A sensitive information detection method and system

Abstract

The application discloses a sensitive information detection method and system, and the method comprises the following steps: obtaining text information to be detected; performing detection processing on the text information to be detected to obtain target text detection information; the target text detection information comprises first target detection information, second target detection information and / or third target detection information; the first target detection information comprises L first target detection word information; the second target detection information comprises M second target detection word information; the third target detection information comprises N third target detection word information; and performing optimization processing on the target text detection information to obtain target detection result information.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of information detection, and in particular to a sensitive information detection method and system. BACKGROUND

[0002] In the digital era, the management and protection of sensitive information is of great importance, but with the explosive growth of its types and quantities, the sensitive information detection technology is facing challenges. On the one hand, the improper dissemination of sensitive information can bring many risks, and enterprises and organizations also need to accurately identify sensitive information. On the other hand, the existing sensitive information detection technology has limitations, such as poor quality of sensitive data sets, insufficient flexibility of traditional rule detection methods, high computational cost and poor adaptability of deep learning models, etc. Therefore, a sensitive information detection method and system are provided to improve the accuracy and efficiency of sensitive information identification and improve the accuracy and comprehensiveness of sensitive information identification. SUMMARY

[0003] The technical problem to be solved by the present application is to provide a sensitive information detection method and system to improve the accuracy and efficiency of sensitive information identification and improve the accuracy and comprehensiveness of sensitive information identification.

[0004] To solve the above technical problems, the first aspect of the embodiment of the present application discloses a sensitive information detection method, the method comprising:

[0005] obtaining text information to be detected;

[0006] detecting and processing the text information to be detected to obtain target text detection information; the target text detection information includes first target detection information, and / or second target detection information, and / or third target detection information; the first target detection information includes L first target detection word information; the second target detection information includes M second target detection word information; and the third target detection information includes N third target detection word information;

[0007] optimizing the target text detection information to obtain target detection result information.

[0008] The second aspect of the embodiment of the present application discloses a sensitive information detection system, the system comprising:

[0009] an acquisition module for acquiring text information to be detected;

[0010] The second processing module is configured to perform detection processing on the text information to be detected to obtain target text detection information; the target text detection information comprises first target detection information, and / or second target detection information, and / or third target detection information; the first target detection information comprises L first target detection word information; the second target detection information comprises M second target detection word information; and the third target detection information comprises N third target detection word information.

[0011] The third processing module is configured to perform optimization processing on the target text detection information to obtain target detection result information.

[0012] The third aspect of the present application discloses another sensitive information detection system, and the system comprises:

[0013] A memory in which executable program codes are stored;

[0014] A processor coupled with the memory;

[0015] The processor invokes the executable program codes stored in the memory to execute part or all of the steps of the sensitive information detection method disclosed in the first aspect of the present application.

[0016] The fourth aspect of the present application discloses a computer readable storage medium, and the computer readable storage medium stores computer instructions; when the computer instructions are invoked, part or all of the steps of the sensitive information detection method disclosed in the first aspect of the present application are executed. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows; obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0018] Figure 1 is a scene schematic diagram of the sensitive information detection system provided by the embodiments of the present application;

[0019] Figure 2 is a flow schematic diagram of the sensitive information detection method disclosed by the embodiments of the present application;

[0020] Figure 3 is a structure schematic diagram of the sensitive information detection system disclosed by the embodiments of the present application;

[0021] Figure 4 is a structure schematic diagram of another sensitive information detection system disclosed by the embodiments of the present application. DETAILED DESCRIPTION

[0022] In order to make the technical personnel in the art better understand the present application, 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 those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0023] The terms "first", "second", and the like in the specification and claims of the present application and the above-described drawings are used to distinguish different objects, rather than to describe a particular order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to the process, method, product, or device.

[0024] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearance of the phrase in various places in the specification does not necessarily all refer to the same embodiment, nor is it necessarily mutually exclusive of other embodiments. It is explicitly and implicitly understood that the embodiments described herein can be combined with other embodiments.

[0025] In this application, the word "exemplary" is used to mean "serving as an example, instance, or illustration." Any implementation described as exemplary in this application is not necessarily to be construed as preferred or advantageous over other implementations. The following description is presented to enable any person skilled in the art to make and use the application. In the following description, for purposes of explanation, specific details are set forth. It is apparent to those skilled in the art that the present application can be practiced without using these specific details. In other instances, well-known structures and processes are not elaborated in order not to obscure the description of the application with unnecessary details. Thus, the present application is not intended to be limited by the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein.

