on the semantic text feature, and calculating the similarity between the weighted associated feature vector and the threat pattern vector.
[0101] In this embodiment, the associated feature vector is a vector form of the associated feature, so as to perform numerical calculation and similarity comparison; the associated feature element includes the projection value and directionality of the pattern represented by the associated feature vector; the threat pattern vector is a vector form of the threat pattern, so as to perform similarity comparison with the associated feature vector; the threat pattern element includes threat indicators and dynamic attributes, the threat indicator is a numerical indicator including the attack source, attack type, target system, exploited vulnerability, etc., and the dynamic attribute is an attribute value including the attack frequency, duration, impact range, etc., and these attribute values can change over time; the similarity is the degree of similarity between two vectors, specifically used to measure the degree of match between the associated feature vector and the threat pattern vector, so as to determine whether the network traffic data contains features similar to the known threat pattern; the similarity between the weighted associated feature vector and the threat pattern vector is the degree of similarity between the associated feature vector and the threat pattern vector calculated after considering the weight, specifically used to more accurately evaluate the degree of match between the associated feature vector and the threat pattern vector;
[0102] Specifically, according to the association type of the association feature output by the threat analysis model, the association feature is converted into an association feature vector, the association feature vector includes a number of association feature elements, the threat feature corresponding to the threat pattern is extracted, and the threat feature is converted into a threat pattern vector, the threat pattern vector includes a number of threat pattern elements, the similarity between the association feature vector and the threat pattern vector is calculated based on a specific similarity calculation formula, a weight is assigned to each association special element and threat pattern element based on semantic text features, and the similarity between the weighted association feature vector and the threat pattern vector is calculated, so as to obtain a more accurate similarity matching result.
[0103] In one embodiment, step S54 comprises the steps of:
[0104] S541: Calculate the similarity between the weighted correlation feature vector and the threat pattern vector. The calculation formula is:
[0105] Among them, σ 加权 is the weighted similarity, A i is the i-th associated characteristic element of the associated characteristic vector, B i is the ith threat pattern element of the threat pattern vector, n is the dimension of the vector, α A,i is the weight of the i-th associated feature element, β B,i is the weight of the i-th threat pattern element.
[0106] In this embodiment, the associated feature vectors and threat pattern vectors are weighted based on TF-IDF. TF-IDF is a commonly used weighting technology for information retrieval and text mining. TF (term frequency) represents the frequency of occurrence of feature elements in the feature text, and IDF (inverse document frequency) represents the general importance of feature elements. A document is a specific text unit in a text data set, specifically a specific data point or record in network traffic data.
[0107] Specifically, the similarity between the weighted associated feature vector and the threat pattern vector is calculated based on a specific similarity calculation formula. By considering the semantic text features, weights are assigned to the associated feature elements and the threat pattern elements, and the weighted similarity is calculated. This not only improves the matching accuracy, but also makes the matching process more in line with the actual context and semantic relationship.
[0108] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.
[0109] In one embodiment, an AI-enhanced network security threat intelligence analysis system is provided, and the AI-enhanced network security threat intelligence analysis system corresponds one-to-one to the AI-enhanced network security threat intelligence analysis method in the above embodiment. An AI-enhanced network security threat intelligence analysis system includes:
[0110] A data processing module is used to obtain threat intelligence data based on network attacks and network traffic data in real time, and to identify threat patterns corresponding to the threat intelligence data and abnormal patterns corresponding to the network traffic data;
[0111] A feature extraction module is used to input threat intelligence data and network traffic data into a pre-trained semantic recognition model to extract semantic text features;
[0112] The data packaging module is used to package the network traffic data with abnormal patterns and threat intelligence data of the same time sequence and their corresponding semantic text features to obtain an abnormal data set;
[0113] The correlation extraction module is used to input the abnormal data set into the pre-trained threat analysis model to extract correlation features based on threat intelligence data, network traffic data and semantic text features;
[0114] A similarity matching module is used to identify the association type of the association feature output by the threat analysis model, and to weight the importance of the association feature and the threat pattern and perform similarity matching based on the association type;
[0115] An execution module, used for executing a preset corresponding processing strategy based on importance weighting and similarity matching results;
