Threat intelligence analysis method based on large-scale model collaboration

By adopting a large-scale model collaboration method in IoT threat intelligence analysis, extracting threat intelligence content from dark web pages and semantic encoding, the problems of low efficiency in threat intelligence acquisition and data complexity in the existing technology are solved, and threats are realized Efficient structured and accurate analysis of intelligence.

CN118982028AInactive Publication Date: 2024-11-19STATE GRID HENAN INFORMATION & TELECOMM CO +2
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202411127448.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-16
Publication Date
2024-11-19
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the prior art, the acquisition of Internet of Things threat intelligence is inefficient, high cost and insufficient timeliness, and the diversity and complexity of threat intelligence data make it difficult to integrate intelligence and intelligent analysis.

Method used

Threat intelligence analysis method based on large-scale model collaboration is adopted, and threat intelligence content information is extracted by obtaining dark web pages, and global content semantic encoding and keyword extraction are performed based on word granularity, and interactive matching results are calculated to obtain structured data.

Benefits of technology

Effectively convert unstructured threat intelligence reports into structured data to improve the utilization efficiency and analysis accuracy of threat intelligence.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982028A_ABST
    Figure CN118982028A_ABST
Patent Text Reader

Abstract

The invention relates to the field of intelligent intelligence analysis, in particular to a threat intelligence analysis method based on large-scale model collaboration. Threat intelligence content information is extracted from a dark web webpage by adopting a natural language processing technology based on deep learning, and global content information semantic feature extraction and keyword information semantic feature extraction are respectively performed on the threat intelligence content information; and performing fine-grained context semantic association reinforcement on global content information based on the keyword information of the threat intelligence so as to realize deep understanding and structured representation of the content of the threat intelligence. Therefore, the unstructured threat intelligence report can be effectively converted into the structured data, and the utilization efficiency and the analysis accuracy of the threat intelligence can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of intelligent intelligence analysis, and specifically to a threat intelligence analysis method based on large-scale model collaboration. Background Art

[0002] With the rapid development of the Internet and the Internet of Things, network security issues are becoming increasingly severe, and various types of network attacks and threats are emerging one after another. Enterprises and individuals are facing unprecedented security challenges. How to timely and accurately identify, predict and prevent network threats has become a core issue in the Internet era. In this context, threat intelligence is an important part of network security defense. The efficiency and quality of its collection, analysis and utilization are directly related to the overall effectiveness of network security.

[0003] However, the current acquisition of IoT threat intelligence has high limitations. The acquisition of traditional threat intelligence mainly relies on manual search and information extraction, which is not only inefficient, but also has high labor costs and lacks timeliness. In addition, threat intelligence data is usually diverse and complex, containing a large amount of unstructured text data, such as security reports, vulnerability warnings, security information, etc. These data have significant differences in format, language and content, and lack a unified structured standard, which further increases the difficulty of intelligence integration and intelligent analysis. Therefore, an optimized threat intelligence analysis method based on large-scale model collaboration is expected. Summary of the invention

[0004] This application is made in view of the above problems. One purpose of this application is to provide a threat intelligence analysis method based on large-scale model collaboration.

[0005] The embodiment of the present application provides a threat intelligence analysis method based on large-scale model collaboration, which includes: Get dark web pages; Extracting threat intelligence content information from the dark web page; Performing word-granularity-based global content semantic encoding on the threat intelligence content information to obtain a set of threat intelligence content word-granularity encoding vectors; Performing keyword extraction and semantic coding on the threat intelligence content information to obtain a set of threat intelligence keyword context semantic coding vectors; Calculate the interactive matching results of each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors relative to the set of threat intelligence keyword context semantic encoding vectors to obtain structured data of the threat intelligence content information.

[0006] For example, according to the threat intelligence analysis method based on large-scale model collaboration of an embodiment of the present application, the threat intelligence content information is subjected to global content semantic encoding based on word granularity to obtain a set of threat intelligence content word granularity encoding vectors, including: After word segmentation processing is performed on the threat intelligence content information, a threat intelligence semantic encoder including a word embedding layer is passed through to obtain a set of word granularity encoding vectors of the threat intelligence content.

[0007] For example, according to the threat intelligence analysis method based on large-scale model collaboration according to an embodiment of the present application, keyword extraction and semantic encoding are performed on the threat intelligence content information to obtain a set of threat intelligence keyword context semantic encoding vectors, including: Extracting keywords from the threat intelligence content information to obtain a set of threat intelligence keywords; The set of threat intelligence keywords is context-associated encoded to obtain a set of threat intelligence keyword context semantic encoding vectors.

[0008] For example, according to the threat intelligence analysis method based on large-scale model collaboration of an embodiment of the present application, keyword extraction is performed on the threat intelligence content information to obtain a set of threat intelligence keywords, including: The threat intelligence content information is input into a keyword extractor based on the AIGC model to obtain a set of threat intelligence keywords.

