Sensitive information identification method and system for archive auditing and opening

Abstract

The invention discloses a sensitive information identification method and system for archive audit opening, and the method comprises the steps: obtaining archive data, and recognizing the archive data through a multi-level recognition model which comprises a first module, a second module and a third module; the first module is used for identifying first sensitive information in the archive data, the second module is used for identifying second sensitive information which is different from the first sensitive information but similar to the first sensitive information in content, and the third module is used for identifying third sensitive information implied in the archive data; and analyzing and processing by combining the first module, the second module and the third module to determine whether the file can be an open file. According to the invention, the multi-level identification model comprising the first module, the second module and the third module is constructed, so that comprehensive, accurate and deep identification of sensitive information in archive data is realized, more comprehensive auditing suggestions and more detailed sensitive information identification can be provided, and the accuracy and reliability of archive auditing are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of archive information data processing, and in particular to a sensitive information identification method and system for archive review and opening. Background Art

[0002] As an indispensable part of information management for organizations and institutions, archives management carries the important mission of recording history, inheriting culture and ensuring information security. With the rapid advancement of information technology, traditional paper archives management is gradually transforming to electronic archives management. This transformation has greatly improved the efficiency of archive storage, the convenience of retrieval and the extensiveness of information sharing. Electronic archives not only save valuable physical space, but also effectively avoid the risk of damage or loss of paper archives due to environmental factors through digital means. However, while enjoying the convenience brought by information technology, the security and privacy protection of electronic archives have become increasingly prominent, especially the protection of sensitive information, which has become a key issue that needs to be urgently addressed in the review of archive opening. Ensuring that archives can meet the public's information needs while strictly guarding the boundaries of information security and privacy during the opening process is an important challenge facing the current archives management field.

[0003] Although there has been a certain amount of technical accumulation and practical exploration in the field of archives open audit, there are still many challenges. First, existing technologies often lack in-depth context analysis capabilities, which makes it difficult to handle polysemous words and sensitive information related to context, and it is easy to misjudge or miss, affecting the accuracy and efficiency of the audit. Secondly, the diversity of archive content makes the types of sensitive information numerous and complex. Simple keyword matching methods are difficult to fully cover and accurately identify various forms of sensitive information, and it is difficult to meet the refined needs of archives open audit. In addition, although the application of advanced technologies such as deep learning in the field of archives audit has broad prospects, it is limited by the scarcity of high-quality annotated data, and model training is often insufficient, making it difficult to fully realize its potential, thus affecting the accuracy and reliability of the audit. Therefore, how to improve the intelligence level and adaptability of archives open audit while ensuring information security is a technical problem that needs to be solved urgently. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In view of the above existing problems, the present invention is proposed. Therefore, the present invention provides a sensitive information identification method for archive review and opening to solve the problem that sensitive information identification in archive review and opening is incomplete, inaccurate and difficult to adapt to diversified information needs.

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

[0007] In a first aspect, the present invention provides a method for identifying sensitive information for file review and opening, comprising:

[0008] Acquire archival data, and identify the archival data using a multi-level recognition model, wherein the multi-level recognition model includes a first module, a second module, and a third module;

[0009] Using the first module to identify first sensitive information in the archival data, using the second module to identify second sensitive information that is different from the first sensitive information but has similar content, and using the third module to identify third sensitive information implicit in the archival data;

[0010] Combine the analysis and processing of the first module, the second module and the third module to determine whether the file can be an open file.

[0011] As a preferred solution of the sensitive information identification method for archive review and opening of the present invention, wherein: using the first module to identify the first sensitive information in the archive data includes:

[0012] Using intelligent algorithms to identify sensitive words in the archive data, accumulating the weights corresponding to each sensitive word to obtain a total weighted score;

[0013] If the total weighted score exceeds a first threshold, the file is marked as a sensitive file;

[0014] If the total weighted score does not exceed the first threshold or no sensitive word is identified, the archive data is transferred to the second recognition module.