[0116] Optionally, the data processing module includes:
[0117] The data acquisition submodule is used to obtain the original threat data and original traffic data based on network attacks in real time;
[0118] The preprocessing submodule is used to clean the original threat data and original traffic data and remove interference data, so as to obtain threat intelligence data based on network attacks and network traffic data;
[0119] Optionally, also include:
[0120] A time series analysis module is used to perform time series analysis on threat intelligence data and network traffic data respectively to determine the timestamp and duration of the threat intelligence data and the network traffic data;
[0121] A timing matching module is used to match threat intelligence data having a timestamp close to its timestamp within a preset range and a duration whose coincidence rate with its duration is greater than a preset value based on network traffic data with an abnormal pattern;
[0122] The data packaging module is specifically used to package the network traffic data with abnormal patterns, the threat intelligence data matched therewith, and the corresponding semantic text features into an abnormal data set;
[0123] Optionally, also include:
[0124] A matching layer module is used to match preset corresponding association extraction rules based on the threat pattern corresponding to the threat intelligence data in the abnormal data set and the abnormal pattern corresponding to the network traffic data;
[0125] The analysis layer module is used to extract correlation features from threat intelligence data, network traffic data, and semantic text features based on matching correlation extraction rules;
[0126] Optionally, also include:
[0127] A correlation feature vector acquisition module is used to represent the correlation feature as a correlation feature vector based on the correlation type of the correlation feature output by the threat analysis model, wherein the correlation feature vector includes a plurality of correlation feature elements;
[0128] A threat pattern vector acquisition module is used to extract threat features corresponding to the threat pattern and express the threat features as a threat pattern vector, wherein the threat pattern vector includes a plurality of threat pattern elements;
[0129] The similarity calculation module is used to calculate the similarity between the associated feature vector and the threat pattern vector. The calculation formula is:
[0130] Among them, σ is the similarity, A i is the i-th associated characteristic element of the associated characteristic vector, B i is the ith threat pattern element of the threat pattern vector, and n is the dimension of the vector;
[0131] A weight assignment module is used to assign a weight to each associated feature element and threat pattern element based on semantic text features, and calculate the similarity between the weighted associated feature vector and the threat pattern vector;
[0132] Optionally, the weight distribution module is specifically used to calculate the similarity between the weighted associated feature vector and the threat pattern vector, and the calculation formula is:
[0133] Among them, σ 加权 is the weighted similarity, A i is the i-th associated characteristic element of the associated characteristic vector, B i is the ith threat pattern element of the threat pattern vector, n is the dimension of the vector, α A,i is the weight of the i-th associated feature element, β B,i is the weight of the i-th threat pattern element.
[0134] For the specific definition of an AI-enhanced network security threat intelligence analysis system, please refer to the definition of an AI-enhanced network security threat intelligence analysis method above, which will not be repeated here. Each module in the above-mentioned AI-enhanced network security threat intelligence analysis system can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.
[0135] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.
Claims
1. An AI-enhanced network security threat intelligence analysis method, characterized by: Includes steps: Obtain threat intelligence data and network traffic data based on network attacks in real time, and identify threat patterns corresponding to threat intelligence data and abnormal patterns corresponding to network traffic data; Input threat intelligence data and network traffic data into the pre-trained semantic recognition model to extract semantic text features; Pack the network traffic data with abnormal patterns and threat intelligence data of the same time series and their corresponding semantic text features to obtain an abnormal data set; Input the abnormal data set into the pre-trained threat analysis model to extract relevant features based on threat intelligence data, network traffic data and semantic text features; Identify the association type of the association features output by the threat analysis model, and weight the importance of the association features and the threat pattern based on the association type and perform similarity matching; The preset corresponding processing strategy is executed based on the importance weighting and similarity matching results.
2. The AI-enhanced network security threat intelligence analysis method according to claim 1, characterized in that: The step of acquiring threat intelligence data and network traffic data based on network attacks in real time, and identifying threat patterns corresponding to the threat intelligence data and abnormal patterns corresponding to the network traffic data, comprises the following steps: Obtain raw threat data and raw traffic data based on network attacks in real time; The original threat data and original traffic data are cleaned and interference data is removed to obtain threat intelligence data based on network attacks and network traffic data.
