Intelligent detection method and system for fragmented sensitive information

Abstract

The invention provides a fragmented sensitive information intelligent detection method and system, and belongs to the technical field of information detection. The method comprises the following steps: creating a sensitive information knowledge base and a retrieval model; to-be-detected fragmented information is obtained from the multi-source document, and a to-be-detected text is generated; identifying sensitive information in the to-be-detected text by using a large language model; based on the sensitive information, utilizing an NLP technology to construct a knowledge graph; performing community division on the knowledge graph by using a graph algorithm, and generating summary information of communities; inputting the summary information into a retrieval model, and matching the retrieval model with entries in a sensitive information knowledge base to retrieve potential sensitive information research and judgment rules related to the summary information; performing context fusion on the potential sensitive information of the community and the potential sensitive information research and judgment rule to generate a sensitive information research and judgment prompt text; and generating an evaluation report about the sensitive information by using a large language model. According to the invention, efficient and accurate detection of fragmented sensitive information is realized.

Description

Technical Field

[0001] The present invention relates to the field of information detection technology, and more specifically to a method and system for intelligently detecting fragmented sensitive information. Background Art

[0002] With the rapid development of information technology, the explosive growth of data volume has made the protection of sensitive information particularly important. However, existing sensitive information detection systems often have problems such as low detection accuracy and difficulty in processing fragmented information. Especially when dealing with complex scenarios such as cross-media and cross-platform, the detection and identification of sensitive information becomes particularly difficult.

[0003] Existing sensitive information detection systems use rules-based and simple model binary classification. Such systems rely on predefined rules and keyword lists and require a large amount of high-quality sensitive data to train models to identify sensitive information in documents or data. Although these systems can effectively detect some sensitive information in specific scenarios, they have the following obvious limitations: 1. When performing information retrieval, traditional methods such as rule-based keyword matching are used. Such methods cannot understand the context and may misjudge harmless sentences as sensitive information. Simple variant processing, such as replacing letters and adding symbols, can bypass keyword matching and reduce the effectiveness of the detection system. In addition, it is difficult to maintain rules and keyword lists, making it difficult to cope with new types of sensitive information.

[0004] 2. It is not effective in detecting fragmented sensitive information, especially when the sensitive information is distributed in multiple documents or data sets.

[0005] 3. The detection accuracy is limited, which is prone to false positives and missed positives.

[0006] It can be seen that how to efficiently and accurately detect fragmented sensitive information has become a problem that needs to be solved urgently. Summary of the invention

[0007] In view of the above problems, the purpose of the present invention is to provide a method and system for intelligent detection of fragmented sensitive information, which utilizes the retrieval enhancement capability of a large model and combines the correlation analysis of a knowledge graph to achieve efficient and accurate detection of fragmented sensitive information.

[0008] In order to achieve the above object, the present invention is implemented through the following technical solutions: In a first aspect, the present invention discloses a method for intelligently detecting fragmented sensitive information, comprising: Define sensitive information and create a sensitive information knowledge base; Use deep learning models to build retrieval models; Obtain fragmented information to be detected from multi-source documents, perform data cleaning on the fragmented information, and generate text to be detected; Use a large language model to identify sensitive information in the text to be detected; Based on sensitive information, use NLP technology to build a knowledge graph; Use graph algorithms to divide knowledge graphs into communities and generate summary information for communities containing sensitive information; Use graph algorithms to divide the knowledge graph into communities and generate summary information for each community; Input the summary information into the retrieval model, and use the retrieval model to match the entries in the sensitive information knowledge base to retrieve the potential sensitive information in the summary information and the corresponding sensitive information analysis rules; Based on the text to be detected, the potential sensitive information and the corresponding sensitive information analysis rules are context-integrated to generate sensitive information analysis prompt text; Based on the sensitive information analysis prompt text, a large language model is used to generate an assessment report on sensitive information. Further, the definition of sensitive information and the creation of a sensitive information knowledge base include: Define personally identifiable information, financial information, health information, and geolocation data as sensitive information; Create a dataset or database containing all defined sensitive information as a sensitive information knowledge base; The sensitive information knowledge base includes: instances of sensitive information, categories of sensitive information and contextual descriptions of sensitive information.

