Information processing method, device and equipment and readable storage medium

By encoding and detecting anomalies in vulnerability description information using multiple anomaly detection algorithms, various anomaly scores are generated and weighted and fused, which solves the problem of outliers in vulnerability description information, optimizes the vulnerability handling process, and ensures information quality.

CN119026140APending Publication Date: 2024-11-26HANGZHOU DBAPPSECURITY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411263822.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-10
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

In existing technologies, vulnerability detection rarely involves anomaly detection processing of vulnerability information, resulting in outliers in the vulnerability description information that affect information quality and hinder subsequent processing.

Method used

Multiple anomaly detection algorithms, such as semantic matching detection, anomaly classification detection, knowledge graph reasoning detection, and structured field anomaly detection, are employed to encode and detect anomalies in the descriptive text of vulnerability data, generating multiple anomaly scores. These scores are then weighted and fused to obtain a final anomaly detection result.

Benefits of technology

Effectively detect and correct outliers in vulnerability description information, optimize vulnerability handling processes, ensure information quality, and guarantee the smooth progress of subsequent processing.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119026140A_ABST
    Figure CN119026140A_ABST
Patent Text Reader

Abstract

The invention discloses an information processing method and device, equipment and a readable storage medium. The method comprises the steps that vulnerability data and description texts of basic information of the vulnerability data are acquired; performing anomaly detection on the description text by using at least two anomaly detection algorithms with different dimensions to obtain at least two anomaly scores; performing weighted fusion on the at least two abnormal scores to obtain a fusion score; and determining an abnormal value detection result of the description text of the vulnerability data by using the fusion score. According to the method, anomaly detection of different dimensions can be carried out on the vulnerability description information, the final anomaly value detection result is determined through weighted fusion, the vulnerability description information can be further optimized, and smooth proceeding of subsequent vulnerability processing is guaranteed.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of security assurance technology, and in particular to an information processing method, apparatus, device and readable storage medium. Background Art

[0002] Information system security is inseparable from effective vulnerability management. During the management process, detailed descriptions of each vulnerability need to be collected for tracking, analysis, and repair. However, there are often various abnormal problems in vulnerability descriptions, such as descriptions that do not match the affected products, syntax errors, and out-of-bounds values. These will affect the quality of information and hinder subsequent vulnerability handling.

[0003] However, for vulnerability detection, more efforts are made to optimize vulnerability detection rules or vulnerability detection models, and abnormal detection processing is rarely performed on vulnerability information.

[0004] In summary, how to effectively solve problems such as abnormal detection of vulnerability information is a technical problem that technical personnel in this field urgently need to solve. Summary of the invention

[0005] The purpose of this application is to provide an information processing method, device, equipment and readable storage medium, which, combined with multiple detection algorithms, can effectively detect abnormal values ​​in vulnerability information.

[0006] In order to solve the above technical problems, this application provides the following technical solutions:

[0007] An information processing method, comprising:

[0008] Get vulnerability data and its basic information description text;

[0009] Using at least two anomaly detection algorithms of different dimensions, perform anomaly detection on the description text to obtain at least two anomaly scores;

[0010] Performing weighted fusion on at least two anomaly scores to obtain a fusion score;

[0011] The fusion score is used to determine an outlier detection result of the description text of the vulnerability data.

[0012] Preferably, the anomaly detection algorithm includes at least two of semantic matching detection, anomaly classification detection, knowledge graph reasoning detection, structured field anomaly detection and clustering detection;

[0013] Accordingly, at least two anomaly detection algorithms of different dimensions are used to perform anomaly detection on the description text to obtain at least two anomaly scores, including:

[0014] Encode the description text of the vulnerability data to obtain a text encoding vector, perform semantic matching detection based on the text encoding vector, and obtain a first anomaly score in the case of a semantic mismatch anomaly;

[0015] Generate a variant text of the description text, encode the variant text to obtain a variant text encoding vector, and perform anomaly detection on the variant text encoding vector using an anomaly classifier mode to obtain a second anomaly score;

[0016] Performing abnormal reasoning on the description text in combination with the knowledge graph to obtain a third abnormality score;

[0017] extracting supplementary features of the description text using the large language model, and determining a fourth anomaly score based on the supplementary features;

[0018] Performing structured field anomaly detection on the supplementary features to obtain a fifth anomaly score.