[0015] As a preferred solution of the sensitive information identification method for archive review and opening described in the present invention, the second sensitive information different from the first sensitive information but similar in content identified by the second module includes:

[0016] The semantic analysis model is used to convert each sensitive word and vocabulary in the text segment into corresponding word vectors, and the cosine similarity of the word vectors is calculated to obtain the semantic similarity;

[0017] If the semantic similarity exceeds a second threshold, marking the file as a sensitive file;

[0018] If the semantic similarity does not exceed the second threshold, the archive data is transferred to the third recognition module.

[0019] As a preferred solution of the sensitive information identification method for archive review and opening of the present invention, wherein: the third module is used to identify the third sensitive information implicit in the archive data;

[0020] Encoding and feature extracting the text information in the archive data to obtain first feature information of the text information;

[0021] Performing multi-level convolution and pooling operations on the first feature information to extract global features of the text information;

[0022] Based on the global features, the probability distribution of the text information is obtained, and it is determined whether the text information contains third sensitive information according to the probability distribution.

[0023] As a preferred solution of the sensitive information identification method for file review and opening of the present invention, wherein: combining the first module, the second module and the third module for analysis and processing, determining whether the file can be an open file includes:

[0024] Collect the outputs of the three modules, assign different weights to the outputs of each module, and perform weighted calculations on the outputs of each module to obtain a comprehensive sensitivity score for the archival data;

[0025] If the comprehensive sensitivity score exceeds the third threshold, the file is marked as a sensitive file; otherwise, it is a non-sensitive file.

[0026] As a preferred solution of the sensitive information identification method for archive review and opening described in the present invention, the weight corresponding to each sensitive word is cumulatively calculated, including:

[0027] Build a sensitive word library based on domain knowledge, regulatory requirements and historical data, and dynamically update the sensitive word library;

[0028] In the sensitive word database, formulate sensitive word weight allocation rules based on the type and severity of sensitive information and its importance in the specific archive content;

[0029] According to the sensitive word weight allocation rule, a weight is allocated to each sensitive word in the sensitive word library.

[0030] As a preferred solution of the sensitive information identification method for archive review and opening described in the present invention, the archive data includes:

[0031] Perform OCR processing on non-text files in the archive data and convert the text content in the file into a parsable text format.

[0032] In a second aspect, the present invention provides a system for identifying sensitive information for file review and opening, comprising:

[0033] A data acquisition module, used to acquire archive data;

[0034] A first identification module, used for identifying first sensitive information in the archival data using a first module of a multi-level identification model;

[0035] A second identification module, used to identify second sensitive information that is different from the first sensitive information but has similar content using a second module of the multi-level identification model;

[0036] A third identification module, used to identify third sensitive information hidden in the archive data using the third module of the multi-level identification model;

[0037] The comprehensive judgment module is used to analyze and process the first module, the second module and the third module to determine whether the file can be an open file.

[0038] In a third aspect, the present invention provides a computing device, comprising:

[0039] Memory and processor;

[0040] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the sensitive information identification method for archive review and opening are implemented.

[0041] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the sensitive information identification method for archive review and opening.

[0042] Compared with the prior art, the present invention has the following beneficial effects: the present invention realizes comprehensive, accurate and in-depth recognition of sensitive information in archival data by constructing a multi-level recognition model including a first module, a second module and a third module. The first module uses intelligent algorithms and sensitive word weight calculations to quickly screen out explicit sensitive information; the second module uses a semantic analysis model to effectively identify sensitive information that is different from sensitive words but has similar content, thereby enhancing the flexibility and accuracy of recognition; the third module uses deep learning technology to deeply explore the implicit sensitive information in archival data, further improving the depth and breadth of recognition. Through the method of the present invention, a more comprehensive audit opinion can be provided. Conventional methods may only give relatively simple results. This method can provide more detailed sensitive information identification, ensuring the accuracy and reliability of archival audits. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0044] Figure 1 A schematic diagram of a general process of a sensitive information identification method for file review and opening according to an embodiment of the present invention;

[0045] Figure 2 This is a schematic diagram of the overall process of a method for identifying sensitive information for file review and opening according to an embodiment of the present invention;