[0009] For example, according to the threat intelligence analysis method based on large-scale model collaboration of an embodiment of the present application, context association encoding is performed on the set of threat intelligence keywords to obtain a set of context semantic encoding vectors of the threat intelligence keywords, including: The set of threat intelligence keywords is input into a keyword context association encoder including a keyword embedding layer to obtain a set of threat intelligence keyword context semantic encoding vectors.

[0010] For example, according to the threat intelligence analysis method based on large-scale model collaboration of an embodiment of the present application, the interactive matching result of each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors relative to the set of threat intelligence keyword context semantic encoding vectors is calculated to obtain structured data of the threat intelligence content information, including: Taking each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors as a query feature vector, the one-way attention granularity scanning interactive matching result of the query feature vector relative to the set of threat intelligence keyword context semantic encoding vectors is calculated to obtain a set of keyword information-guided threat intelligence content word granularity encoding vectors as structured data of the threat intelligence content information.

[0011] For example, according to the threat intelligence analysis method based on large-scale model collaboration of an embodiment of the present application, each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors is used as a query feature vector, and the query feature vector is calculated relative to the set of threat intelligence keyword context semantic encoding vectors. The one-way attention granularity scanning interactive matching result is obtained to obtain a set of keyword information-guided threat intelligence content word granularity encoding vectors as structured data of the threat intelligence content information, including: Calculating semantic metric coefficients between the query feature vector and each threat intelligence keyword context semantic coding vector in the set of threat intelligence keyword context semantic coding vectors to obtain a sequence of one-way matching semantic metric coefficients; Normalizing the sequence of one-way matching semantic metric coefficients through a softmax function to obtain a sequence of one-way matching semantic metric weights; Using the sequence of the one-way matching semantic measurement weights as weights, calculating the weighted sum of the set of threat intelligence keyword context semantic encoding vectors to obtain a threat intelligence keyword context semantic reshaping representation vector; The query feature vector and the threat intelligence keyword context semantic reshaping representation vector are calculated by point-by-point multiplication to obtain the keyword information-guided threat intelligence content word granularity encoding vector.

[0012] For example, according to the threat intelligence analysis method based on large-scale model collaboration of an embodiment of the present application, the semantic metric coefficient between the query feature vector and each threat intelligence keyword context semantic encoding vector in the set of threat intelligence keyword context semantic encoding vectors is calculated to obtain a sequence of one-way matching semantic metric coefficients, including: Calculating the square of each eigenvalue of each threat intelligence keyword context semantic coding vector in the set of threat intelligence keyword context semantic coding vectors to obtain a set of non-linear transformation threat intelligence keyword context semantic coding vectors; Cascading the query feature vector and each non-linear transformation threat intelligence keyword context semantic encoding vector in the set of non-linear transformation threat intelligence keyword context semantic encoding vectors to obtain a sequence of semantic association feature vectors; Using a preset transformation vector to multiply each semantic association feature vector in the sequence of semantic association feature vectors respectively to obtain a sequence of semantic similarities; Each semantic similarity in the sequence of semantic similarities is added with a bias term and then passed through a sigmoid activation function to obtain a sequence of one-way matching semantic measurement coefficients.

[0013] According to the threat intelligence analysis method based on large-scale model collaboration of the embodiment of the present application, it can effectively convert unstructured threat intelligence reports into structured data, thereby improving the utilization efficiency and analysis accuracy of threat intelligence. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application are briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present application, and are not intended to limit the present application.

[0015] Figure 1 A schematic diagram of the application architecture of a threat intelligence analysis method based on large-scale model collaboration in an embodiment of the present application is shown; Figure 2 A flowchart of a threat intelligence analysis method based on large-scale model collaboration in an embodiment of the present application is shown; Figure 3 A flowchart of sub-step S540 of the threat intelligence analysis method based on large-scale model collaboration in an embodiment of the present application is shown; Figure 4 The dark web threat intelligence extraction framework in the embodiment of the present application is shown; Figure 5 A schematic diagram of the structure of a threat intelligence analysis system based on large-scale model collaboration in an embodiment of the present application is shown; and Figure 6 An application scenario diagram of a threat intelligence analysis method based on large-scale model collaboration in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0017] The terms used in this specification are those common terms currently widely used in the art in consideration of the functions of the present application, but these terms may vary according to the intentions of those skilled in the art, precedents, or new technologies in the art. In addition, specific terms may be selected, and in this case, their detailed meanings will be described in the detailed description of the present application. Therefore, the terms used in the specification should not be understood as simple names, but rather as a general description based on the meaning of the terms and the present application.

[0018] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0019] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. At the same time, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0020] Figure 1 A schematic diagram of the application architecture of a threat intelligence analysis method based on large-scale model collaboration in an embodiment of the present application is shown, including a server 100 and a terminal device 200.

