Sensitive file identification processing method and system based on clustering model

Abstract

The invention discloses a sensitive file identification processing method and system based on a clustering model, and relates to the technical field of data processing, and the method comprises the following steps: 1, data preprocessing: carrying out text extraction, word segmentation processing and cleaning processing on obtained to-be-detected file data in sequence; the to-be-detected file data comprises text data, image data or multimedia data; and step 2, clustering recognition: performing feature extraction on the cleaned text. According to the sensitive file recognition processing method and system based on the clustering model, a sensitive file recognition strategy based on the clustering model is provided, sensitive files can be automatically found and classified, and manual rules or keyword lists are not needed. Through automatic clustering and feature extraction, the sensitive file identification efficiency and precision are greatly improved, automatic identification, automatic encryption and automatic alarm of the sensitive file are realized, and the automation level and accuracy of data processing are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a sensitive file identification and disposal method and system based on a clustering model. BACKGROUND

[0002] With the explosive growth of digital information, the management and protection of sensitive files have become crucial. In the modern information environment, enterprises, institutions and individuals face increasingly complex business scenarios, and the amount of data is growing continuously. The informationization framework based on data flow is evolving, which puts higher requirements on data security. Sensitive files contain important business secrets, personal privacy information or other critical data, and once leaked, may cause serious economic losses, legal risks and reputation damage.

[0003] Traditional sensitive file identification and disposal methods have many drawbacks. These methods usually rely on manually configured rules and keywords for text matching. In the context of a large amount of data, manual configuration work is not only inefficient, but also prone to errors. Since the rules and keywords are set to meet only part of the specific scenarios, they cannot truly understand the deep meaning of the text content, so the accuracy of the judgment is low, and misjudgment and missed judgment may occur. For example, when dealing with some professional or semantically obscure files, simple keyword matching may not be able to accurately identify sensitive information. Moreover, after identifying sensitive files, traditional methods lack effective processing mechanisms, making it difficult to achieve timely and automatic protection and alerting of sensitive files, and cannot meet the current strict requirements for data security.

[0004] To solve the above problems, we propose a sensitive file identification and disposal method and system based on a clustering model. SUMMARY

[0005] To address the deficiencies of the prior art, the present application provides a sensitive file identification and disposal method and system based on a clustering model, which solves the problems raised in the background art.

[0006] To achieve the above purpose, the present application is implemented by the following technical solution: a sensitive file identification and disposal method based on a clustering model, comprising the following steps:

[0007] Step 1, data preprocessing: sequentially performing text extraction, word segmentation processing and cleaning processing on the obtained file data to be detected; the file data to be detected includes text data, image data or multimedia data, the original file is stored in a file storage server, and the file identifier, creator, capture time and path are stored in a message queue; the cleaning result and file identifier are stored in a new message queue after processing;

[0008] Step two, cluster identification: the text after cleaning is used to extract feature vectors with TF-IDF model, and k-means clustering is used to cluster the feature vectors. The cluster center is updated by the cluster center iterative optimization formula during clustering:

[0009]

[0010] wherein, is the cluster center of the i-th cluster after the t+1 iteration, is the cluster center after the t iteration, and β is the iteration update coefficient (0.4-0.8), is the number of files in the i-th cluster, and x is the feature vector of the file in the cluster;

[0011] Step three, sensitive processing: sensitive information containing clusters are identified by similarity detection, and the files in the identified clusters are encrypted, and an alarm information is generated. The sensitive density of each cluster is calculated, and the clusters with sensitive density exceeding the threshold are disposed as a whole. The calculation formula of the sensitive density D is: wherein is the number of files in the cluster with high similarity to the sensitive information library, is the total number of files in the cluster, is the average similarity score of all files in the cluster to the sensitive information library.

[0012] Preferably, in the data preprocessing step, the original files are stored in a file storage server, and each original file corresponds to a file identifier. The file identifier, file creator, file capture time and file storage path are stored in a message queue. When performing text extraction, word segmentation processing and cleaning processing, the file identifier and storage path in the message queue are read, the original file is read from the file server according to the original file storage path, and text extraction is performed on the text data, image data or multimedia data. Then the text extraction result is sequentially subjected to word segmentation processing and cleaning processing, and the file identifier and cleaning processing result are stored in a new message queue.