[0019] Preferably, performing semantic matching detection based on the text encoding vector includes:

[0020] Encode the product name to obtain the product semantic vector;

[0021] Calculating the cosine similarity between the text encoding vector and the product semantic vector;

[0022] The cosine similarity is used to perform semantic matching detection.

[0023] Preferably, generating a variant text of the description text includes:

[0024] The description text is perturbed using a mask model to obtain the variant text.

[0025] Preferably, performing abnormal reasoning on the description text in combination with the knowledge graph to obtain a third abnormality score includes:

[0026] Obtain product knowledge graph;

[0027] Based on the specified information in the description text, link the description text to the knowledge graph;

[0028] After linking, reasoning is performed on the knowledge graph to determine missing information;

[0029] The third anomaly score is determined based on the missing information.

[0030] Preferably, performing structured field anomaly detection on the supplementary features to obtain a fifth anomaly score comprises:

[0031] Acquire structured field information from the description text;

[0032] Comparing the structured field information with the normalized representation to detect version number outliers;

[0033] In the event that an anomaly is detected, the fifth anomaly score is determined based on the anomaly.

[0034] Preferably, it also includes:

[0035] In the case where the outlier detection result indicates the presence of abnormal vulnerability information, based on the description text, combined with semantic understanding and association mining, and with reference to the knowledge graph to construct supplementary information to supplement the missing information, a complete and accurate new description text is generated.

[0036] An information processing device, comprising:

[0037] The text acquisition module is used to obtain the description text of vulnerability data and its basic information;

[0038] An anomaly detection module, used to perform anomaly detection on the description text using at least two anomaly detection algorithms of different dimensions to obtain at least two anomaly scores;

[0039] A fusion module, used for weighted fusion of at least two anomaly scores to obtain a fusion score;

[0040] The result determination module is used to determine the outlier detection result of the description text of the vulnerability data by using the fusion score.

[0041] An electronic device, comprising:

[0042] Memory for storing computer programs;

[0043] A processor is used to implement the steps of the above-mentioned information processing method when executing the computer program.

[0044] A readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned information processing method are implemented.

[0045] Apply the method provided in the embodiment of the present application to obtain a description text of vulnerability data and its basic information; use at least two anomaly detection algorithms of different dimensions to perform anomaly detection on the description text to obtain at least two anomaly scores; perform weighted fusion on the at least two anomaly scores to obtain a fusion score; and use the fusion score to determine the outlier detection result of the description text of the vulnerability data.

[0046] In order to perform outlier detection on vulnerability description information, first obtain vulnerability data and a description text of its basic information, which is the vulnerability information. Then, use at least two different flavors of anomaly detection algorithms to perform anomaly detection on the description text, thereby obtaining at least two anomaly scores. By weighted fusion of the anomaly scores, a fusion score is obtained. Finally, based on the fusion score, the outlier detection result of the vulnerability description information can be determined. That is, the present application can perform anomaly detection on vulnerability description information in different dimensions, and determine the final outlier detection result through weighted fusion, and further optimize the vulnerability description information, thereby ensuring the smooth progress of subsequent vulnerability processing.

[0047] Correspondingly, the embodiments of the present application also provide information processing devices, equipment and readable storage media corresponding to the above-mentioned information processing methods, which have the above-mentioned technical effects and are not repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 This is a flowchart of an information processing method in an embodiment of the present application;

[0050] Figure 2 Schematic diagram of various detection algorithms in the embodiments of this application;

[0051] Figure 3 This is a flowchart of an information processing method in an embodiment of the present application;

[0052] Figure 4 This is a schematic diagram of the structure of an information processing device in an embodiment of the present application;

[0053] Figure 5 This is a schematic diagram of the structure of an electronic device in an embodiment of the present application;

[0054] Figure 6 This is a schematic diagram of the specific structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0055] In order to enable those skilled in the art to better understand the present application, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0056] Please refer to Figure 1 , Figure 1 This is a flowchart of an information processing method in an embodiment of the present application, the method comprising the following steps:

[0057] S101. Obtain vulnerability data and description text of its basic information.