[0009] Further, the use of a deep learning model to construct a retrieval model includes: Obtain text data from the sensitive information knowledge base and pre-process the text data; Use word embedding technology to convert each word in the text data into a fixed-length vector; Use sentence vector models to convert sentences or phrases in text data into vectors; Select a deep learning model based on retrieval requirements, train the deep learning model using vectors converted from text data, and generate a retrieval model.

[0010] Further, the use of a large language model to identify sensitive information in the text to be detected includes: Use natural language processing technology to extract meaningful features from the text to be tested; Use large language models to identify potentially sensitive information patterns in the text to be tested; Define rules for sensitive information, match the extracted features and identified potential sensitive information patterns with the rules for sensitive information to determine whether the text to be detected contains sensitive information; If the match is successful, the corresponding text will be marked as sensitive information and recorded as text to be identified. Furthermore, the knowledge graph is constructed based on sensitive information using NLP technology, including: Use NLP technology to identify entities in the text to be identified; Extract the relationship between entities from the text to be recognized; Based on entities and the relationships between entities, a knowledge graph containing entities and relationships is constructed.

[0011] Furthermore, the method uses a graph algorithm to divide the knowledge graph into communities and generates summary information for each community, including: Use the Leiden algorithm to divide the knowledge graph into multiple communities based on different topics or contexts; Extract key information and sensitive points from each community; Generate summary information based on the extracted key information and sensitive points.

[0012] Furthermore, the summary information is input into the retrieval model, and the retrieval model is used to match the entries in the sensitive information knowledge base to retrieve the potential sensitive information in the summary information and the corresponding sensitive information analysis rules, including: Input the summary information into the retrieval model, convert the summary information into a vector representation, and generate a summary information vector; The retrieval model is used to calculate the similarity between the summary information vector and each entry vector in the sensitive information knowledge base; Based on the preset similarity threshold, entries with similarities higher than the similarity threshold are determined, the corresponding entries are regarded as potential sensitive information, and the corresponding sensitive information analysis rules are extracted.

[0013] Furthermore, based on the text to be detected, context fusion processing is performed on the potential sensitive information and the corresponding sensitive information analysis rules to generate a sensitive information analysis prompt text, including: According to the sensitive information analysis rules, the potential sensitive information is integrated with the text to be detected to form a context environment containing complete information and generate sensitive information analysis prompt text; Based on the generated sensitive information analysis prompt text, the existence and sensitivity of sensitive information in the text to be detected are evaluated.

[0014] Furthermore, the sensitive information assessment prompt text is used to generate an assessment report on sensitive information using a large language model, including: Based on the sensitive information analysis prompt text, use the large language model to identify the type of sensitive information, mark the location of sensitive information, and encrypt or desensitize it; A large language model is used to assign a sensitivity score to each entity or relationship, evaluate the risk level of sensitive information, and generate processing suggestions for sensitive information to form an assessment report for sensitive information.

[0015] In a second aspect, the present invention further discloses a fragmented sensitive information intelligent detection system, comprising: A knowledge base creation module is used to define sensitive information and create a sensitive information knowledge base; A retrieval model building module is used to build a retrieval model using a deep learning model; The data preparation module is used to obtain fragmented information to be detected from multi-source documents, clean the fragmented information, and generate text to be detected; A sensitive information identification module is used to identify sensitive information in the text to be detected using a large language model; The knowledge graph construction module is used to build a knowledge graph based on sensitive information using NLP technology; The community detection module is used to divide the knowledge graph into communities using graph algorithms and generate summary information for each community; A retrieval matching module is used to input the summary information into the retrieval model, and use the retrieval model to match the entries in the sensitive information knowledge base to retrieve the potential sensitive information in the summary information and the corresponding sensitive information analysis rules; A fusion module is used to perform context fusion processing on potential sensitive information and corresponding sensitive information analysis rules based on the text to be detected, and generate sensitive information analysis prompt text; The response generation module is used to analyze prompt text based on sensitive information and use a large language model to generate an assessment report on sensitive information.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention uses advanced large-scale models to enhance the understanding of context. Through contextual learning, the large model can understand sensitive information in complex contexts and identify potential sensitive information fragments, thereby improving detection accuracy.