[0026] It should be noted that the method of the present application is executed in a computer device, and the processing objects of each computer device exist in the form of data or information, such as time, which is essentially time information. It can be understood that if the size, quantity, position, etc. are mentioned in subsequent embodiments, they all exist in the form of corresponding data for processing by the computer device, and specific details are not described here.

[0027] It should be noted that the artificial intelligence related technologies involved in the present application are briefly described. Artificial intelligence (AI) is the use of digital computers or digital computer controlled machines to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology of computer science, which attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines, so that machines have the functions of perception, reasoning and decision-making.

[0028] Artificial intelligence technology is a comprehensive discipline, involving a wide range of fields, both hardware and software technologies. Artificial intelligence basic technologies generally include technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc.

[0029] Computer vision (CV) is a science that studies how to make machines "see". Further, it refers to using cameras and computers to replace human eyes to identify and measure targets, and further to do image processing, so that the computer processing becomes more suitable for human eye observation or image transmission to instrument detection. As a scientific discipline, computer vision researches related theories and technologies, trying to establish artificial intelligence systems that can obtain information from images or multidimensional data. Computer vision technology usually includes image processing, image recognition, image semantic understanding, image retrieval, OCR, video processing, video semantic understanding, video content / behavior recognition, three-dimensional object reconstruction, 3D technology, virtual reality, augmented reality, simultaneous localization and mapping, etc. It also includes common face recognition, fingerprint recognition and other biometric identification technologies.

[0030] Single modal information is only one type of data, such as text, image, audio, video, electromagnetic signal, etc. Multi-modal information is data information including at least two single modal information. Further, multi-modal information is suitable for complex tasks that require the integration of multiple information sources, such as sentiment analysis, robot interaction, autonomous driving, etc. By integrating information from multiple modalities, higher performance and accuracy can usually be achieved in tasks.

[0031] A large model refers to an artificial neural network model with a very large number of parameters. In the field of artificial intelligence, a large model usually refers to a model with hundreds of millions to trillions of parameters. The model usually needs to be trained on a large-scale dataset and requires a large amount of computing resources for optimization and adjustment. Large models are usually used to solve complex natural language processing, computer vision, and speech recognition tasks. Generative AI is an AI that can create new content and ideas, including conversations, stories, images, videos, and music. In the embodiments of the present application, the large model can be a large language model such as ChatGPT, BERT, XLNet, Zhibu model, Claude, Moonshot AI model, ChatGLM model, Tongwen Qianyi model, MiniMax model, Xinghuo model, Llama model, 360GPT model, Qwen model, Baichuan model, Yunque model, vivoLM model, and Wenxin Yiyang, etc. The embodiments of the present application are not limited.

[0032] The embodiments of the present application provide a sensitive information detection method, system, computer device and computer readable storage medium, which are described in detail below.

[0033] Please refer to Figure 1 , Figure 1 The sensitive information detection system provided by the embodiments of the present application is a scene schematic diagram, which can include a computer device 100, and the computer device 100 is integrated with a sensitive information detection system, such as Figure 1 computer device in the prior art.

[0034] In the embodiments of the present application, the computer device 100 is mainly used to obtain text information to be detected;

[0035] detecting and processing the text information to be detected to obtain target text detection information; the target text detection information includes first target detection information, and / or second target detection information, and / or third target detection information; the first target detection information includes L first target detection word information; the second target detection information includes M second target detection word information; and the third target detection information includes N third target detection word information;

[0036] optimizing the target text detection information to obtain target detection result information.

[0037] It can improve the accuracy and efficiency of sensitive information recognition, and improve the accuracy and comprehensiveness of sensitive information recognition.

[0038] In this embodiment, the computer device 100 can be a standalone server, a server network, or a server cluster. For example, the computer device 100 described in this embodiment includes, but is not limited to, a computer, a network host, a single network server, a set of multiple network servers, or a cloud server composed of multiple servers. The cloud server is composed of a large number of computers or network servers based on cloud computing.