[0013] Preferably, in the cluster identification step, the k value is determined by the gap statistical method, and the minimum k value satisfying gap(k)>gap(k+1)-s_k is selected, s_k=sqrt(1+1 / n×sd_k), and sd_k is the standard deviation of gap(k).

[0014] Preferably, in the sensitive processing step, the file storage path, file identifier, file clustering score and file capture time of the file hitting the sensitive information are stored in the database, and when the sensitive file is identified, an alarm information is generated and sent to a specific terminal.

[0015] The application also discloses a sensitive file identification and disposal system based on a clustering model.

[0016] The preprocessing module is configured to sequentially perform text extraction, word segmentation processing and cleaning processing on the obtained file data to be detected.

[0017] The clustering identification module is configured to perform feature extraction on the cleaned text, and perform clustering on the extracted features, similar files are classified into the same group, and a class cluster containing sensitive information is identified through similarity detection.

[0018] The storage module is used for storing original files, intermediate data in the processing process and identification results and the like.

[0019] The alarm module is responsible for generating alarm information and sending the alarm information to a specific terminal when sensitive files are identified.

[0020] Preferably, the preprocessing module further comprises:

[0021] The text extraction unit is responsible for extracting text content from different types of original files.

[0022] The word segmentation unit performs word segmentation operation on the text extraction result.

[0023] The cleaning unit cleans the text after word segmentation, and removes noise and irrelevant information.

[0024] Preferably, the clustering identification module comprises:

[0025] The feature extraction unit extracts features from the cleaned text data using a tf-idf model.

[0026] The clustering unit adopts a k-means algorithm to perform clustering on the extracted features.

[0027] The sensitive detection unit identifies a class cluster containing sensitive information through similarity detection.

[0028] Preferably, the storage module comprises:

[0029] The file storage server unit stores original files.

[0030] The message queue unit is used for storing file identification, file creator, file capture time and file storage path and the like, and transferring intermediate data in the data processing process.

[0031] The database unit stores relevant information of files containing sensitive information, such as file storage path, file identification, file clustering score and file capture time.

[0032] Preferably, the alarm module comprises:

[0033] The alarm generating unit generates corresponding alarm information according to the identification result of the sensitive file;

[0034] The terminal communication unit sends the alarm information to a specific terminal;

[0035] The alarm recording unit records the alarm information for subsequent query and analysis.

[0036] Advantages

[0037] The present application provides a sensitive file identification and disposal method and system based on a clustering model. Compared with the prior art, the following advantages are possessed:

[0038] (1) The present application proposes a sensitive file identification strategy based on a clustering model, which can automatically discover and classify sensitive files without manual rules or keyword lists. Through automatic clustering and feature extraction, the efficiency and accuracy of sensitive file identification are greatly improved, realizing automatic identification, automatic encryption and automatic alarm of sensitive files, effectively improving the automation level and accuracy of data processing.

[0039] (2) The present application does not require manual rules, can accurately, automatically and quickly identify sensitive files in a large number of files, and take automatic protection measures, reducing the risk of data leakage. After the leakage of sensitive files, it can produce alarm and automatically dispose in time, significantly improving the data security, providing strong protection for the data security of enterprises, institutions and individuals. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 The flowchart of the present application;

[0041] Figure 2 The system block diagram of the present application;

[0042] Figure 3 The schematic diagram of the preprocessing module of the present application;

[0043] Figure 4 The schematic diagram of the clustering identification module of the present application;

[0044] Figure 5 The schematic diagram of the storage module of the present application;

[0045] Figure 6 The schematic diagram of the alarm module of the present application.