[0058] In this embodiment, the vulnerability data and the description text of its basic information can be obtained from the storage device, that is, the description text is the vulnerability description information of the vulnerability data.

[0059] S102: Perform anomaly detection on the description text using anomaly detection algorithms of at least two different dimensions to obtain at least two anomaly scores.

[0060] In a specific implementation of the present application, the anomaly detection algorithm includes at least two of semantic matching detection, anomaly classification detection, knowledge graph reasoning detection, structured field anomaly detection, and clustering detection;

[0061] Accordingly, anomaly detection is performed on the description text using at least two anomaly detection algorithms in different dimensions to obtain at least two anomaly scores, including:

[0062] Encode the description text of the vulnerability data to obtain a text encoding vector, perform semantic matching detection based on the text encoding vector, and obtain a first anomaly score in the case of a semantic mismatch anomaly;

[0063] Generate variant text of the description text, encode the variant text to obtain a variant text encoding vector, and use an anomaly classifier mode to perform anomaly detection on the variant text encoding vector to obtain a second anomaly score;

[0064] Combine the knowledge graph to perform abnormal reasoning on the description text and obtain the third abnormality score;

[0065] extracting supplementary features of the description text using the large language model, and determining a fourth anomaly score based on the supplementary features;

[0066] Structured field anomaly detection is performed on the supplementary features to obtain the fifth anomaly score.

[0067] Among them, semantic matching detection based on text encoding vectors includes:

[0068] Encode the product name to obtain the product semantic vector;

[0069] Calculate the cosine similarity between the text encoding vector and the product semantic vector;

[0070] Use cosine similarity to perform semantic matching detection.

[0071] Among them, the variant text of the description text is generated, including:

[0072] The mask model is used to perturb the description text to obtain variant text.

[0073] Among them, the description text is combined with the knowledge graph to perform abnormal reasoning to obtain the third abnormal score, including:

[0074] Obtain product knowledge graph;

[0075] Based on the specified information in the description text, link the description text to the knowledge graph;

[0076] After linking, reasoning is performed on the knowledge graph to determine the missing information;

[0077] Based on the missing information, a third anomaly score is determined.

[0078] Among them, the structured field anomaly detection is performed on the supplementary features to obtain the fifth anomaly score, including:

[0079] Get structured field information from the description text;

[0080] Compare the structured field information with the normalized representation to detect version number outliers;

[0081] In the event that an anomaly is detected, a fifth anomaly score is determined based on the anomaly.

[0082] In practical applications, after obtaining the description text, a preprocessing step can be performed first, that is, standard text preprocessing is performed on the description text: word segmentation, stop word removal, capitalization unification, etc.

[0083] Then, use a large model’s mask model (such as BERT’s masked language model) to perturb the input text and generate several variant texts. Next, perform feature engineering steps:

[0084] 1.1, use the encoder model based on the large model (such as ChatGPT, BERT, etc.) to encode the original description text and obtain the text encoding vector as the main feature.

[0085] 1.2, the variant text generated by the mask model of the large model is also encoded using the same encoder model as auxiliary features.

[0086] 1.3, extract additional semantic features from the middle layer of the large model encoder as supplementary features. Finally, three features are retained: the large model encoding vector of the original text (main feature), the large model encoding vector of the variant text (auxiliary feature), and the semantic features extracted from the middle layer of the large model (supplementary feature).

[0087] To facilitate outlier detection, the specific meaning of outliers is defined below.