[0017] 2. The present invention integrates a knowledge graph containing entities, relationships, and attributes related to sensitive information. This structured information helps to enhance the model's understanding and detection accuracy. The knowledge graph is crucial to providing a comprehensive framework to understand the relationships between different pieces of sensitive information, even if this information is scattered across multiple documents or data sets, and it is still possible to mine implicit associations between information. The entity relationships and attribute information stored in the knowledge graph can assist the model in identifying and understanding potential sensitive information, and improve the comprehensiveness and accuracy of detection.

[0018] 3. The present invention can analyze fragmented information scattered in different documents or data sets, link entities and relationships through semantic matching, and ensure comprehensive and accurate detection by using information aggregation methods.

[0019] 4. The present invention integrates large-scale model decision-making and fragmented information integration technology to perform intelligent detection, judge the sensitivity and level of information, and can effectively reduce false alarms and missed alarms.

[0020] It can be seen that compared with the prior art, the present invention has outstanding substantive features and significant progress, and the beneficial effects of its implementation are also obvious. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0022] Figure 1 It is a method flow chart of a specific implementation mode of the present invention. Figure 2 It is a system structure diagram of a specific implementation mode of the present invention.

[0023] In the figure, 1. Knowledge base creation module; 2. Retrieval model construction module; 3. Data preparation module; 4. Sensitive information identification module; 5. Knowledge graph construction module; 6. Community detection module; 7. Retrieval matching module; 8. Fusion module; 9. Response generation module. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the scheme of the present invention, the present invention 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 invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0025] See also Figure 1 As shown, this embodiment provides a fragmented sensitive information intelligent detection method, including the following steps: S1: Define sensitive information and create a sensitive information knowledge base.

[0026] In a specific implementation, personal identity information (such as name, ID number), financial information (such as bank account, credit card number), health information, and geographic location data are first defined as sensitive information. The definition can be specific to keywords, phrases, or pattern matching rules.

[0027] Then, a dataset or database containing all defined sensitive information is created as a sensitive information knowledge base.

[0028] Specifically, the sensitive information knowledge base includes: Specific instances of sensitive information (for example, "John Doe", "1234567890123456").

[0029] The category of sensitive information (e.g., “Personal Information,” “Financial Information”).

[0030] A contextual description of the sensitive information (for example, "a sentence containing a full name and date of birth").

[0031] S2: Build a retrieval model using a deep learning model.

[0032] In a specific implementation, text data in a sensitive information knowledge base is first obtained, and the text data is preprocessed, including steps such as noise removal, word segmentation, and stop word removal, to ensure the accuracy of subsequent processing.

[0033] Then, use word embedding technology (such as Word2Vec, GloVe) to convert each word in the text data into a fixed-length vector. And use sentence vector models such as (Sentence-BERT) to convert sentences or phrases in the text data into vectors.

[0034] Finally, a deep learning model is selected according to the retrieval requirements, and the deep learning model is trained using vectors based on text data conversion to generate a retrieval model. Among them, the deep learning model can select an embedding model to achieve more complex context understanding.

[0035] S3: Obtain fragmented information to be detected from multi-source documents, perform data cleaning on the fragmented information, and generate text to be detected.

[0036] In a specific implementation, the fragmented information to be detected is collected from multi-source documents through various sources (such as emails, social media, documents, databases, etc.) Then, data cleaning is performed on the fragmented information to be detected to remove irrelevant information and noise data and ensure data quality.

[0037] S4: Use a large language model to identify sensitive information in the text to be detected.

[0038] In a specific implementation, first, natural language processing technology is used to extract meaningful features from the text to be detected. The natural language processing technology can use word embedding technology and named entity recognition technology, and its purpose is to extract meaningful features from the text.

[0039] Then, the large language model is used to identify potential sensitive information patterns in the text to be detected. In this method, the large language model can be a Llama model or a Qwen model. It should be noted that a fine-tuned domain small model or fine-tuning directly on the large language model can also be used to identify potential sensitive information patterns in the text.