[0046] In the figure: 01, preprocessing module; 02, clustering identification module; 03, storage module; 04, alarm module; 011, text extraction unit; 012, word segmentation unit; 013, cleaning unit; 021, feature extraction unit; 022, clustering unit; 023, sensitive detection unit; 031, file storage server unit; 032, message queue unit; 033, database unit; 041, alarm generation unit; 042, terminal communication unit; 043, alarm record unit. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0048] The embodiments of the present application provide three technical solutions, specifically including the following embodiments:

[0049] Embodiment 1

[0050] Please refer to Figure 1 A sensitive file identification and disposal method based on a clustering model, comprising the following steps:

[0051] Step one, data preprocessing: sequentially performing text extraction, word segmentation processing and cleaning processing on the obtained to-be-detected file data. The to-be-detected file data includes text data, image data or multimedia data. The original file is stored in a file storage server, and each original file corresponds to a file identifier. The file identifier, file creator, file capture time and file storage path are stored in a message queue. When performing text extraction, word segmentation processing and cleaning processing, the file identifier and storage path in the message queue are read, the original file is read from the file server according to the original file storage path, and text extraction is performed on the text data, image data or multimedia data. Then, the text extraction result is sequentially subjected to word segmentation processing and cleaning processing, and the file identifier and cleaning processing result are stored in a new message queue;

[0052] Step two, clustering identification: extracting a feature vector from the cleaned text using a TF-IDF model, and clustering the feature vector using k-means clustering. In the clustering process, the cluster center is updated through a cluster center iterative optimization formula, and the cluster center iterative optimization formula is as follows:

[0053]

[0054] Wherein, is the cluster center after the t+1 th iteration of the i th cluster, is the cluster center after the tth iteration, β is the iteration update coefficient (0.4-0.8), is the number of files in the ith cluster, x is the feature vector of the files in the cluster;

[0055] Step three, sensitive processing: identify the class cluster containing sensitive information through similarity detection, encrypt the files in the identified class cluster, and generate an alarm information. The file storage path, file identification, file clustering score and file capture time that hit the sensitive information are stored in the database, and when the sensitive file is identified, an alarm information is generated and sent to a specific terminal.

[0056] In the sensitive processing step, the specific implementation is as follows:

[0057] a. Calculate the class cluster sensitive density:

[0058] For each class cluster generated after clustering , calculate its sensitive density , the calculation formula is: , wherein: is the total number of files in the class cluster , is the number of files in the class cluster whose similarity to the sensitive information library exceeds the preset basic threshold , is the average similarity score of all files in the class cluster to the sensitive information library. b. Risk determination and disposal:

[0059] Set a sensitive density threshold , when the sensitive density

[0060] of a class cluster is greater than , the system determines that the class cluster is a high-risk sensitive class cluster. Then, all files under the class cluster are uniformly encrypted, and an alarm information is generated.

[0061] In the data preprocessing step, the original files are stored in the file storage server, and each original file corresponds to a file identification. The file identification, file creator, file capture time and file storage path are stored in the message queue. When performing text extraction, word segmentation processing and cleaning processing, read the file identification and storage path in the message queue, read the original file from the file server according to the original file storage path, and perform text extraction on the text data, image data or multimedia data. Then, the text extraction result is sequentially subjected to word segmentation processing and cleaning processing, and the file identification and cleaning processing result are stored in a new message queue.

[0062] ​​In the clustering identification step, the value of k is determined by the gap statistics method. The smallest k value that satisfies gap(k)>gap(k+1)-s_k is selected, s_k=sqrt(1+1 / n×sd_k), where sd_k is the standard deviation of gap(k).

[0063] In the sensitive processing step, the storage path, file identifier, file clustering score, and file capture time of the file containing sensitive information are stored in the database. When a sensitive file is identified, an alarm message is generated and sent to a specific terminal.

[0064] Example 2

[0065] Based on Example 1, see Figures 2-4 As shown, this invention also discloses a sensitive document identification and processing system based on a clustering model, comprising:

[0066] Preprocessing module 01 is configured to sequentially perform text extraction, word segmentation, and cleaning on the acquired file data to be detected. This module further includes a text extraction unit, a word segmentation unit, and a cleaning unit. The text extraction unit is responsible for extracting text content from different types of original files (text data, image data, or multimedia data); the word segmentation unit performs word segmentation on the extracted text results; and the cleaning unit cleans the segmented text, removing noise and irrelevant information.

[0067] Clustering and Recognition Module 02: This module is configured to extract features from the cleaned text, cluster the extracted features, group similar files into the same group, identify clusters containing sensitive information through similarity detection, encrypt files within the identified clusters, and generate alarm messages. This module includes a feature extraction unit, a clustering unit, and a sensitivity detection and processing unit. The feature extraction unit uses the TF-IDF model to extract features from the cleaned text data; the clustering unit uses the k-means algorithm to cluster the extracted features; and the sensitivity detection and processing unit identifies clusters containing sensitive information through similarity detection, encrypts files within the sensitive clusters, and generates alarm messages.