[0088] Clearly define the abnormal situations that may appear in the basic information of the vulnerability, laying the foundation for subsequent outlier detection and correction. The main types of outliers include:

[0089] 2.1. Format anomaly: The description content does not conform to the general vulnerability description format and grammatical specifications. For example, the version range of product A and product B is not clearly distinguished in the description, resulting in parsing and confusion of version data.

[0090] 2.2. Semantic mismatch: There is an obvious mismatch between the product name and the description content, for example: a vulnerability description information does not correspond to multiple different products.

[0091] 2.3. Content inconsistency: There are logical inconsistencies in the description. For example, the vulnerability description mentions that a specific version of the product has a vulnerability, but then mentions that the vulnerability has been fixed in that version.

[0092] 2.4. Value out of range: If the affected product version number, rating data, etc. are out of a reasonable range, for example, the affected product version number is a negative number or exceeds the currently released version range.

[0093] 2.5. Error message: Clear data error, such as the CVE number does not exist, the number has been withdrawn, etc.

[0094] In order to improve the effectiveness of detection, this application designs multiple detection methods from different perspectives and integrates the detection results.

[0095] Figure 2 This is a schematic diagram of anomaly detection, which explains the components of outlier detection, including multiple algorithms based on large model semantic matching, large model anomaly classifier, knowledge graph reasoning detection, optimized deep clustering detection, and structured field anomaly detection.

[0096] Please refer to Figure 3,The figure describes the complete process of outlier detection and correction, including text preprocessing, feature engineering, multiple algorithm detection and anomaly score fusion, outlier determination, model generation correction description, and final rule filtering and manual review.

[0097] The following is a detailed description of the five algorithms:

[0098] 3.1. Semantic matching algorithm process based on large model:

[0099] a) Enter the basic information text of the vulnerability, that is, the description text.

[0100] b) Use the pre-trained BERT large model to encode the description text and obtain the description semantic vector (i.e., the main feature).

[0101] c) Use the same large model to encode the product name and obtain the product semantic vector.

[0102] d) Calculate the cosine similarity sim between the description vector and the product vector.

[0103] e) If sim < preset threshold θ: mark it as semantic mismatch anomaly and output anomaly score S1; otherwise: judge it as semantic match normal.

[0104] 3.2. Abnormal classifier process based on large model:

[0105] a) Enter the basic information text of the vulnerability, that is, the description text.

[0106] b) Use the pre-trained ChatGPT large model to encode the description text and obtain the semantic vector representation (auxiliary features).

[0107] c) Input the semantic vector into the end-to-end anomaly classifier model.

[0108] d) The anomaly classifier model directly predicts and outputs the anomaly score S2.

[0109] 3.3. Knowledge graph reasoning detection algorithm process:

[0110] a) Enter the basic information text of the vulnerability, that is, the description text.

[0111] b) Based on the pre-built product knowledge graph: link the product, version number and other information in the description text to the knowledge graph; based on graph reasoning, discover the potentially affected products and other information missing in the description.

[0112] c) For the case of missing information, output the anomaly score S3.

[0113] 3.4、Optimized deep clustering detection algorithm process:

[0114] a) Input the normal vulnerability description training set X.

[0115] b) Use the BERT large model to semantically encode X and obtain a vector representation of X (i.e., a supplementary feature).

[0116] c) In the encoding vector space, use K-Means and other algorithms for clustering to obtain k normal semantic clusters.

[0117] d) Input a new sample x', which is the text description of the vulnerability data.

[0118] e) Use the large model to encode x' and obtain a vector representation.

[0119] f), calculate the minimum distance d_min from x' to each normal cluster.

[0120] g) If d_min > threshold r, it is marked as abnormal and the abnormal score S4 is output; otherwise, it is judged as normal.

[0121] 3.5. Structured field anomaly detection algorithm flow:

[0122] a) Enter the version number and other structured field information in the vulnerability information.

[0123] b) Label pre-training capability based on the BERT large model: learn common patterns and representations of version numbers (a supplementary feature) from the training corpus and construct a standardized version number representation.

[0124] c) Normalize the input version number information.

[0125] d) Check the semantic consistency of the normalized representation with the description context.