[0021] The terminal device 200 and the server 100 can be connected via the Internet to achieve mutual communication. Optionally, the above-mentioned Internet uses standard communication technology and / or protocol. The Internet is usually the Internet, but it can also be any network, including but not limited to any combination of a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a mobile, wired or wireless network, a private network or a virtual private network. In some embodiments, the data exchanged through the network can be represented by technologies and / or formats including Hyper Text Markup Language (HTML), Extensible Markup Language (XML), etc. In addition, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), Internet Protocol Security (IPsec), etc. can be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can also be used to replace or supplement the above-mentioned data communication technologies.

[0022] The server 100 can provide various network services for the terminal device 200, wherein the server 100 can be a single server, a server cluster consisting of several servers, or a cloud computing center. Specifically, the server 100 may include a processor 110 (Center Processing Unit, CPU), a memory 120, an input device 130, and an output device 140, etc. The input device 130 may include a keyboard, a mouse, a touch screen, etc., and the output device 140 may include a display device, such as a liquid crystal display (Liquid Crystal Display, LCD), a cathode ray tube (Cathode Ray Tube, CRT), etc.

[0023] The memory 120 may include a read-only memory (ROM) and a random access memory (RAM), and provides the processor 110 with program instructions and data stored in the memory 120. In an embodiment of the present application, the memory 120 may be used to store the program of the threat intelligence analysis method based on large-scale model collaboration in an embodiment of the present application.

[0024] The processor 110 calls the program instructions stored in the memory 120, and the processor 110 is used to execute the steps of any threat intelligence analysis method based on large-scale model collaboration in the embodiments of the present application according to the obtained program instructions.

[0025] In addition, the application architecture diagram in the embodiment of the present application is intended to more clearly illustrate the technical solution in the embodiment of the present application, and does not constitute a limitation on the technical solution provided in the embodiment of the present application. Of course, for other application architectures and business applications, the technical solution provided in the embodiment of the present application is also applicable to similar problems.

[0026] The following is a non-restrictive description of the threat intelligence analysis method based on large-scale model collaboration provided according to at least one embodiment of the present application through several examples or embodiments. As described below, different features in these specific examples or embodiments can be combined with each other without conflicting with each other to obtain new examples or embodiments, and these new examples or embodiments also fall within the scope of protection of the present application.

[0027] In response to the above technical problems, the technical concept of this application is to use natural language processing technology based on deep learning to extract threat intelligence content information from dark web pages, and extract global content information semantic features and keyword information semantic features from the threat intelligence content information, and then strengthen the global content information based on the keyword information of the threat intelligence. Fine-grained contextual semantic associations are performed to achieve a deep understanding and structured representation of the threat intelligence content. In this way, unstructured threat intelligence reports can be effectively converted into structured data, improving the utilization efficiency and analysis accuracy of threat intelligence.

[0028] Based on this, Figure 2 The flowchart of the threat intelligence analysis method based on large-scale model collaboration in the embodiment of the present application is shown. For example, the threat intelligence analysis method based on large-scale model collaboration can be executed by a server, which can be Figure 1 The server 100 shown in FIG. Figure 2As shown, according to the threat intelligence analysis method based on large-scale model collaboration according to the embodiment of the present application, the steps include: S510, obtaining a dark web page; S520, extracting threat intelligence content information from the dark web page; S530, performing global content semantic encoding based on word granularity on the threat intelligence content information to obtain a set of threat intelligence content word granularity encoding vectors; S540, performing keyword extraction and semantic encoding on the threat intelligence content information to obtain a set of threat intelligence keyword context semantic encoding vectors; S550, calculating the interactive matching results of each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors relative to the set of threat intelligence keyword context semantic encoding vectors to obtain structured data of the threat intelligence content information.

[0029] Specifically, in the technical solution of the present application, first, a dark web page is obtained, and threat intelligence content information is extracted from the dark web page. It should be understood that the dark web refers to those hidden network parts on the Internet that are difficult to access through conventional search engines. Due to the concealment of the dark web, many network hacker attacks and malware transactions tend to be carried out on the dark web, which makes the dark web pages contain rich threat intelligence information. In other words, since the information in the dark web pages is highly relevant and targeted, therefore, in the technical solution of the present application, based on natural language processing technology, effective IoT threat intelligence is extracted from mixed dark web page data, which can understand the latest network attack trends, identify new threats, track hacker activity trajectories, etc., and significantly improve the timeliness and accuracy of threat intelligence.