[0046] Figure 3 A schematic diagram of a DFA algorithm Trie tree of a sensitive information identification method for file review and opening according to an embodiment of the present invention;

[0047] Figure 4 A schematic diagram of a semantic similarity calculation process of a sensitive information identification method for file review and opening according to an embodiment of the present invention;

[0048] Figure 5 A schematic diagram of the BERT-DPCNN model structure of a sensitive information identification method for archive review and opening according to an embodiment of the present invention;

[0049] Figure 6 This is a schematic diagram of an interface of a sensitive information identification system of a sensitive information identification method for file review and opening according to an embodiment of the present invention;

[0050] Figure 7 A schematic diagram of identification results of an example test of a sensitive information identification method for archive review and opening according to an embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0052] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0053] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0054] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0055] At the same time, in the description of the present invention, it should be noted that the orientations or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the orientations or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the system or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0056] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0057] Example 1

[0058] Reference Figure 1-5 , as an embodiment of the present invention, provides a sensitive information identification method for file review and opening, comprising:

[0059] S100: Obtaining archive data;

[0060] S102: using a multi-level recognition model to identify archival data, the multi-level recognition model comprising a first module, a second module and a third module;

[0061] S102-1: using the first module to identify first sensitive information in the archive data;

[0062] S102-2: using the second module to identify second sensitive information that is different from the first sensitive information but has similar content;

[0063] S102-3: using the third module to identify the third sensitive information hidden in the archive data;

[0064] S104: Combine the first module, the second module and the third module for analysis and processing to determine whether the file can be an open file;

[0065] It should be noted that the present invention comprehensively and meticulously screens various types of sensitive information in archival data through a phased and hierarchical identification method. First, the first module quickly locks obvious sensitive content, laying the foundation for subsequent processing; then, the second module uses semantic analysis technology to effectively capture sensitive points that are similar to the first sensitive information but expressed differently, enhancing the depth and breadth of identification; finally, the third module deeply mines the potential sensitive information in the archival data to ensure the integrity of identification. The three modules work together to not only improve the accuracy and efficiency of sensitive information identification, but also provide a scientific and reliable decision-making basis for the open review of archives, effectively balancing the open use of archival information and confidentiality and security.

[0066] In the embodiment of the present application, the data acquisition in the above step S100 further includes the following sub-steps A1-A3;

[0067] In A1, the archive data to be reviewed is extracted from the archive management system;

[0068] In an optional embodiment, the archival data may be obtained by database query, API interface call, local upload, etc. The archival data may include text files, PDF files, image files, etc.

[0069] In A2, for the acquired archive data, if the archive data is a non-text file, the non-text file should be converted into a parsable text format;

[0070] In an optional embodiment, the conversion technology for converting a non-text file into a parsable text format may be a PDF to text tool, a third-party text recognition service, or ORC technology processing.

[0071] In an embodiment of the present application, OCR processing is performed on non-text type files, such as scanned PDFs or images, to convert the text content in the image into a parsable text format. At the same time, the accuracy of the OCR is checked and recognition errors are corrected to ensure the integrity and accuracy of the text content.

[0072] In A3, the obtained archive data is cleaned and segmented;

[0073] For example, the redundant characters in the text, such as spaces and punctuation marks, are cleaned up, and the text format is standardized. Synonyms and words with the same root are normalized to ensure the consistency of the text content. Word segmentation is performed to divide continuous text paragraphs into independent vocabulary units for subsequent sensitive information detection.

[0074] It should be noted that the archival data extracted from the archive management system by the present invention covers a variety of file types, providing a rich data source for subsequent multi-level sensitive information identification; for non-text type files, they are converted into a parsable text format through technical means such as OCR processing, ensuring the integrity and accuracy of the text content; the obtained archival data is subjected to text cleaning and word segmentation processing, which improves the quality of the data, makes the data more standardized and structured, and facilitates subsequent sensitive information detection and analysis.