[0039] It is understood that the computer device 100 used in the embodiments of this application can be a device that includes both receiving and transmitting hardware, that is, a device having receiving and transmitting hardware capable of performing bidirectional communication on a bidirectional communication link. Such a device may include: cellular or other communication devices having a single-line display, a multi-line display, or a cellular or other communication device without a multi-line display. Specifically, the computer device 100 may be a desktop terminal or a mobile terminal, and may also be one of a mobile phone, tablet computer, laptop computer, etc.

[0040] Those skilled in the art will understand that Figure 1 The application environment shown is merely one application scenario of the solution in this application and does not constitute a limitation on the application scenario of the solution in this application. Other application environments may include those that are more specific to this application. Figure 1 The number of computer devices shown is more or less, for example Figure 1 Only one computer device is shown in the document. It is understood that the sensitive information detection system may also include one or more other services, which are not specified here.

[0041] In addition, such as Figure 1 As shown, the sensitive information detection system may also include a memory 200 for storing data, such as image data and location information.

[0042] It should be noted that, Figure 1 The schematic diagram of the sensitive information detection system shown is merely an example. The sensitive information detection system and scenario described in this application are for the purpose of more clearly illustrating the technical solutions of this application and do not constitute a limitation on the technical solutions provided in this application. As those skilled in the art will know, with the evolution of sensitive information detection systems and the emergence of new business scenarios, the technical solutions provided in this application are also applicable to similar technical problems.

[0043] This invention discloses a sensitive information detection method and system that improves the accuracy and efficiency of sensitive information identification, and enhances the accuracy and comprehensiveness of sensitive information identification. These will be described in detail below.

[0044] Example 1

[0045] Please refer to Figure 2 , Figure 2 is a flowchart of a sensitive information detection method disclosed by an embodiment of the present application. Wherein, Figure 2 The sensitive information detection method described is applied to a management system, such as a local server or a cloud server for management, etc., which is not limited by the present application. As Figure 2 shown, the sensitive information detection method can include the following operations:

[0046] 101, obtaining text information to be detected.

[0047] 102, detecting the text information to be detected to obtain target text detection information.

[0048] In the present application, the target text detection information includes first target detection information, and / or second target detection information, and / or third target detection information; the first target detection information includes L first target detection word information; the second target detection information includes M second target detection word information; and the third target detection information includes N third target detection word information.

[0049] 103, optimizing the target text detection information to obtain target detection result information.

[0050] In the present application, the target order is used to indicate the picking of the target goods for delivery.

[0051] It should be noted that L, M and N are positive integers not less than 0, which are not limited by the present application.

[0052] It should be noted that the sensitive information detection method of the present application can realize efficient and accurate sensitive information detection, i.e. combining the rapid screening based on rules and the deep semantic analysis ability of the deep learning model, and making improvements for the pain points in the prior art, by constructing a high-quality sensitive data set and fusing the rule engine and deep learning, the system can quickly adapt to the changing context and emerging sensitive information types, while significantly reducing the computing resource consumption in large-scale data processing, providing an efficient and flexible solution for sensitive information detection in complex context, which is not limited by the present application. Further, through various forms of sensitive word detection, the high precision of rule recognition and the high recall rate of deep learning recognition can be combined to ensure comprehensive coverage, which is not limited by the present application.

[0053] It should be noted that the text information to be detected can be obtained by web crawler or input by user, which is not limited by the present application.

[0054] It should be noted that the above target detection result information can be used for text content deletion of the to-be-detected text information to optimize and improve the text, so as to avoid the violation of the text expression due to the sensitive word, or find the violation expression problem of the text, and the embodiments of the present application are not limited.

[0055] It can be seen that the sensitive information detection method described in the embodiments of the present application is beneficial to improve the accuracy and efficiency of sensitive information recognition, and improve the accuracy and comprehensiveness of sensitive information recognition.

[0056] In an optional embodiment, the above step of detecting the to-be-detected text information to obtain target text detection information includes:

[0057] In response to the selection operation of the user, the detection type information is obtained;

[0058] It is judged whether the detection type information is the first detection type, and a type judgment result is obtained;

[0059] When the type judgment result is yes, the to-be-detected text information is detected based on the rule to obtain the first target detection information and the second target detection information;

[0060] The to-be-detected text information is detected by using a target information detection model to obtain third target detection information; the target information detection model includes a first detection model, a second detection model and a third detection model;

[0061] When the type judgment result is no, the to-be-detected text information is detected based on the rule to obtain the first target detection information and the second target detection information.