[0068] Storage Module 03: This module stores the original files, intermediate data during processing, and identification results. It includes a file storage server unit, a message queue unit, and a database unit. The file storage server unit stores the original files; the message queue unit stores information such as file identifiers, file creators, file capture times, and file storage paths, and transmits intermediate data during data processing; the database unit stores information related to files that have hit sensitive information, such as file storage paths, file identifiers, file clustering scores, and file capture times.

[0069] Alarm Module 04: When a sensitive file is identified, this module generates an alarm message and sends it to a specific terminal. This module includes an alarm generation unit, a terminal communication unit, and an alarm recording unit. The alarm generation unit generates corresponding alarm messages based on the sensitive file identification results; the terminal communication unit sends the alarm messages to the specific terminal; and the alarm recording unit records the alarm messages for subsequent querying and analysis.

[0070] Preprocessing module 01 further includes:

[0071] Text extraction unit 011: Responsible for extracting text content from different types of source files;

[0072] Word segmentation unit 012: Performs word segmentation on the extracted text results;

[0073] Cleaning Unit 013: Cleans the segmented text to remove noise and irrelevant information.

[0074] Clustering identification module 02 includes:

[0075] Feature extraction unit 021: Extracts features from cleaned text data using the TF-IDF model;

[0076] Clustering unit 022: The extracted features are clustered using the k-means algorithm;

[0077] Sensitive detection unit 023: Identifies clusters containing sensitive information through similarity detection.

[0078] Example 3

[0079] Based on Example 2, see Figures 5-6 As shown, storage module 03 includes:

[0080] File storage server unit 031: Stores the original files;

[0081] Message queue unit 032: Used to store information such as file identifier, file creator, file capture time and file storage path, and to transmit intermediate data during data processing;

[0082] Database Unit 033: Stores information related to files that contain sensitive information, such as file storage path, file identifier, file cluster score, and file capture time.

[0083] Alarm module 04 includes:

[0084] Alarm generation unit 041: Generates corresponding alarm information based on the identification results of sensitive files;

[0085] Terminal communication unit 042: Sends alarm information to a specific terminal;

[0086] Alarm record unit 043: Record alarm information for subsequent query and analysis.

[0087] Example 4

[0088] Based on Example 3, 1. Data acquisition and preprocessing:

[0089] Text extraction: Read the file identifier and storage path from the message queue, and read the original file from the file server according to the storage path. For text data, its content can be directly read; for image data, OCR (Optical Character Recognition) technology is adopted, such as using OCR tools like Tesseract to convert the text in the image into text; for multimedia data, first extract the text information therein through corresponding multimedia parsing tools, for example, extract the text from the subtitle track of the video.

[0090] Word segmentation processing: Process the extracted text using a word segmentation tool. For example, for Chinese, the Jieba word segmentation tool can be used. Jieba word segmentation can accurately segment continuous text into individual words by constructing a dictionary and a statistical language model. For example, segment the text "This is a sensitive commercial document" into "This", "is", "a", "sensitive", "commercial", "document".

[0091] Cleaning processing: Remove stop words (such as虚词like "de", "shi", "zai", etc. that have no practical meaning), punctuation marks, and low-frequency words (words with extremely low occurrence frequencies and little contribution to text features) from the word segmentation results. Match and filter out characters and words that do not meet the requirements through regular expressions. For example, use the regular expression "[^\w\s]" to match and remove all characters that are not letters, numbers, or spaces. The cleaned text and the corresponding file identifier are stored in a new message queue.

[0092] 2. Feature extraction and clustering

[0093] Feature Extraction: Data is read from the new message queue, and text features are extracted using the tf-idf (Term Frequency - Inverse Document Frequency) model. The tf-idf model measures the importance of a word to a document by calculating the term frequency (tf) of each word in the document and the inverse document frequency (idf) of that word in the entire document set. Assuming there are N documents in the document set, the number of times a word w appears in document d is tf(w, d), and the number of documents containing word w is n(w), then the idf value of word w is calculated as: idf(w) = log(N / n(w)), and the tf-idf value of word w in document d is: tf-idf(w, d) = tf(w, d) * idf(w). For example, in a collection of 100 documents, if the word "sensitive" appears 5 times in one document and is contained in 10 other documents, then the tf value for "sensitive" in that document is 5, the idf value is log(100 / 10) = 2, and the tf - idf value is 5 * 2 = 10. The calculated tf - idf values ​​form the feature vector of the text.