[0126] e) Check whether there are any version number anomalies such as value out of bounds, time contradiction, etc.

[0127] f) If an anomaly is detected, the anomaly score S5 is output.

[0128] S103: Perform weighted fusion on at least two anomaly scores to obtain a fusion score.

[0129] For example, Figure 2 The anomaly scores of the five detection algorithms shown are weighted fused to obtain the fusion score.

[0130] S104: Determine an outlier detection result of the description text of the vulnerability data by using the fusion score.

[0131] If the fusion score exceeds the threshold, it is determined that there are outliers in the description text and it is sent to the subsequent correction module for processing; if the fusion score does not exceed the threshold, it is determined that there are no outliers in the description text.

[0132] Apply the method provided in the embodiment of the present application to obtain a description text of vulnerability data and its basic information; use at least two anomaly detection algorithms of different dimensions to perform anomaly detection on the description text to obtain at least two anomaly scores; perform weighted fusion on the at least two anomaly scores to obtain a fusion score; and use the fusion score to determine the outlier detection result of the description text of the vulnerability data.

[0133] In order to perform outlier detection on vulnerability description information, first obtain vulnerability data and a description text of its basic information, which is the vulnerability information. Then, use at least two different flavors of anomaly detection algorithms to perform anomaly detection on the description text, thereby obtaining at least two anomaly scores. By weighted fusion of the anomaly scores, a fusion score is obtained. Finally, based on the fusion score, the outlier detection result of the vulnerability description information can be determined. That is, the present application can perform anomaly detection on vulnerability description information in different dimensions, and determine the final outlier detection result through weighted fusion, and further optimize the vulnerability description information, thereby ensuring the smooth progress of subsequent vulnerability processing.

[0134] It should be noted that, based on the above embodiments, the embodiments of the present application also provide corresponding improved solutions. In the preferred / improved embodiments, the same steps or corresponding steps as those in the above embodiments can be referenced to each other, and the corresponding beneficial effects can also be referenced to each other, which will not be described one by one in the preferred / improved embodiments of this article.

[0135] In a specific implementation of the present application, it also includes:

[0136] In the case where the outlier detection results indicate the presence of abnormal vulnerability information, a new description text is generated based on the description text, combined with semantic understanding and association mining, and with reference to the knowledge graph to construct supplementary information that is missing, thus generating a complete and accurate new description text.

[0137] For abnormal vulnerability information detected, a correction mechanism based on a large model is designed: semantic understanding and association mining are performed based on the large model, using its text generation capabilities, and referencing the knowledge graph to build and supplement the missing product terms + version information to generate a more complete and accurate description document. For complex situations that cannot be automatically corrected, an expert manual review and correction mechanism is provided to output the final standardized description.

[0138] It can fully utilize the semantic understanding and generation capabilities of the large model to automatically correct the detected outliers, thereby outputting high-quality standardized vulnerability descriptions. At the same time, it also retains the manual review process to ensure the reliability of the system output.

[0139] Corresponding to the above method embodiment, the embodiment of the present application further provides an information processing device. The information processing device described below and the information processing method described above can refer to each other.

[0140] See also Figure 4 As shown, the device includes the following modules:

[0141] The text acquisition module 101 is used to obtain the description text of the vulnerability data and its basic information;

[0142] Anomaly detection module 102, used to perform anomaly detection on the description text using at least two anomaly detection algorithms of different dimensions to obtain at least two anomaly scores;

[0143] A fusion module 103 is used to perform weighted fusion on at least two anomaly scores to obtain a fusion score;

[0144] The result determination module 104 is used to determine the outlier detection result of the description text of the vulnerability data by using the fusion score.

[0145] By using the device provided in the embodiment of the present application, a description text of vulnerability data and its basic information is obtained; anomaly detection is performed on the description text using at least two anomaly detection algorithms of different dimensions to obtain at least two anomaly scores; weighted fusion is performed on the at least two anomaly scores to obtain a fusion score; and the fusion score is used to determine the outlier detection result of the description text of the vulnerability data.