[0062] It should be noted that the above first detection type represents that the to-be-detected text information is detected by using the target information detection model, that is, before the to-be-detected sensitive word text is processed, the opinion of the user is inquired first, and the user performs a selection operation, if the user selects to use the deep model, the text will enter the recognition link based on the target information detection model; if the user selects not to use the target information detection model, the detection step based on the target information detection model will be skipped, and the subsequent process will be directly entered. After the above two specific detection methods in the rule-based detection branch and the target information detection model (if selected to be used), the detection results obtained by each branch will be merged, and the sensitive information found by different detection methods will be summarized together, which can ensure the comprehensive coverage of sensitive word detection.

[0063] It should be noted that the training sample set for the target information detection model is obtained by the following method:

[0064] (1) Construct a sensitive word library: construct a custom sensitive word library after data cleaning and classification.

[0065] (2) Constructing training sample set (annotated data set and test data set): Based on the sensitive word library, annotated data and test data are generated respectively, and the test data contains a certain proportion of positive and negative examples.

[0066] Further, the construction of the sensitive word library is divided into the following steps:

[0067] (a) Data collection and cleaning: Use the collected sensitive words to clean the data, remove duplicate, invalid and noise data, and ensure the accuracy and integrity of the data.

[0068] (b) Sensitive word classification: classify sensitive words to ensure that each category has clear definitions and standards for constructing sensitive word libraries and generating sentences containing sensitive words.

[0069] (c) Sentence evaluation: According to each classification, make sentences containing multiple classified sensitive words, and use them as training and test data. This helps improve the model's ability to recognize different contexts and expressions.

[0070] Further, the process of the target information detection model is as follows:

[0071] 1. Training data preparation

[0072] (A) A labeled JSON data file, such as output.json, needs to be prepared, which should be organized in the following format:

[0073]

[0074]

[0075] text: represents the text to be annotated.

[0076] label: annotation information, including category and corresponding entity index range.

[0077] 2. Training data preprocessing

[0078] (B) Expand the annotated data

[0079] Use the script to preprocess the JSON data and convert it to a flat format for subsequent processing. After the script is executed, the corresponding train.json file will be generated.

[0080] (C) Prepare the data directory

[0081] Create two sub-folders under the data directory, ori_data (for storing the original data file train.json) and ner_data (storing the preprocessed training file)

[0082] (D) Generate training data

[0083] Execute the data processing script to generate the model training required files dev.txt, labels.txt and train.txt under the ori_data folder.

[0084] 3、Training

[0085] (E) Create a model save directory (such as checkpoint / sensitive).

[0086] (F) Modify the dataset path and name in the training script.

[0087] (G) Execute the training script to start model training. After training is complete, the model file will be saved in the checkpoint directory.

[0088] It should be noted that the above rule-based detection processing of the to-be-detected text information is a fast screening mechanism based on rules, and a multi-level rule engine is designed to support customization and configuration of rules and sensitive word libraries. The embodiments of the present application are not limited. Further, the rule-based detection processing has explicit recognition ability for personal sensitive information with fixed format or obvious characteristics and sensitive word library, and the false positive rate is low. The embodiments of the present application are not limited.

[0089] It should be noted that the above detection processing of the to-be-detected text information by the target information detection model can process complex semantics, rules difficult to define, and implicit expressions, and identify potential sensitive information. The embodiments of the present application are not limited.

[0090] It can be seen that the sensitive information detection method described in the embodiments of the present application is beneficial to improve the accuracy and efficiency of sensitive information recognition, and improve the accuracy and comprehensiveness of sensitive information recognition.

[0091] In another optional embodiment, the rule-based detection processing of the to-be-detected text information obtains first target detection information and second target detection information, including:

[0092] The to-be-detected text information is detected by using a regular expression to obtain the first target detection information;

[0093] The to-be-detected text information is detected by using a regular expression to obtain the first target detection information;

[0094] It should be noted that the above detection and processing of the to-be-detected text information by using the regular expression is based on rich and accurate regular expression configuration, which can quickly locate and mark various sensitive information with high accuracy, effectively guarantee information security and compliance, meet the strict requirements of sensitive information identification in various scenarios, and provide comprehensive, reliable and efficient sensitive information protection barrier. The embodiments of the present application are not limited. Further, the setting of the regular expression for feature description and matching of various possible sensitive information is as follows:

[0095] Regular expression configuration: configured through a yaml format file, and direct modification in the system is also supported. The configuration file style is as follows:

[0096]

[0097]

[0098] Modify the regular expression: the json style can be used for modification in the system.