[0094] Clustering: The extracted feature vectors are subjected to k-means clustering. The following steps are followed to determine the value of k: First, the sample data is clustered, with k ranging from 0 to k_max (e.g., k_max = 10). The intra-cluster deviation sum w_k is calculated for different k values. The intra-cluster deviation sum is the sum of the squared distances from each sample point to the center of its cluster. Assuming the center of the i-th cluster is c_i, and this cluster has n_i sample points x_j, then the intra-cluster deviation sum for this cluster is: sum_{j= 1}^{n_i} dist (x_j, c_i)^2. The intra-cluster deviation sums of all clusters are summed to obtain w_k. Then, n inference data points (e.g., n = 50) are generated through a random uniform distribution. These inference data points are clustered, again with k ranging from 0 to k_max. The average intra-cluster deviation sum w_kn for different k values ​​on the n inference data points is calculated. Next, the logarithmic deviation gap(k) between w_k and w_kn is calculated, gap(k) = log(w_kn) - log(w_k), and the standard deviation sd_k of this deviation is also calculated. Let s_k = sqrt(1 + 1 / n*sd_k). Finally, the smallest k is selected, such that gap(k) > gap(k + 1) - s_k + 1, and this k value is used as the k value in the k-means clustering. After clustering, similar files are grouped into the same group.

[0095] 3. Sensitive document identification and processing

[0096] Sensitivity Detection: Identifying clusters containing sensitive information through similarity detection. A sensitive information database can be pre-built, containing known sensitive words, phrases, or semantic patterns. For each clustered group, the similarity between the feature vectors of files in the cluster and the feature vectors in the sensitive information database is calculated, for example, using cosine similarity. Assuming the feature vector of a file in a cluster is A = (a_1, a_2, …, a_n), and the feature vector in the sensitive information database is B = (b_1, b_2, …, b_n), then the cosine similarity is calculated as: cosine (A, B) = sum_{i = 1}^{n} a_i * b_i / (sqrt(sum_{i = 1}^{n} a_i^2) * sqrt (sum_{i = 1}^{n} b_i^2)). When the similarity exceeds a certain threshold (e.g., 0.8), the cluster is considered to contain sensitive information.

[0097] Encryption Processing: Files within the identified clusters containing sensitive information are encrypted. The AES (Advanced Encryption Standard) algorithm can be used, offering high security and efficiency. First, an encryption key is generated, for example, using a 128-bit key. Then, the file content is encrypted in blocks. The file is divided into fixed-size blocks (e.g., 128-bit blocks), and each block is encrypted sequentially, converting plaintext into ciphertext, ensuring file security during storage and transmission.

[0098] Alarm Generation: This function generates alarm information, including key information such as the file identifier, storage path, cluster score, and capture time of sensitive files. The alarm information is sent to specific terminals, such as the security administrator's mobile phone or computer, via SMS, email, or internal system messages to facilitate timely further action. Simultaneously, this information is stored in a database for subsequent querying, statistics, and analysis.

[0099] Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0100] The embodiments of the invention have been described in detail above, but the content described is only a preferred embodiment of the invention and should not be considered as limiting the scope of the invention. All equivalent changes and improvements made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for identifying and handling sensitive documents based on a clustering model, characterized in that: Includes the following steps: Step 1, Data Preprocessing: The acquired file data to be detected is subjected to text extraction, word segmentation, and cleaning in sequence; the file data to be detected includes text data, image data, or multimedia data; the original file is stored on a file storage server, and the file identifier, creator, capture time, and path are stored in a message queue; after processing, the cleaning result and the file identifier are stored in a new message queue. Step 2, Clustering and Identification: Feature vectors are extracted from the cleaned text using the TF-IDF model. K-means clustering is then used to cluster these feature vectors. During the clustering process, the cluster centers are updated using an iterative optimization formula: in, Let i be the cluster center of the i-th cluster after the (t+1)-th iteration. Let t be the cluster center after the t-th iteration, and β be the iteration update coefficient (0.4-0.8). Let x be the number of files in the i-th cluster, and x be the feature vector of the files within the cluster. Step 3, Sensitive Processing: Clusters containing sensitive information are identified through similarity detection. Files within the identified clusters are encrypted, and alarm messages are generated. The sensitivity density of each cluster is calculated, and clusters with sensitivity densities exceeding a threshold are processed collectively. The formula for calculating the sensitivity density D is: ,in This represents the number of files in this cluster that are highly similar to sensitive information databases. The total number of files in this cluster. The average similarity score between all files in this cluster and the sensitive information database.