[0075] In the embodiment of the present application, the identification using the first module in the above step S102-1 also includes sub-steps B1-B4;

[0076] In B1, a sensitive word library is established to assign weights to sensitive words;

[0077] In an optional embodiment, the basic data for establishing the sensitive word library can be derived from the knowledge in the field, regulatory requirements or historical data, and the initial sensitive word library is constructed based on these data. The sensitive word library covers various types of sensitive information that may be involved, such as personal privacy, commercial secrets, state secrets, etc.

[0078] In the embodiment of the present application, a sensitive word weight allocation rule is formulated according to the type and severity of sensitive information and its importance in the specific archive content. According to the sensitive word weight allocation rule, a weight is allocated to each sensitive word in the sensitive word library. The weight allocation can be based on expert scoring or statistical analysis.

[0079] In the embodiment of the present application, the weights of sensitive words are verified and tested to ensure the rationality of the weight distribution. According to the test results, the weight distribution of sensitive words is adjusted to improve accuracy and flexibility.

[0080] In an optional embodiment, the sensitive word weight verification test method may be through simulation test or historical data backtest;

[0081] It should be noted that the sensitive word library can be updated dynamically to ensure that the sensitive word library can be adjusted in a timely manner in accordance with new regulations and needs.

[0082] In B2, an intelligent algorithm is used to identify sensitive words in the archive data, and the weight corresponding to each sensitive word is accumulated and calculated to obtain a total weighted score;

[0083] In an optional embodiment, the intelligent algorithm may be a deterministic finite element automaton DFA, a support vector machine, a convolutional neural network CNN, a recurrent neural network RNN, a regular expression matching method, etc.;

[0084] Exemplarily, after the text preprocessing is completed, the DFA algorithm is used to quickly identify and match the explicit sensitive information in the text content; DFA is a finite automaton model based on state transition, which can quickly match the input text word by word by constructing a state diagram and state transition rules. When the input character sequence matches an entry in the sensitive word library, the DFA enters the accepting state and immediately triggers a successful match mark;

[0085] The DFA algorithm scans each character in the file in turn and inputs it into the DFA state machine; the state machine transfers along the state transition diagram according to the current state and input character until it reaches the acceptance state or the match fails. If the match is successful, the location of the sensitive word and related information are recorded.

[0086] For each sensitive word identified, a cumulative calculation is performed based on its preset weight in the sensitive word library. Each matched sensitive word will increase the overall sensitivity score. The weight values ​​of all identified sensitive words are added up to obtain the overall sensitivity score of the file, that is, the total weighted score.

[0087] In B3, if the total weighted score exceeds the first threshold, the file is marked as a sensitive file;

[0088] It should be noted that if the total weighted score exceeds the first threshold, the file will be marked as a sensitive file, no subsequent processing will be performed, and the multi-level recognition model identification will end; the setting of the first threshold can be based on statistical analysis of the weight distribution of sensitive information in historical archival data, the experience judgment of domain experts, the prediction results of machine learning algorithms, or a comprehensive analysis of ROC curves, aiming to balance the risks of false positives and missed negatives and ensure the accuracy and efficiency of sensitive file identification.

[0089] In B4, if the total weighted score does not exceed the first threshold or no sensitive word is identified, the archive data is transferred to the second recognition module.

[0090] It should be noted that the identification of explicit sensitive information, i.e., the first sensitive information, not only establishes the first solid line of defense for the archive review process, but also provides a key information basis for subsequent processing. By quickly and efficiently identifying explicit and intuitive sensitive words in the archives, and performing weight division and scoring, the system can quickly determine the sensitivity of the archives, thereby effectively avoiding the leakage of sensitive information;

[0091] It should also be noted that when the first module completes the recognition task, the files will be divided into two categories according to the results of the total weighted score: sensitive files and non-sensitive files. For sensitive files, the system will immediately terminate the subsequent multi-level recognition model processing, and for files that do not exceed the first threshold or do not recognize sensitive words, they will be transferred to the second recognition module for further in-depth analysis.