[0146] In order to perform outlier detection on vulnerability description information, first obtain vulnerability data and a description text of its basic information, which is the vulnerability information. Then, use at least two different flavors of anomaly detection algorithms to perform anomaly detection on the description text, thereby obtaining at least two anomaly scores. By weighted fusion of the anomaly scores, a fusion score is obtained. Finally, based on the fusion score, the outlier detection result of the vulnerability description information can be determined. That is, the present application can perform anomaly detection on vulnerability description information in different dimensions, and determine the final outlier detection result through weighted fusion, and further optimize the vulnerability description information, thereby ensuring the smooth progress of subsequent vulnerability processing.

[0147] In a specific embodiment of the present application, the anomaly detection algorithm includes at least two of semantic matching detection, anomaly classification detection, knowledge graph reasoning detection, structured field anomaly detection and clustering detection;

[0148] Accordingly, the anomaly detection module 102 is specifically configured to:

[0149] Encode the description text of the vulnerability data to obtain a text encoding vector, perform semantic matching detection based on the text encoding vector, and obtain a first anomaly score in the case of a semantic mismatch anomaly;

[0150] Generate variant text of the description text, encode the variant text to obtain a variant text encoding vector, and use an anomaly classifier mode to perform anomaly detection on the variant text encoding vector to obtain a second anomaly score;

[0151] Combine the knowledge graph to perform abnormal reasoning on the description text and obtain the third abnormality score;

[0152] extracting supplementary features of the description text using the large language model, and determining a fourth anomaly score based on the supplementary features;

[0153] Structured field anomaly detection is performed on the supplementary features to obtain the fifth anomaly score.

[0154] In a specific implementation of the present application, the anomaly detection module is specifically used to:

[0155] Encode the product name to obtain the product semantic vector;

[0156] Calculate the cosine similarity between the text encoding vector and the product semantic vector;

[0157] Use cosine similarity to perform semantic matching detection.

[0158] In a specific implementation of the present application, the anomaly detection module is specifically used to: use a mask model to perturb the description text to obtain a variant text.

[0159] In a specific implementation of the present application, the anomaly detection module is specifically used to: obtain a product knowledge graph;

[0160] Based on the specified information in the description text, link the description text to the knowledge graph;

[0161] After linking, reasoning is performed on the knowledge graph to determine the missing information;

[0162] Based on the missing information, a third anomaly score is determined.

[0163] In a specific implementation of the present application, the anomaly detection module is specifically used to: obtain structured field information from the description text;

[0164] Compare the structured field information with the normalized representation to detect version number outliers;

[0165] In the event that an anomaly is detected, a fifth anomaly score is determined based on the anomaly.

[0166] In a specific implementation of the present application, it also includes:

[0167] The exception handling module is used to generate a complete and accurate new description text based on the description text, combined with semantic understanding and association mining, and referring to the knowledge graph to construct supplementary information to supplement the missing information when the outlier detection results indicate the existence of abnormal vulnerability information.

[0168] Corresponding to the above method embodiment, an embodiment of the present application further provides an electronic device. The electronic device described below and the information processing method described above can refer to each other.

[0169] See also Figure 5 As shown, the electronic device includes:

[0170] A memory 332, for storing computer programs;

[0171] The processor 322 is used to implement the steps of the information processing method of the above method embodiment when executing a computer program.

[0172] For details, please refer to Figure 6 , Figure 6 A schematic diagram of the specific structure of an electronic device provided for this embodiment, which may have relatively large differences due to different configurations or performances, may include one or more processors (central processing units, CPU) (for example, one or more processors) and a memory 332, and the memory 332 stores one or more computer programs 342 or data 344. Among them, the memory 332 can be a temporary storage or a permanent storage. The program stored in the memory 332 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the data processing device. Furthermore, the processor 322 can be configured to communicate with the memory 332 to execute a series of instruction operations in the memory 332 on the electronic device 301.

[0173] The electronic device 301 may further include one or more power supplies 326 , one or more wired or wireless network interfaces 350 , one or more input and output interfaces 358 , and / or one or more operating systems 341 .