[0030] Secondly, in order to achieve fine-grained semantic understanding of the threat intelligence content information, the threat intelligence content information is further segmented and then semantically encoded through a threat intelligence semantic encoder that includes a word embedding layer. Specifically, word embedding technology can convert each vocabulary unit in the threat intelligence content information into a dense vector in the vector space to capture the semantic information of each keyword. Furthermore, the self-attention mechanism of the threat intelligence semantic encoder is used to perform context modeling on the embedding vector of each vocabulary unit, capture the potential semantic dependencies between vocabulary units, and incorporate the semantic associations of surrounding vocabulary into the embedding vector of each vocabulary unit to obtain a set of threat intelligence content word granularity encoding vectors, thereby achieving more accurate semantic understanding and feature representation, and providing strong support for subsequent semantic analysis and association.

[0031] Correspondingly, in step S530, the threat intelligence content information is subjected to global content semantic encoding based on word granularity to obtain a set of threat intelligence content word granularity encoding vectors, including: performing word segmentation processing on the threat intelligence content information and then passing it through a threat intelligence semantic encoder including a word embedding layer to obtain a set of threat intelligence content word granularity encoding vectors.

[0032] Next, considering that in text data, keywords often carry the most core information and intentions, and are the key to understanding and analyzing text content. Therefore, in the technical solution of this application, in order to analyze threat intelligence more focusedly, a keyword extractor based on the AIGC model is introduced to extract keywords from the threat intelligence content information. Among them, the AIGC (Artificial Intelligence Generated Content) model can deeply understand the semantics of the text, identify and extract key information points in the text, quickly extract representative keywords from the threat intelligence content information, and generate a set of threat intelligence keywords, providing important clues for in-depth analysis of threat intelligence.

[0033] Then, in order to further understand the semantic information of each threat intelligence keyword and the contextual semantic association between keywords and improve the accuracy of threat intelligence analysis, the set of threat intelligence keywords is similarly input into a keyword context association encoder including a keyword embedding layer for processing. Through word embedding technology and self-attention mechanism, the deep semantic features of each keyword are mined, and the intrinsic connection between keywords is understood, thereby obtaining a set of threat intelligence keyword context semantic encoding vectors.

[0034] Accordingly, in step S540, if Figure 3 As shown, keyword extraction and semantic encoding are performed on the threat intelligence content information to obtain a set of threat intelligence keyword context semantic encoding vectors, including: S541, keyword extraction is performed on the threat intelligence content information to obtain a set of threat intelligence keywords; S542, context association encoding is performed on the set of threat intelligence keywords to obtain a set of threat intelligence keyword context semantic encoding vectors.

[0035] Among them, in step S541, keyword extraction is performed on the threat intelligence content information to obtain a set of threat intelligence keywords, including: inputting the threat intelligence content information into a keyword extractor based on the AIGC model to obtain the set of threat intelligence keywords.

[0036] Among them, in step S542, the set of threat intelligence keywords is context-associated encoded to obtain a set of threat intelligence keyword context semantic encoding vectors, including: inputting the set of threat intelligence keywords into a keyword context-associated encoder including a keyword embedding layer to obtain a set of threat intelligence keyword context semantic encoding vectors.

[0037] Next, considering that threat intelligence keywords represent key information and topics in the intelligence text, they have important guiding significance for the in-depth understanding of threat intelligence content. Therefore, in the technical solution of the present application, in order to strengthen the semantic feature expression of threat intelligence content and improve the effectiveness of feature representation, each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors is further used as a query feature vector, and the query feature vector is calculated relative to the set of threat intelligence keyword context semantic encoding vectors. The one-way attention granularity scanning interactive matching results are obtained to obtain a set of keyword information-guided threat intelligence content word granularity encoding vectors, which is used as the structured data of the threat intelligence content information. That is, the vocabulary unit in the threat intelligence content is used as a query feature, and through the one-way attention mechanism, the encoding features of each threat intelligence keyword are weighted and fused according to the semantic similarity between the query vocabulary and each threat intelligence keyword, so as to obtain a comprehensive representation of the semantic features related to the query vocabulary in the set of threat intelligence keywords, and then through its positional dot multiplication with the semantic features of the query vocabulary, the key semantic information in the query vocabulary is strengthened to obtain the keyword information-guided threat intelligence content word granularity encoding vector. In this way, the semantic correlation between each vocabulary unit and keywords in the threat intelligence content is enhanced, making the semantic expression of the threat intelligence content more precise, thereby achieving a deep understanding and structured representation of the threat intelligence content.

[0038] Correspondingly, in step S550, the interactive matching results of each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors relative to the set of threat intelligence keyword context semantic encoding vectors are calculated to obtain structured data of the threat intelligence content information, including: taking each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors as a query feature vector, and calculating the unidirectional attention granularity scanning interactive matching results of the query feature vector relative to the set of threat intelligence keyword context semantic encoding vectors to obtain a set of keyword information-guided threat intelligence content word granularity encoding vectors as the structured data of the threat intelligence content information.