2. The method for identifying and handling sensitive documents based on a clustering model according to claim 1, characterized in that: In the data preprocessing step, the original files are stored in the file storage server. Each original file has a corresponding file identifier. The file identifier, file creator, file capture time, and file storage path are all stored in the message queue. When performing text extraction, word segmentation, and cleaning, the system reads the file identifier and storage path from the message queue, reads the original file from the file server according to the original file storage path, and performs text extraction on the text data, image data, or multimedia data. Then, the text extraction results are sequentially processed by word segmentation and cleaning, and the file identifier and cleaning results are stored in a new message queue.

3. The method for identifying and handling sensitive documents based on a clustering model according to claim 1, characterized in that: In the clustering identification step, the value of k is determined by the gap statistics method. The smallest k value that satisfies gap(k)>gap(k+1)-s_k is selected, s_k=sqrt(1+1 / n×sd_k), where sd_k is the standard deviation of gap(k).

4. The method for identifying and handling sensitive documents based on a clustering model according to claim 1, characterized in that: In the sensitive processing step, the storage path, file identifier, file clustering score, and file capture time of the file containing sensitive information are stored in the database. When a sensitive file is identified, an alarm message is generated and sent to a specific terminal.

5. A sensitive document identification and handling system based on a clustering model, characterized in that: include: Preprocessing module (01): is configured to sequentially perform text extraction, word segmentation and cleaning on the acquired data of the file to be detected; Clustering identification module (02): is configured to extract features from the cleaned text, cluster the extracted features, group similar files into the same group, and identify clusters containing sensitive information through similarity detection; Storage module (03): Used to store the original file, intermediate data during processing, and recognition results, etc. Alarm module (04): When a sensitive file is identified, it is responsible for generating alarm information and sending it to a specific terminal.

6. A sensitive document identification and processing system based on a clustering model according to claim 5, characterized in that: The preprocessing module (01) further includes: Text extraction unit (011): Responsible for extracting text content from different types of source files; Word segmentation unit (012): Performs word segmentation on the extracted text results; Cleaning unit (013): Cleans the segmented text to remove noise and irrelevant information.

7. A sensitive document identification and processing system based on a clustering model according to claim 5, characterized in that: The clustering identification module (02) includes: Feature extraction unit (021): Extracts features from the cleaned text data using the TF-IDF model; Clustering unit (022): The extracted features are clustered using the k-means algorithm; Sensitive detection unit (023): Identifies clusters containing sensitive information through similarity detection.

8. A sensitive document identification and processing system based on a clustering model according to claim 5, characterized in that: The storage module (03) includes: File storage server unit (031): Stores the original files; Message queue unit (032): Used to store information such as file identifier, file creator, file capture time and file storage path, and to transmit intermediate data during data processing; Database unit (033): Stores relevant information about the file that contains sensitive information, such as file storage path, file identifier, file cluster score and file capture time.

9. A sensitive document identification and processing system based on a clustering model according to claim 5, characterized in that: The alarm module (04) includes: Alarm generation unit (041): Generates corresponding alarm information based on the identification results of sensitive files; Terminal communication unit (042): Sends alarm information to a specific terminal; Alarm recording unit (043): Records alarm information for subsequent query and analysis.

10. A sensitive document identification and processing system based on a clustering model according to claim 5, characterized in that: The system stores a program that, when executed by a processor, implements the steps of the sensitive document identification and handling method based on a clustering model as described in any one of claims 1-4.

Images