[0174] The steps in the information processing method described above can be implemented by the structure of an electronic device.

[0175] Corresponding to the above method embodiment, the embodiment of the present application further provides a readable storage medium. The readable storage medium described below and the information processing method described above can refer to each other.

[0176] A readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the information processing method of the above method embodiment are implemented.

[0177] The readable storage medium may specifically be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or other readable storage medium that can store program codes.

[0178] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0179] Those skilled in the art may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0180] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0181] Finally, it should be noted that, in this article, relationships such as first and second, etc. are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms include, include or any other variations are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0182] Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for those skilled in the art, according to the idea of ​​the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. An information processing method, characterized in that: include: Get vulnerability data and its basic information description text; Using at least two anomaly detection algorithms of different dimensions, perform anomaly detection on the description text to obtain at least two anomaly scores; Performing weighted fusion on at least two anomaly scores to obtain a fusion score; The fusion score is used to determine an outlier detection result of the description text of the vulnerability data.

2. The method according to claim 1, characterized in that The anomaly detection algorithm includes at least two of semantic matching detection, anomaly classification detection, knowledge graph reasoning detection, structured field anomaly detection and clustering detection; Accordingly, at least two anomaly detection algorithms of different dimensions are used to perform anomaly detection on the description text to obtain at least two anomaly scores, including: Encode the description text of the vulnerability data to obtain a text encoding vector, perform semantic matching detection based on the text encoding vector, and obtain a first anomaly score in the case of a semantic mismatch anomaly; Generate a variant text of the description text, encode the variant text to obtain a variant text encoding vector, and perform anomaly detection on the variant text encoding vector using an anomaly classifier mode to obtain a second anomaly score; Performing abnormal reasoning on the description text in combination with the knowledge graph to obtain a third abnormality score; extracting supplementary features of the description text using the large language model, and determining a fourth anomaly score based on the supplementary features; Performing structured field anomaly detection on the supplementary features to obtain a fifth anomaly score.

3. The method according to claim 2, characterized in that Performing semantic matching detection based on the text encoding vector includes: Encode the product name to obtain the product semantic vector; Calculating the cosine similarity between the text encoding vector and the product semantic vector; The cosine similarity is used to perform semantic matching detection.

4. The method according to claim 2, characterized in that: Generate variant text of the description text, including: The description text is perturbed using a mask model to obtain the variant text.

5. The method according to claim 2, characterized in that: The description text is subjected to abnormal reasoning in combination with the knowledge graph to obtain a third abnormality score, including: Obtain product knowledge graph; Based on the specified information in the description text, link the description text to the knowledge graph; After linking, reasoning is performed on the knowledge graph to determine missing information; The third anomaly score is determined based on the missing information.

6. The method according to claim 2, characterized in that Performing structured field anomaly detection on the supplementary features to obtain a fifth anomaly score includes: Acquire structured field information from the description text; Comparing the structured field information with the normalized representation to detect version number outliers; In the event that an anomaly is detected, the fifth anomaly score is determined based on the anomaly.

7. The method according to any one of claims 1 to 6, characterized in that: Also includes: In the case where the outlier detection result indicates the presence of abnormal vulnerability information, based on the description text, combined with semantic understanding and association mining, and with reference to the knowledge graph to construct supplementary information to supplement the missing information, a complete and accurate new description text is generated.

8. An information processing device, characterized in that: include: The text acquisition module is used to obtain the description text of vulnerability data and its basic information; An anomaly detection module, used to perform anomaly detection on the description text using at least two anomaly detection algorithms of different dimensions to obtain at least two anomaly scores; A fusion module, used for weighted fusion of at least two anomaly scores to obtain a fusion score; A result determination module is used to determine an outlier detection result of the description text of the vulnerability data by using the fusion score.

9. An electronic device, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the steps of the information processing method according to any one of claims 1 to 7 when executing the computer program.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the information processing method according to any one of claims 1 to 7 are implemented.