[0092] In the embodiment of the present application, the identification using the second module in the above step S102-2 also includes sub-steps C1-C3;

[0093] In C1, the semantic analysis model is used to convert each sensitive word and vocabulary in the text segment into a corresponding word vector, and the cosine similarity of the word vector is calculated to obtain the semantic similarity;

[0094] In an optional embodiment, the semantic analysis model may be a word vector model such as Word2Vec, GloVe, and FastTex, etc. Word2Vec is a commonly used word embedding technology that can map words from a high-dimensional discrete space to a low-dimensional continuous space, thereby capturing the semantic relationship between words;

[0095] For example, the pre-trained word vector model Word2Vec is used for recognition analysis, and the semantic analysis process is: using the Word2Vec model to convert each sensitive word and word in the text segment into a corresponding word vector. During the conversion process, the Word2Vec model will learn the word vector representation of each word based on the co-occurrence relationship of the word in a large text corpus. These word vectors have certain distribution characteristics in space, and similar words will be mapped to similar positions;

[0096] Calculate the cosine similarity between the sensitive word vector and each vocabulary word vector in the text segment. Cosine similarity is a commonly used method to measure the similarity between two vectors. Its value range is between [-1,1]. When the directions of two vectors are closer, their cosine similarity is closer to 1; conversely, when the directions of two vectors are farther apart, their cosine similarity is closer to -1.

[0097] By calculating the cosine similarity, the semantic similarity between each sensitive word and the vocabulary in the text segment is obtained.

[0098] In C2, if the semantic similarity exceeds the second threshold, the file is marked as a sensitive file;

[0099] It should be noted that if the semantic similarity exceeds the second threshold, the file is marked as a sensitive file, no subsequent processing is performed, and the multi-level recognition model recognition ends.

[0100] In C3, if the semantic similarity does not exceed the second threshold, the archive data is transferred to the third recognition module.

[0101] It should be noted that semantic similarity analysis and identification is a key step in the sensitive information identification process. It aims to identify potential implicit sensitive information that has not been explicitly matched by the first module's intelligent algorithm by calculating the semantic similarity between the text and the words in the sensitive word library. This step uses semantic analysis technology to make up for the deficiency that the first module can only identify explicit sensitive words.

[0102] In the embodiment of the present application, the identification using the third module in the above step S102-3 also includes sub-steps D1-D3;

[0103] In D1, the text information in the archive data is encoded and feature extracted to obtain first feature information of the text information;

[0104] In D2, the first feature information is subjected to multi-level convolution and pooling operations to extract the global features of the text information;

[0105] In D3, based on the global features, the probability distribution of the text information is obtained, and whether the text information contains the third sensitive information is determined according to the probability distribution;

[0106] In the embodiment of the present application, the third module applies BERT and DPCNN models for collaborative recognition analysis, specifically: first, the preprocessed text is encoded by the BERT model to generate a deep semantic vector of each word in the context; the output of the last layer or the middle layer of the BERT model is extracted as features, which will contain the contextual semantic representation of the text fragment;

[0107] Then, using the deep convolutional structure of DPCNN, multi-level convolution and pooling operations are performed on the features extracted from BERT. DPCNN can gradually extract the global features of the text through layer-by-layer compression and feature extraction. Finally, the output of DPCNN is classified through a fully connected layer. The model will output a probability distribution, indicating the possibility that the input text belongs to sensitive information or non-sensitive information. The output sensitivity score is used to determine whether the text fragment contains implicit sensitive information.

[0108] It should be noted that the identification of implicit sensitive information, that is, the third sensitive information, is mainly aimed at sensitive content that cannot be directly detected by explicit vocabulary. This part of the content may be difficult to be discovered by traditional methods due to obscure expressions, complex contexts or terminology in specific fields. In order to identify these implicit sensitive information, the present invention mainly uses BERT and DPCNN models for collaborative analysis to improve the recognition accuracy of complex and obscure sensitive information and identify potential sensitive information from a semantic level.

[0109] S104: Combine the first module, the second module and the third module for analysis and processing to determine whether the file can be an open file;

[0110] In the embodiment of the present application, the above step S104 also includes sub-steps E1-E5;

[0111] In E1, multi-module output fusion: the multi-level analysis model combines the recognition results from multiple levels and conducts a comprehensive analysis of the sensitivity of each text fragment according to preset rules or models; obtains explicit sensitive words that match the sensitive word library and the context in which these words are located from the first module; collects possible sensitive information after semantic similarity analysis from the second module, including potential sensitive information that shows high similarity in the context; obtains the prediction results of the BERT-DPCNN model and the sensitivity score of the text fragment from the third module.