[0039] Specifically, each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors is used as a query feature vector, and the one-way attention granularity scanning interactive matching result of the query feature vector relative to the set of threat intelligence keyword context semantic encoding vectors is calculated to obtain a set of keyword information-guided threat intelligence content word granularity encoding vectors as structured data of the threat intelligence content information, including: calculating the semantic measurement coefficient between the query feature vector and each threat intelligence keyword context semantic encoding vector in the set of threat intelligence keyword context semantic encoding vectors to obtain a sequence of one-way matching semantic measurement coefficients; normalizing the sequence of one-way matching semantic measurement coefficients through a softmax function to obtain a sequence of one-way matching semantic measurement weights; using the sequence of one-way matching semantic measurement weights as weights, calculating the weighted sum of the set of threat intelligence keyword context semantic encoding vectors to obtain a threat intelligence keyword context semantic reshaping representation vector; calculating the position point multiplication between the query feature vector and the threat intelligence keyword context semantic reshaping representation vector to obtain the keyword information-guided threat intelligence content word granularity encoding vector.

[0040] Among them, the semantic measurement coefficient between the query feature vector and each threat intelligence keyword context semantic coding vector in the set of threat intelligence keyword context semantic coding vectors is calculated to obtain a sequence of one-way matching semantic measurement coefficients, including: calculating the square of each eigenvalue of each threat intelligence keyword context semantic coding vector in the set of threat intelligence keyword context semantic coding vectors to obtain a set of non-linear transformation threat intelligence keyword context semantic coding vectors; cascading the query feature vector with each non-linear transformation threat intelligence keyword context semantic coding vector in the set of non-linear transformation threat intelligence keyword context semantic coding vectors to obtain a sequence of semantic association feature vectors; using a preset transformation vector to multiply each semantic association feature vector in the sequence of semantic association feature vectors to obtain a sequence of semantic similarities; adding a bias term to each semantic similarity in the sequence of semantic similarities and then passing through a sigmoid activation function to obtain a sequence of one-way matching semantic measurement coefficients.

[0041] In a specific example, each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors is used as a query feature vector, and a one-way attention granularity scanning interactive matching result of the query feature vector relative to the set of threat intelligence keyword context semantic encoding vectors is calculated to obtain a set of keyword information-guided threat intelligence content word granularity encoding vectors as structured data of the threat intelligence content information, including: processing the query feature vector and the set of threat intelligence keyword context semantic encoding vectors with the following one-way interactive matching formula to obtain the keyword information-guided threat intelligence content word granularity encoding vector, wherein the one-way interactive matching formula is: in, represents the query feature vector, The first one in the set of semantic encoding vectors of the threat intelligence keyword context Threat intelligence keyword context semantic encoding vector, Indicates cascade operation, is the preset transformation vector, is the bias parameter, express function, Represents the query feature vector and the The one-way matching semantic metric coefficient between the context semantic encoding vectors of threat intelligence keywords, is the number of semantic encoding vectors of the threat intelligence keyword context, represents the exponential function operation with e as the base, Represents the query feature vector and the The one-way matching semantic measurement weight between the context semantic encoding vectors of threat intelligence keywords, Representing threat intelligence keywords context semantics reshape representation vector, It means point multiplication by position. It represents the keyword information guiding the threat intelligence content word granularity encoding vector.

[0042] Preferably, in a preferred embodiment of the present application, the set of keyword information-guided threat intelligence content word granularity encoding vectors is optimized, including the steps of: cascading the set of keyword information-guided threat intelligence content word granularity encoding vectors into a keyword information-guided threat intelligence content word granularity encoding cascade feature vector, and cascading the set of threat intelligence content word granularity encoding vectors into a threat intelligence content word granularity encoding cascade feature vector; calculating the first norm and second norm of the mean vector of the keyword information-guided threat intelligence content word granularity encoding cascade feature vector and the threat intelligence content word granularity encoding cascade feature vector; calculating the weighted sum of the inverse of the square root of the second norm and the first norm, and performing point multiplication with the point addition sum vector of the keyword information-guided threat intelligence content word granularity encoding cascade feature vector and the threat intelligence content word granularity encoding cascade feature vector to obtain a first threat intelligence content word granularity correction subvector. ; Perform dot multiplication of the dot product vector of the keyword information guided threat intelligence content word granularity encoding cascade feature vector and the threat intelligence content word granularity encoding cascade feature vector and the square root of the length of the threat intelligence content word granularity encoding vector to obtain a second threat intelligence content word granularity correction subvector; calculate the weighted sum of the first threat intelligence content word granularity correction subvector and the second threat intelligence content word granularity correction subvector to obtain a threat intelligence content word granularity correction vector; perform dot multiplication of the threat intelligence content word granularity correction vector and the keyword information guided threat intelligence content word granularity encoding cascade feature vector to obtain a corrected keyword information guided threat intelligence content word granularity encoding cascade feature vector; and, split the corrected keyword information guided threat intelligence content word granularity encoding cascade feature vector according to the cascade mode to obtain a set of optimized keyword information guided threat intelligence content word granularity encoding vectors.