3. The AI-enhanced network security threat intelligence analysis method according to claim 1, characterized in that: The step of packaging network traffic data with abnormal patterns, threat intelligence data and their corresponding semantic text features in the same time sequence to obtain an abnormal data set includes the following steps: Perform time series analysis on threat intelligence data and network traffic data respectively to determine timestamps and durations of threat intelligence data and network traffic data; Based on network traffic data with abnormal patterns, match threat intelligence data that has a timestamp close to its timestamp within a preset range and a duration whose overlap rate with its duration is greater than a preset value; The network traffic data with abnormal patterns, the threat intelligence data matched with it, and the corresponding semantic text features are packaged into an abnormal data set.
4. The AI-enhanced network security threat intelligence analysis method according to claim 1, characterized in that: The threat analysis model includes a matching layer and an analysis layer. The step of inputting the abnormal data set into the pre-trained threat analysis model to extract associated features based on threat intelligence data, network traffic data and semantic text features includes the steps of: The matching layer matches the preset corresponding association extraction rules based on the threat patterns corresponding to the threat intelligence data in the abnormal data set and the abnormal patterns corresponding to the network traffic data; The analysis layer extracts correlation features from threat intelligence data, network traffic data, and semantic text features based on matching correlation extraction rules.
5. The AI-enhanced network security threat intelligence analysis method according to claim 1, characterized in that: The step of identifying the association type of the association feature output by the threat analysis model, and weighting the importance of the association feature and the threat pattern based on the association type and performing similarity matching includes the following steps: Based on the association type of the association feature output by the threat analysis model, the association feature is represented as an association feature vector, wherein the association feature vector includes a plurality of association feature elements; Extracting a threat feature corresponding to the threat pattern, and expressing the threat feature as a threat pattern vector, wherein the threat pattern vector includes a plurality of threat pattern elements; Calculate the similarity between the associated feature vector and the threat pattern vector. The calculation formula is: Among them, σ is the similarity, A i is the i-th associated characteristic element of the associated characteristic vector, B i is the ith threat pattern element of the threat pattern vector, and n is the dimension of the vector; A weight is assigned to each associated feature element and threat pattern element based on the semantic text features, and the similarity between the weighted associated feature vector and the threat pattern vector is calculated.
6. The AI-enhanced network security threat intelligence analysis method according to claim 5, characterized in that: The step of assigning a weight to each associated feature element and threat pattern element based on the semantic text feature, and calculating the similarity between the weighted associated feature vector and the threat pattern vector, comprises the steps of: Calculate the similarity between the weighted correlation feature vector and the threat pattern vector. The calculation formula is: Among them, σ 加权 is the weighted similarity, A i is the i-th associated characteristic element of the associated characteristic vector, B i is the ith threat pattern element of the threat pattern vector, n is the dimension of the vector, α A, i is the weight of the i-th associated feature element, β B,i is the weight of the i-th threat pattern element.
7. An AI-enhanced cybersecurity threat intelligence analysis system, characterized by: include: A data processing module is used to obtain threat intelligence data based on network attacks and network traffic data in real time, and to identify threat patterns corresponding to the threat intelligence data and abnormal patterns corresponding to the network traffic data; A feature extraction module is used to input threat intelligence data and network traffic data into a pre-trained semantic recognition model to extract semantic text features; The data packaging module is used to package the network traffic data with abnormal patterns and threat intelligence data of the same time sequence and their corresponding semantic text features to obtain an abnormal data set; The correlation extraction module is used to input the abnormal data set into the pre-trained threat analysis model to extract correlation features based on threat intelligence data, network traffic data and semantic text features; A similarity matching module is used to identify the association type of the association feature output by the threat analysis model, and to weight the importance of the association feature and the threat pattern and perform similarity matching based on the association type; The execution module is used to execute the preset corresponding processing strategy based on the importance weighting and similarity matching results.
8. The AI-enhanced network security threat intelligence analysis system according to claim 7, characterized in that: The data processing module comprises: The data acquisition submodule is used to obtain the original threat data and original traffic data based on network attacks in real time; The preprocessing submodule is used to clean the original threat data and original traffic data and remove interference data, so as to obtain threat intelligence data based on network attacks and network traffic data.
Citation Information
Cited By
User identity threat detection method and system based on large language model
CN120316757A
A user identity threat detection method and system based on large language model
CN120316757B