[0112] In E2, the results are weighted and scored: different weights are assigned according to the accuracy and applicability of each level. The deep learning model may have a higher context understanding ability, so it can be given a higher weight; although the vocabulary matching results are simple, they have a high accuracy in identifying explicit sensitive information, so a certain weight also needs to be retained; the results of each level are weighted and calculated to generate a comprehensive sensitivity score for each text fragment.

[0113] In E3, the audit results are output: a threshold of sensitivity score is set, namely the third threshold, and the comprehensive sensitivity score is compared with the third threshold. If the comprehensive sensitivity score of a certain file exceeds the third threshold, the file is judged as a sensitive file, otherwise it is a non-sensitive file;

[0114] In E4, detected sensitive information files are annotated for user convenience, and each sensitive file is accompanied by an identification source, sensitivity score, and category.

[0115] It should be noted that comprehensive analysis and processing is the last step in the sensitive information identification process, which aims to integrate the results generated by the first three levels and give the final sensitive information judgment through the review mechanism. This step not only relies on the results of each level, but also requires multi-angle evaluation and synthesis of the results to ensure the accuracy and comprehensiveness of the judgment.

[0116] The above is a schematic scheme of a sensitive information identification method for archive review and opening in this embodiment. It should be noted that the technical scheme of the system for sensitive information identification for archive review and opening and the technical scheme of the sensitive information identification method for archive review and opening belong to the same concept, and the details not described in detail in the technical scheme of the sensitive information identification system for archive review and opening in this embodiment can be found in the description of the technical scheme of the sensitive information identification method for archive review and opening.

[0117] The sensitive information identification system for file review and opening in this embodiment includes:

[0118] A data acquisition module, used to acquire archive data;

[0119] A first identification module, used to identify first sensitive information in the archive data using a first module of the multi-level identification model;

[0120] A second identification module, used to identify second sensitive information that is different from the first sensitive information but has similar content using the second module of the multi-level identification model;

[0121] A third identification module, used to identify third sensitive information hidden in the archive data using the third module of the multi-level identification model;

[0122] The comprehensive judgment module is used to analyze and process the first module, the second module and the third module to determine whether the file can be an open file.

[0123] This embodiment also provides a computing device suitable for identifying sensitive information for file review and opening, including:

[0124] Memory and processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the sensitive information identification method for archive review and opening as proposed in the above embodiment.

[0125] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for identifying sensitive information for archive review and opening as proposed in the above embodiment is implemented.

[0126] The storage medium proposed in this embodiment and the sensitive information identification method for archive review and opening proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0127] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ReadOnly, Memory, ROM), random access memory (RandomAccess Memory, RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform the methods of various embodiments of the present invention.

[0128] Example 2

[0129] Reference Figure 2-7 Tables 1-3 are an embodiment of the present invention, which provides a sensitive information identification method for archive review and opening. In order to verify its beneficial effects, the test results of the method of the present invention are provided.

[0130] S1: archival data acquisition and preprocessing;

[0131] S2: weight division of sensitive words;

[0132] The weight division of sensitive words is one of the core steps. Its purpose is to assign a quantitative weight value to each sensitive word according to its importance, frequency of occurrence and potential risk level. Through reasonable weight division, the accuracy of sensitive word identification can be improved, thereby better judging the open status of archives.

[0133] Definition of sensitive word weights: Assume that the sensitive word library contains n sensitive words, denoted by W = {w1,w2…w n}, for each sensitive word w i , whose weight is defined as α i , α i is a non-negative number, indicating the importance of the sensitive word in the file review. i The calculation of can be based on the following factors: the importance of sensitive words I i : Indicates the influence of sensitive words in a specific context;

[0134] For example, certain names of people or places have high importance in a certain period; the frequency of occurrence of sensitive words F i : Indicates the frequency of occurrence of sensitive words in the archive data set. The higher the frequency, the greater the sensitivity. The risk level of sensitive words R i: Indicates the risk of information leakage involved in sensitive words. The risk level can be set based on the evaluation of domain experts or historical data.