[0043] The calculation process of the threat intelligence content word granularity correction vector is expressed as follows: and They are respectively the keyword information guided threat intelligence content word granularity encoding cascade feature vector and the threat intelligence content word granularity encoding cascade feature vector, is the mean vector of the keyword information guided threat intelligence content word granularity encoding cascade feature vector and the threat intelligence content word granularity encoding cascade feature vector, represents the one-norm of the eigenvector, represents the bi-norm of the eigenvector, is the length of the threat intelligence content word granularity encoding vector, and and is the weighted sum weight as a hyperparameter, Indicates adding by position point. It means point multiplication by position. Represents the threat intelligence content word granularity correction vector.

[0044] Here, it is considered that the set of threat intelligence content word granularity encoding vectors guided by the keyword information and the set of threat intelligence content word granularity encoding vectors respectively represent the keyword text structured query semantic features of the threat intelligence keywords based on the threat intelligence content information as the retrieval corpus and the word granularity text semantic features of the threat intelligence content information. In this way, when each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors is used as a query feature vector, and the one-way attention granularity scanning interactive matching result of the query feature vector relative to the set of threat intelligence keyword context semantic encoding vectors is calculated, it is expected to further compensate for the semantic extreme deviation caused by the imbalance of the correspondence ratio under the fine-grained distribution of the sequence, and improve the accuracy of the text semantic structured encoding of the threat intelligence keywords.

[0045] Thus, based on the mean vector norm between the set of threat intelligence content word granularity encoding vectors guided by the keyword information and the set of threat intelligence content word granularity encoding vectors, the constrained representation of the structured foreground and background distinction of the superfluid of the set of threat intelligence content word granularity encoding vectors guided by the keyword information and the set of threat intelligence content word granularity encoding vectors is used to model the feature-level key correspondence between the feature vectors of the set of threat intelligence content word granularity encoding vectors guided by the keyword information and the set of threat intelligence content word granularity encoding vectors, and adjust the global association relationship between the corresponding features, so as to make positive fine-grained corresponding suggestions through the unbalanced proportion control between the corresponding feature values ​​of the set of threat intelligence content word granularity encoding vectors guided by the keyword information and the set of threat intelligence content word granularity encoding vectors, so as to avoid the extreme semantic imbalance between the feature vectors of the set of threat intelligence content word granularity encoding vectors guided by the keyword information and the set of threat intelligence content word granularity encoding vectors by focusing on the focus. In this way, the accuracy of the textual semantic structured encoding of threat intelligence keywords is improved.

[0046] Furthermore, it is worth mentioning that dark web pages contain a large amount of threat content, such as 0day vulnerability information, information about a certain intrusion incident, and leaked data sales, but these contents are included in a large number of dark web pages. By using the natural language processing method based on deep learning to establish an extraction model, by applying Doc2vec and deep clustering to the acquisition of dark web threat intelligence, information about malware, hacker techniques, and network attacks can be effectively extracted from a large amount of web page information.

[0047] Dark Web Threat Intelligence Extraction Framework Figure 4As shown in the figure. First, we need to crawl a large amount of data information from the dark web based on the characteristics of the dark web, extract features from the collected content using Doc2vec, and then apply the deep clustering algorithm that combines autoencoders and clustering to the extracted features. After clustering, the web page content is divided into each cluster according to the content type, and the text containing threat intelligence information is filtered out based on the classification results, without the need to mark a large number of collected posts.

[0048] Web page content is text data. After the data is collected, it needs to be pre-processed by tokenization, cleaning, normalization, stemming, and stop words. Natural language processing is also required to use this data as input for machine learning and extract features suitable for classification. Doc2vec is an unsupervised algorithm that can learn a fixed-length vector representation for variable-length text. It is an upgrade of Word2Vec and is used to calculate document vectors. Use Doc2vec for natural language processing and feature extraction to obtain features that reflect the meaning of the word in context.

[0049] Deep clustering combining autoencoders and clustering is applied to the extracted features. The features obtained by Doc2vec are input into the autoencoder to obtain a low-dimensional feature representation, and the encoder output is clustered and classified by k-means. The encoder output of the autoencoder is regarded as a representation of the label {0,1}k, which represents the relationship between the original text and k possible predefined classes. The clustering result is represented as y={0,1}k. The classification result is used as a pseudo-label and as supervision to update the weights of the autoencoder. Clustering is performed alternately and the autoencoder network is updated by predicting the clustering results. The features are iteratively grouped to obtain the optimal clustering result.