[0099] Add regular expression: when adding rules in the system, only the corresponding content needs to be input, and the system will automatically convert it to yaml format and write it to the configuration file, simplifying the addition process and reducing the operation difficulty.

[0100] Add sensitive word library: select the import sensitive word library function to add the sensitive word library. Support txt format file, each line of content corresponds to a sensitive word. When importing, the word library will be imported into the corresponding category rule. If the category does not exist, it will be automatically created; if the category already exists, it will be merged after de-duplication with the existing category to ensure the accuracy and integrity of the word library. In addition, the system also supports editing operation of the word library.

[0101] It should be noted that the above sensitive information matching detection and processing of the to-be-detected text information is through the preset vector of the sensitive word to quickly filter out a large amount of non-sensitive information, greatly improving the sensitive word detection speed, solving the problem of lack of fast preprocessing means in the prior art, and the embodiments of the present application are not limited.

[0102] It can be seen that the sensitive information detection method described in the embodiments of the present application is beneficial to improve the accuracy and efficiency of sensitive information identification, and improve the accuracy and comprehensiveness of sensitive information identification.

[0103] In another optional embodiment, the sensitive information matching detection and processing of the to-be-detected text information obtains second target detection information, including:

[0104] The to-be-detected text information is processed by word segmentation to obtain first processed text information; the first processed text information includes a plurality of first text word information;

[0105] For any first text word information in the first processed text information, the first text word information and the sensitive word symbol vector information in the sensitive word symbol information are calculated and processed by using the association calculation model to obtain the word symbol association value information corresponding to the first text word information; the word symbol association value information includes a plurality of word symbol association values;

[0106] The association calculation model is:

[0107]

[0108] The CFGLZ represents the word symbol association value; the WBC and the YFC represent the first text word information and the sensitive word symbol vector information respectively; the qz1 and the qz2 represent the first weight coefficient and the second weight information respectively;

[0109] It is judged whether there is a word symbol association value greater than or equal to the association threshold value in the word symbol association value information, and an association value judgment result is obtained;

[0110] When the association value judgment result is yes, the first text word information is determined as a second target detection word information;

[0111] When the association value judgment result is no, the detection process corresponding to the first text word information is ended.

[0112] It should be noted that the first weight coefficient and the second weight information described above can be set by the user or can be preset by the system, and the value is a number between 0 and 1, which is not limited by the embodiments of the present application.

[0113] It should be noted that the sensitive word symbol vector information in the sensitive word symbol information described above represents the data after the sensitive words in the sensitive word library preset in the system are vectorized, and the vectorization can be based on the LSTM model or a large model, which is not limited by the embodiments of the present application.

[0114] It should be noted that the association threshold value described above can be set by the user or can be preset by the system, and the value is a number between 0 and 2, which is not limited by the embodiments of the present application.

[0115] It can be seen that the sensitive information detection method described in the embodiments of the present application is beneficial to improve the accuracy and efficiency of sensitive information recognition, and improve the accuracy and comprehensiveness of sensitive information recognition.

[0116] In yet another optional embodiment, a target information detection model is used to detect and process the to-be-detected text information to obtain third target detection information, which includes:

[0117] The to-be-detected text information is text preprocessed to obtain first text processing information;

[0118] The first text processing information is processed by bidirectional encoding using the first detection model to obtain second text processing information;

[0119] The second text processing information is processed by bidirectional feature extraction using a second detection model to obtain third text processing information; the second detection model includes a first detection submodel, a second detection submodel, and a third detection submodel;

[0120] The third text processing information is processed by detection and recognition using a third detection model to obtain third target detection information.

[0121] It should be noted that the above text preprocessing of the text information to be detected includes data cleaning and word segmentation processing, and the embodiments of the present application are not limited.