[0040] At this point, define rules for sensitive information, match the extracted features and identified potential sensitive information patterns with the rules for sensitive information to determine whether the text to be detected contains sensitive information. The rules for sensitive information can be defined according to industry standards or legal requirements, such as personally identifiable information (PII), health records, financial data, etc.

[0041] If the match is successful, the corresponding text will be marked as sensitive information and recorded as text to be identified.

[0042] S5: Based on sensitive information, use NLP technology to build a knowledge graph.

[0043] In a specific implementation, first, NLP technology is used to identify entities in the text to be identified, such as names, organizations, dates, etc. Then, the relationships between entities are extracted from the text to be identified, such as "belongs to", "works at", etc. Based on the entities and the relationships between them, a knowledge graph containing entities and relationships is constructed, which helps to understand the connections between fragmented information.

[0044] S6: Use graph algorithms to divide the knowledge graph into communities and generate summary information for each community.

[0045] In a specific implementation, the Leiden algorithm is used to divide the knowledge graph into multiple communities based on different topics or contexts; each community may represent a topic or context. Then, key information and sensitive points are extracted from each community. Finally, summary information is generated based on the extracted key information and sensitive points to be sent to the big voice model for review.

[0046] S7: Input the summary information into the retrieval model, and use the retrieval model to match the entries in the sensitive information knowledge base to retrieve the potential sensitive information in the summary information and the corresponding sensitive information analysis rules.

[0047] In a specific implementation, the summary information is first input into the retrieval model, the summary information is converted into a vector representation, and a summary information vector is generated. At this time, the retrieval model is used to calculate the similarity between the summary information vector and each entry vector in the sensitive information knowledge base. Among them, cosine similarity, Euclidean distance, etc. can be used to calculate the similarity between the input text vector and the entry in the sensitive information database.

[0048] Finally, based on the preset similarity threshold, the entries with similarity higher than the similarity threshold are determined, the corresponding entries are regarded as potential sensitive information, and the corresponding sensitive information analysis rules are extracted. It should be noted that the preset similarity threshold is used to determine which matching items are considered relevant. Matching items higher than the threshold are regarded as potential sensitive information.

[0049] S8: Based on the text to be detected, context fusion processing is performed on the potential sensitive information and the corresponding sensitive information analysis rules to generate a sensitive information analysis prompt text.

[0050] In a specific implementation, the potential sensitive information is integrated with the text to be detected to form a context environment containing complete information, and a sensitive information analysis prompt text is generated; the sensitive information analysis prompt text can ensure that sufficient context information is retained for the next step of analysis.

[0051] In addition, based on the sensitive information analysis prompt text, the existence and sensitivity of sensitive information in the text to be detected are evaluated.

[0052] S9: Based on the sensitive information analysis prompt text, use the large language model to generate an assessment report on sensitive information.

[0053] In a specific implementation, based on the sensitive information analysis prompt text, a large language model is used to identify the type of sensitive information, mark the location of sensitive information, and encrypt or desensitize it; at the same time, a large language model is used to assign a sensitivity score to each entity or relationship, evaluate the risk level of sensitive information, and generate processing suggestions for sensitive information to form an assessment report on sensitive information.

[0054] As an example, based on the integrated sensitive information analysis prompt text, a large language model is used to generate an assessment report or suggestion on sensitive information. The generated content may include: Confirm the existence of sensitive information: Indicate what types of sensitive information exist.

[0055] Sensitivity Scoring: Use a large language model to assign a sensitivity score to each entity or relation.

[0056] Mark sensitive information: Mark the location of sensitive information and encrypt or desensitize it.

[0057] Make suggestions: Provide suggestions on how to handle this sensitive information, such as deletion, replacement, encryption, etc.

[0058] Risk assessment: Assess the risk level of leaking this sensitive information.

[0059] It can be seen that the present invention provides a fragmented sensitive information intelligent detection method, which can use the contextual understanding of large model retrieval enhancement technology to more accurately identify sensitive information. At the same time, the use of knowledge graph technology can mine the implicit associations between information and improve the comprehensiveness and accuracy of detection, especially in processing complex and fragmented sensitive information in cross-media, cross-platform and other scenarios.