[0050] Furthermore, the information on the dark web is generally very up-to-date. The value of the threat intelligence information obtained from the dark web can be judged by whether it appears on the clear web and the frequency of its appearance. The spread of information takes time. After a period of time, the information with strong up-to-dateness appears on the clear web. As the up-to-dateness decreases, the amount of information related to it will gradually accumulate in the clear web. Therefore, by calculating the amount of information on the dark web in the clear web, based on the rule that the more up-to-date, the less information, it can be indirectly evaluated whether a certain dark web information has appeared in the clear web or has been paid attention to. To calculate and evaluate the amount of information on the dark web in the clear web, the search engine can be used to retrieve the results of information on the keywords that appear in the dark web data. The extracted IoC objects are searched on the clear web. The search results displayed by the search engine are arranged from high to low according to the correlation between the search results and the searched words. The higher the correlation, the higher the display position, and the lower the correlation.

[0051] Based on the above embodiments, see Figure 5As shown, it is a structural diagram of a threat intelligence analysis system 800 based on large-scale model collaboration in an embodiment of the present application. The threat intelligence analysis system 800 based on large-scale model collaboration includes: a web page acquisition module 810, used to acquire dark web pages; an information extraction module 820, used to extract threat intelligence content information from the dark web pages; a global content semantic encoding module 830, used to perform global content semantic encoding based on word granularity on the threat intelligence content information to obtain a set of threat intelligence content word granularity encoding vectors; a vector extraction module 840, used to perform keyword extraction and semantic encoding on the threat intelligence content information to obtain a set of threat intelligence keyword context semantic encoding vectors; a structured data calculation module 850, used to calculate the interactive matching results of each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors relative to the set of threat intelligence keyword context semantic encoding vectors to obtain structured data of the threat intelligence content information.

[0052] Here, those skilled in the art can understand that the specific functions and operations of each module in the above-mentioned threat intelligence analysis system 800 based on large-scale model collaboration have been referred to above. Figures 2 to 4 The description of the threat intelligence analysis method based on large-scale model collaboration has been introduced in detail, and therefore, its repeated description will be omitted.

[0053] Figure 6 FIG. 1 is an application scenario diagram of a threat intelligence analysis method based on large-scale model collaboration according to an embodiment of the present application. Figure 6 As shown, in this application scenario, first, obtain the dark web page (for example, Figure 6 Then, the dark web page is input into a server (e.g., Figure 6 In S) shown in , the server is capable of using the threat intelligence analysis algorithm based on large-scale model collaboration to process the dark web page to obtain structured data of threat intelligence content information.

[0054] Based on the above embodiments, another exemplary embodiment of an electronic device is also provided in the embodiments of the present application. In some possible implementations, the electronic device in the embodiments of the present application may include a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor can implement the steps of the threat intelligence analysis method based on large-scale model collaboration in the above embodiments when executing the program.

[0055] For example, in the case of electronic equipment Figure 1Taking the server 100 in the example as an example, the processor in the electronic device is the processor 110 in the server 100, and the memory in the electronic device is the memory 120 in the server 100.

[0056] The embodiment of the present application also provides a computer-readable storage medium, on which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, the threat intelligence analysis method based on large-scale model collaboration according to the embodiment of the present application described with reference to the above figures can be executed. The computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may, for example, include a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may, for example, include a read-only memory (ROM), a hard disk, a flash memory, etc.

[0057] The embodiment of the present application also provides a computer program product or a computer program, which includes computer executable instructions, and the computer executable instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer executable instructions from the computer-readable storage medium, and the processor executes the computer executable instructions, so that the computer device performs the threat intelligence analysis method based on large-scale model collaboration according to the embodiment of the present application.

[0058] Those skilled in the art will appreciate that the contents disclosed in this application may be subject to various modifications and improvements. For example, the various devices or components described above may be implemented by hardware, or by software, firmware, or a combination of some or all of the three.

[0059] In addition, although the present application makes various references to certain units in the system according to embodiments of the present application, any number of different units can be used and run on the client and / or server. The units are illustrative only, and different aspects of the system and method can use different units.

[0060] Those skilled in the art will appreciate that all or part of the steps in the above method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a disk or an optical disk. Optionally, all or part of the steps in the above embodiment can also be implemented using one or more integrated circuits. Accordingly, each module / unit in the above embodiment can be implemented in the form of hardware or in the form of software function modules. The present application is not limited to any particular form of combination of hardware and software.

[0061] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which this application belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an idealized or extremely formal sense, unless explicitly defined as such herein.

[0062] The above is an explanation of the present application and should not be considered as limiting thereof. Although several exemplary embodiments of the present application have been described, those skilled in the art will readily appreciate that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present application.

Claims

1. A threat intelligence analysis method based on large-scale model collaboration, It is characterized in that include: Get dark web pages; Extracting threat intelligence content information from the dark web page; Performing global content semantic encoding based on word granularity on the threat intelligence content information to obtain a set of threat intelligence content word granularity encoding vectors; Performing keyword extraction and semantic coding on the threat intelligence content information to obtain a set of threat intelligence keyword context semantic coding vectors; Calculate the interactive matching results of each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors relative to the set of threat intelligence keyword context semantic encoding vectors to obtain structured data of the threat intelligence content information.