[0135] Weight calculation formula, sensitive word w i The weight α i It can be calculated by the following formula:

[0136] α i =λ1·I i +λ2·F i +λ3·R i

[0137] Among them, λ1, λ2, and λ3 are weight coefficients, satisfying λ1+λ2+λ3=1, which are used to balance the impact of different factors on the weight of sensitive words. These coefficients can be adjusted through experience or training based on actual data.

[0138] Normalization of weights: To ensure that the weights of different sensitive words are within a reasonable range, the weight α i Normalization is required. Normalization can be performed using the following formula:

[0139]

[0140] in, is the normalized weight of sensitive words, ensuring that the sum of the weights of all sensitive words is 1. This normalization process helps avoid bias caused by excessive weights of some sensitive words.

[0141] S3: first sensitive information, i.e. identification of explicit sensitive information;

[0142] In the first step S3.1 of explicit sensitive information identification, the DFA algorithm is used to match sensitive words, and the archive text stream is matched word by word with the entries in the sensitive word library. Once a sensitive word is found, the DFA algorithm will immediately mark and record the position of the word in the archive;

[0143] For example, if the file content is "This file contains confidential information and cannot be disclosed without permission", and the sensitive word library contains the following sensitive word "confidential information", then a Trie tree based on sensitive words needs to be constructed, such as Figure 3 Then, the DFA algorithm is used for retrieval and matching. This step is highly efficient and can quickly screen out explicit sensitive information, providing data support for subsequent sensitivity assessment.

[0144] In the matching result analysis and weight calculation step S3.2, the identified sensitive words are analyzed and assigned different weights;

[0145] In the matching result processing S3.3, the archives are classified according to the accumulated weight scores;

[0146] Some core pseudo codes are shown in Table 1 below;

[0147] Table 1

[0148]

[0149]

[0150] S4: Second sensitive information identification;

[0151] Use the Word2Vec model to vectorize text and sensitive words. Word2Vec converts words into vector representations in a high-dimensional space, so that the semantics of words can be measured by the distance between vectors.

[0152] The trained Word2Vec model can be used to convert sensitive words and each word in the text fragment into word vectors to reflect their semantic meanings. In the similarity calculation, the cosine similarity between these word vectors is calculated to measure the semantic similarity between the text fragment and the sensitive word. The formula is:

[0153]

[0154] in, Representation vector and The dot product of Represents vectors and In this way, we can effectively identify potential sensitive information that does not directly use sensitive words on the surface but has semantic associations.

[0155] A threshold of semantic similarity is set to distinguish which text fragments have sensitive tendencies. Specifically, if the cosine similarity between a text fragment and a sensitive word is higher than the set threshold, the text fragment is considered to express a meaning similar to that of the sensitive word and is considered to contain potential hidden sensitive information. The system will record these matching fragments, similar sensitive words and their specific locations, and also record their similarity values. This sophisticated semantic similarity calculation can help the system capture more sensitive information hidden under the surface text and improve the comprehensiveness of sensitive information identification.

[0156] Some core pseudo codes are shown in Table 2 below;

[0157] Table 2

[0158]

[0159]

[0160] S5: The third sensitive information, i.e., implicit sensitive information identification;

[0161] In the third module, the BERT-DPCNN model is used to work together to process sensitive information that cannot be detected by explicit vocabulary. Implicit sensitive information is usually expressed through complex contexts, obscure expressions, or terminology in specific fields, which is difficult to identify by traditional methods. Through the combination of these two models, the system can deeply analyze the semantic level of the text and identify potential sensitive information.