[0060] See also Figure 2 As shown, the present invention also discloses a fragmented sensitive information intelligent detection system, including: a knowledge base creation module 1, a retrieval model construction module 2, a data preparation module 3, a sensitive information identification module 4, a knowledge graph construction module 5, a community detection module 6, a retrieval matching module 7, a fusion module 8 and a response generation module 9. The knowledge base creation module 1 is used to define sensitive information and create a sensitive information knowledge base.

[0061] The retrieval model building module 2 is used to build a retrieval model using a deep learning model.

[0062] The data preparation module 3 is used to obtain fragmented information to be detected from multi-source documents, and perform data cleaning on the fragmented information to generate the text to be detected.

[0063] The sensitive information identification module 4 is used to use a large language model to identify sensitive information in the text to be detected.

[0064] The knowledge graph construction module 5 is used to construct a knowledge graph based on sensitive information using NLP technology.

[0065] The community detection module 6 is used to divide the knowledge graph into communities using a graph algorithm and generate summary information for each community.

[0066] The retrieval matching module 7 is used to input the summary information into the retrieval model, and use the retrieval model to match the entries in the sensitive information knowledge base to retrieve the potential sensitive information in the summary information and the corresponding sensitive information analysis rules.

[0067] The fusion module 8 is used to perform context fusion processing on potential sensitive information and corresponding sensitive information analysis rules based on the text to be detected, and generate sensitive information analysis prompt text.

[0068] The response generation module 9 is used to analyze the prompt text based on the sensitive information and use the large language model to generate an assessment report on the sensitive information.

[0069] The specific implementation of the fragmented sensitive information intelligent detection system of this embodiment is basically the same as the specific implementation of the above-mentioned fragmented sensitive information intelligent detection method, and will not be repeated here.

[0070] In summary, the present invention utilizes the retrieval enhancement capability of the large model and combines it with the correlation analysis of the knowledge graph to achieve efficient and accurate detection of fragmented sensitive information.

[0071] 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. As for the method disclosed in the embodiment, since it corresponds to the system disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0072] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in 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 the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians 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 the present invention.

[0073] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of systems or units, which can be electrical, mechanical or other forms.

[0074] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0075] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, or each module may exist physically separately, or two or more modules may be integrated into one unit.

[0076] Similarly, each processing unit in each embodiment of the present invention may be integrated into one functional module, or each processing unit may exist physically, or two or more processing units may be integrated into one functional module.

[0077] 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.

[0078] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used 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", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including 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. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0079] The above is a detailed introduction to the intelligent detection method and system for fragmented sensitive information provided by the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present invention, several improvements and modifications can be made to the present invention, and these improvements and modifications also fall within the scope of protection of the present invention.

Claims

1. A method for intelligently detecting fragmented sensitive information, characterized in that: include: Define sensitive information and create a sensitive information knowledge base; Use deep learning models to build retrieval models; Obtain fragmented information to be detected from multi-source documents, perform data cleaning on the fragmented information, and generate text to be detected; Use a large language model to identify sensitive information in the text to be detected; Based on sensitive information, use NLP technology to build a knowledge graph; Use graph algorithms to divide the knowledge graph into communities and generate summary information for each community; Input the summary information into the retrieval model, and use the retrieval model to match the entries in the sensitive information knowledge base to retrieve the potential sensitive information in the summary information and the corresponding sensitive information analysis rules; Based on the text to be detected, the potential sensitive information and the corresponding sensitive information analysis rules are context-integrated to generate sensitive information analysis prompt text; Based on the sensitive information analysis prompt text, a large language model is used to generate an assessment report on sensitive information.

2. The intelligent detection method for fragmented sensitive information according to claim 1, characterized in that: Defining sensitive information and creating a sensitive information knowledge base includes: Define personally identifiable information, financial information, health information, and geolocation data as sensitive information; Create a dataset or database containing all defined sensitive information as a sensitive information knowledge base; The sensitive information knowledge base includes: instances of sensitive information, categories of sensitive information and contextual descriptions of sensitive information.

3. The intelligent detection method for fragmented sensitive information according to claim 2, characterized in that: The method of using a deep learning model to construct a retrieval model includes: Obtain text data from the sensitive information knowledge base and pre-process the text data; Use word embedding technology to convert each word in the text data into a fixed-length vector; Use sentence vector models to convert sentences or phrases in text data into vectors; Select a deep learning model based on retrieval requirements, train the deep learning model using vectors converted from text data, and generate a retrieval model.