2. According to the threat intelligence analysis method based on large-scale model collaboration according to claim 1, It is characterized in that The threat intelligence content information is subjected to global content semantic encoding based on word granularity to obtain a set of threat intelligence content word granularity encoding vectors, including: After word segmentation processing is performed on the threat intelligence content information, a threat intelligence semantic encoder including a word embedding layer is passed through to obtain a set of word granularity encoding vectors of the threat intelligence content.

3. According to the threat intelligence analysis method based on large-scale model collaboration according to claim 2, It is characterized in that The threat intelligence content information is subjected to keyword extraction and semantic coding to obtain a set of threat intelligence keyword context semantic coding vectors, including: Extracting keywords from the threat intelligence content information to obtain a set of threat intelligence keywords; The set of threat intelligence keywords is context-associated encoded to obtain a set of threat intelligence keyword context semantic encoding vectors.

4. According to the threat intelligence analysis method based on large-scale model collaboration according to claim 3, It is characterized in that Extracting keywords from the threat intelligence content information to obtain a set of threat intelligence keywords includes: The threat intelligence content information is input into a keyword extractor based on the AIGC model to obtain a set of threat intelligence keywords.

5. According to the threat intelligence analysis method based on large-scale model collaboration according to claim 4, It is characterized in that Performing context association coding on the set of threat intelligence keywords to obtain a set of threat intelligence keyword context semantic coding vectors includes: The set of threat intelligence keywords is input into a keyword context association encoder including a keyword embedding layer to obtain a set of threat intelligence keyword context semantic encoding vectors.

6. According to the threat intelligence analysis method based on large-scale model collaboration according to claim 5, It is characterized in that Calculating the interactive matching result of each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors relative to the set of threat intelligence keyword context semantic encoding vectors to obtain structured data of the threat intelligence content information, including: Taking each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors as a query feature vector, the one-way attention granularity scanning interactive matching result of the query feature vector relative to the set of threat intelligence keyword context semantic encoding vectors is calculated to obtain a set of keyword information-guided threat intelligence content word granularity encoding vectors as structured data of the threat intelligence content information.

7. According to the threat intelligence analysis method based on large-scale model collaboration according to claim 6, It is characterized in that Taking each threat intelligence content word granularity encoding vector in the set of threat intelligence content word granularity encoding vectors as a query feature vector, calculating the one-way attention granularity scanning interactive matching result of the query feature vector relative to the set of threat intelligence keyword context semantic encoding vectors to obtain a set of keyword information-guided threat intelligence content word granularity encoding vectors as structured data of the threat intelligence content information, including: Calculating semantic metric coefficients between the query feature vector and each threat intelligence keyword context semantic coding vector in the set of threat intelligence keyword context semantic coding vectors to obtain a sequence of one-way matching semantic metric coefficients; Normalizing the sequence of one-way matching semantic metric coefficients through a softmax function to obtain a sequence of one-way matching semantic metric weights; Using the sequence of the one-way matching semantic measurement weights as weights, calculating the weighted sum of the set of threat intelligence keyword context semantic encoding vectors to obtain a threat intelligence keyword context semantic reshaping representation vector; The query feature vector and the threat intelligence keyword context semantic reshaping representation vector are calculated by point-by-point multiplication to obtain the keyword information-guided threat intelligence content word granularity encoding vector.

8. According to the threat intelligence analysis method based on large-scale model collaboration according to claim 7, It is characterized in that Calculating the semantic metric coefficient between the query feature vector and each threat intelligence keyword context semantic coding vector in the set of threat intelligence keyword context semantic coding vectors to obtain a sequence of one-way matching semantic metric coefficients, including: Calculating the square of each eigenvalue of each threat intelligence keyword context semantic coding vector in the set of threat intelligence keyword context semantic coding vectors to obtain a set of non-linear transformation threat intelligence keyword context semantic coding vectors; Cascading the query feature vector and each non-linear transformation threat intelligence keyword context semantic encoding vector in the set of non-linear transformation threat intelligence keyword context semantic encoding vectors to obtain a sequence of semantic association feature vectors; Using a preset transformation vector to multiply each semantic association feature vector in the sequence of semantic association feature vectors respectively to obtain a sequence of semantic similarities; Each semantic similarity in the sequence of semantic similarities is added with a bias term and then passed through a sigmoid activation function to obtain a sequence of one-way matching semantic measurement coefficients.

Citation Information

Cited By

  • Intelligent management cloud platform based on artificial intelligence and management method

    CN119025750A

  • Artificial Intelligence-Based Intelligent Management Cloud Platform and Management Method

    CN119025750B

  • Large language model driven open source intelligence analysis system and method

    CN120430416A