[0162] Some core pseudo codes are shown in Table 3 below;

[0163] Table 3

[0164]

[0165] S6: comprehensive analysis and processing;

[0166] The present invention is a system for reviewing sensitive information of archives. Figure 6 As shown, Figure 6 In the embodiment, the first level is the first module described in the present invention, the second level is the second module described in the present invention, and the third level is the third module described in the present invention;

[0167] The method of the present invention is used to identify the archives, and the final result is shown in the figure below: Figure 7 As shown, from Figure 7 It can be seen that the current total sensitivity score of the archive is 83.7 points; after multiple tests, it can be well identified whether the archive contains sensitive information. Using the method of the present invention, the archive identification accuracy is high and the reliability is strong.

[0168] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for identifying sensitive information for file review and opening, characterized in that: include: Acquire archival data, and identify the archival data using a multi-level recognition model, wherein the multi-level recognition model includes a first module, a second module, and a third module; Using the first module to identify first sensitive information in the archival data, using the second module to identify second sensitive information that is different from the first sensitive information but has similar content, and using the third module to identify third sensitive information implicit in the archival data; Combine the analysis and processing of the first module, the second module and the third module to determine whether the file can be an open file.

2. The sensitive information identification method for file review and opening as claimed in claim 1, characterized in that: Using the first module to identify the first sensitive information in the archive data includes: Using intelligent algorithms to identify sensitive words in the archive data, accumulating the weights corresponding to each sensitive word to obtain a total weighted score; If the total weighted score exceeds a first threshold, the file is marked as a sensitive file; If the total weighted score does not exceed the first threshold or no sensitive word is identified, the archive data is transferred to the second recognition module.

3. The sensitive information identification method for file review and opening as claimed in claim 1 or 2, characterized in that: Using the second module to identify second sensitive information that is different from the first sensitive information but has similar content includes: The semantic analysis model is used to convert each sensitive word and vocabulary in the text segment into corresponding word vectors, and the cosine similarity of the word vectors is calculated to obtain the semantic similarity; If the semantic similarity exceeds a second threshold, marking the file as a sensitive file; If the semantic similarity does not exceed the second threshold, the archive data is transferred to the third recognition module.

4. The sensitive information identification method for file review and opening as claimed in claim 3, characterized in that: Using the third module to identify third sensitive information hidden in the archival data; Encoding and feature extracting the text information in the archive data to obtain first feature information of the text information; Performing multi-level convolution and pooling operations on the first feature information to extract global features of the text information; Based on the global features, the probability distribution of the text information is obtained, and it is determined whether the text information contains third sensitive information according to the probability distribution.

5. The sensitive information identification method for file review and opening as claimed in claim 4, characterized in that: Combining the analysis and processing of the first module, the second module and the third module to determine whether the file can be an open file includes: Collect the outputs of the three modules, assign different weights to the outputs of each module, and perform weighted calculations on the outputs of each module to obtain a comprehensive sensitivity score for the archival data; If the comprehensive sensitivity score exceeds the third threshold, the file is marked as a sensitive file; otherwise, it is a non-sensitive file.

6. The sensitive information identification method for file review and opening as claimed in claim 2, characterized in that: The weight corresponding to each sensitive word is accumulated and calculated, including: Build a sensitive word library based on domain knowledge, regulatory requirements and historical data, and dynamically update the sensitive word library; In the sensitive word database, formulate sensitive word weight allocation rules based on the type and severity of sensitive information and its importance in the specific archive content; According to the sensitive word weight allocation rule, a weight is allocated to each sensitive word in the sensitive word library.

7. The sensitive information identification method for file review and opening as claimed in claim 6, characterized in that: The archival data includes: Perform OCR processing on non-text files in the archive data and convert the text content in the file into a parsable text format.

8. A sensitive information identification system for file review and opening, characterized in that: include, A data acquisition module, used to acquire archive data; A first identification module, used for identifying first sensitive information in the archival data using a first module of a multi-level identification model; A second identification module, used to identify second sensitive information that is different from the first sensitive information but has similar content using a second module of the multi-level identification model; A third identification module, used to identify third sensitive information hidden in the archive data using the third module of the multi-level identification model; The comprehensive judgment module is used to analyze and process the first module, the second module and the third module to determine whether the file can be an open file.

9. An electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the sensitive information identification method for archive review and opening as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the sensitive information identification method for archive review and opening as described in any one of claims 1 to 7.

Images