4. The intelligent detection method for fragmented sensitive information according to claim 3 is characterized in that: The method of using a large language model to identify sensitive information in the text to be detected includes: Use natural language processing technology to extract meaningful features from the text to be tested; Use large language models to identify potentially sensitive information patterns in the text to be tested; Define rules for sensitive information, match the extracted features and identified potential sensitive information patterns with the rules for sensitive information to determine whether the text to be detected contains sensitive information; If the match is successful, the corresponding text will be marked as sensitive information and recorded as text to be identified.

5. The intelligent detection method for fragmented sensitive information according to claim 4, characterized in that: The knowledge graph is constructed based on sensitive information using NLP technology, including: Use NLP technology to identify entities in the text to be identified; Extract the relationship between entities from the text to be recognized; Based on entities and the relationships between entities, a knowledge graph containing entities and relationships is constructed.

6. The intelligent detection method for fragmented sensitive information according to claim 5, characterized in that: The graph algorithm is used to divide the knowledge graph into communities and generate summary information for each community, including: Use the Leiden algorithm to divide the knowledge graph into multiple communities based on different topics or contexts; Extract key information and sensitive points from each community; Generate summary information based on the extracted key information and sensitive points.

7. The intelligent detection method for fragmented sensitive information according to claim 6, characterized in that: The summary information is input into the retrieval model, and the retrieval model is used to match the entries in the sensitive information knowledge base to retrieve the potential sensitive information in the summary information and the corresponding sensitive information analysis rules, including: Input the summary information into the retrieval model, convert the summary information into a vector representation, and generate a summary information vector; The retrieval model is used to calculate the similarity between the summary information vector and each entry vector in the sensitive information knowledge base; Based on the preset similarity threshold, entries with similarities higher than the similarity threshold are determined, the corresponding entries are regarded as potential sensitive information, and the corresponding sensitive information analysis rules are extracted.

8. The intelligent detection method for fragmented sensitive information according to claim 7, characterized in that: Based on the text to be detected, the potential sensitive information and the corresponding sensitive information analysis rules are subjected to context fusion processing to generate a sensitive information analysis prompt text, including: According to the sensitive information analysis rules, the potential sensitive information is integrated with the text to be detected to form a context environment containing complete information and generate sensitive information analysis prompt text; Based on the sensitive information analysis prompt text, the existence and sensitivity of sensitive information in the text to be detected are evaluated.

9. The intelligent detection method for fragmented sensitive information according to claim 8, characterized in that: The sensitive information-based judgment prompt text uses a large language model to generate an assessment report on sensitive information, including: Based on the sensitive information analysis prompt text, use the large language model to identify the type of sensitive information, mark the location of sensitive information, and encrypt or desensitize it; A large language model is used to assign a sensitivity score to each entity or relationship, evaluate the risk level of sensitive information, and generate processing suggestions for sensitive information to form an assessment report for sensitive information.

10. A fragmented sensitive information intelligent detection system, characterized in that: include: A knowledge base creation module is used to define sensitive information and create a sensitive information knowledge base; A retrieval model building module is used to build a retrieval model using a deep learning model; The data preparation module is used to obtain fragmented information to be detected from multi-source documents, clean the fragmented information, and generate text to be detected; A sensitive information identification module is used to identify sensitive information in the text to be detected using a large language model; The knowledge graph construction module is used to build a knowledge graph based on sensitive information using NLP technology; The community detection module is used to divide the knowledge graph into communities using graph algorithms and generate summary information for each community; A retrieval matching module is used to input the summary information into the retrieval model, and use the retrieval model to match the entries in the sensitive information knowledge base to retrieve the potential sensitive information in the summary information and the corresponding sensitive information analysis rules; A fusion module is used to perform context fusion processing on potential sensitive information and corresponding sensitive information analysis rules based on the text to be detected, and generate sensitive information analysis prompt text; The response generation module is used to analyze prompt text based on sensitive information and use a large language model to generate an assessment report on sensitive information.

Images