[0122] It should be noted that the bidirectional encoding of the first text processing information using the first detection model is through context-aware embedding of sensitive word text, understanding the subtle differences of words in the context, capturing complex semantic relationships in the text, solving the deficiencies of the prior art in processing polysemy and complex context, and improving detection accuracy. Its implementation can be based on a large model, or it can be implemented based on a BERT model, and the embodiments of the present application are not limited. Further, the first detection model mainly undertakes the task of deep semantic understanding and feature extraction of sensitive word text, deeply mines the semantic representation of words and sentences, generates high-quality word vectors containing multi-dimensional information such as semantics and syntax through bidirectional encoding of input sensitive word text, and fully considers the context information before and after the word. These word vectors provide a strong semantic feature basis for subsequent sensitive information judgment, so that the target information detection model can more accurately understand the semantic content of the sensitive word text, thereby effectively distinguishing between normal text and text containing sensitive information. For example, in identifying sensitive text involving implicit semantic expression, the first detection model can accurately capture the potential sensitive intent therein by virtue of its deep semantic understanding ability, greatly improving the accuracy and recall rate of sensitive information recognition, and enhancing the overall recognition effect of the system, and the embodiments of the present application are not limited.

[0123] It should be noted that the above-mentioned use of the second detection model to process the second text processing information for bidirectional feature extraction processing is to capture the long-distance dependency relationship in the sensitive word text, process the information flow in the long text, and enhance the context understanding capability to solve the problem that the existing single-directional LSTM model cannot be matched, and the embodiments of the application are not limited. Further, when the second detection model processes the sensitive information recognition in the sensitive word text sequence, each word is processed in turn according to the natural order of the sensitive word text during forward processing, and the information flow from the beginning to the end is accurately captured; during reverse processing, the words are processed from the end to the beginning to obtain the reverse information. By integrating the information in the two directions of the forward direction, the second detection model can generate a comprehensive feature representation for each word that contains rich context information. This feature representation is extremely critical in the sequence labeling task of the present application, because sensitive information recognition often needs to consider the order of the text sequence and the context dependency relationship. For example, when identifying a text containing sensitive words but depending on the context to judge its sensitivity, the feature representation generated by the second detection model can provide strong support for accurate judgment, effectively reduce the misjudgment or omission caused by insufficient context understanding, and improve the stability and reliability of the system for sensitive information recognition in complex text sequences, and the embodiments of the application are not limited.

[0124] It should be noted that the above-mentioned use of the third detection model to process the third text processing information for detection and recognition is to consider the correlation between sensitive word labels, and to ensure that the dependency relationship between sensitive word labels is effectively mapped through sequence labeling to improve the accuracy of sensitive word detection and solve the functions that the existing independent classifier does not have, and the embodiments of the application are not limited. Further, in the sequence labeling link of sensitive information recognition, each input sequence element needs to be assigned an accurate label, and these labels usually have specific constraint relationships. The third detection model can learn the transition probability between these labels, calculate the conditional probability of all possible label sequences by constructing a probability model, and then determine the label sequence with the highest probability as the output. The third detection model can also use the constraints between labels to correct the errors of individual label prediction. For example, when facing some text fragments that are easy to produce ambiguity, individual models may give incorrect label prediction, but the third detection model can correct it according to the constraint relationship between the labels, so that the final label sequence is more in line with the language rules and the requirements of the sensitive information recognition task, thereby effectively improving the accuracy of the sequence labeling task, ensuring that the entire sensitive information recognition system can output more accurate and reliable recognition results when processing text sequences, reducing the risk of false positives and omissions, and improving the overall performance and credibility of the system, and the embodiments of the application are not limited.

[0125] It can be seen that the sensitive information detection method described in the embodiment of the present application is beneficial to improving the accuracy and efficiency of sensitive information recognition, and improving the accuracy and comprehensiveness of sensitive information recognition.

[0126] In an optional embodiment, the second text processing information is processed by using a second detection model for bidirectional feature extraction, to obtain third text processing information, including:

[0127] The second text processing information is processed by using a first detection sub-model for bidirectional semantic feature extraction, to obtain first sub-text processing information; the first sub-text processing information represents bidirectional semantic features of forward propagation and backward propagation in the text;

[0128] The first sub-text processing information is processed by using a second detection sub-model for front-back bidirectional data splicing, to obtain second sub-text processing information;

[0129] The second sub-text processing information is processed by using a third detection sub-model for sequence labeling, to obtain the third text processing information.

[0130] It should be noted that the bidirectional semantic feature extraction of the second text processing information by using the first detection sub-model can be that a bidirectional text information capturing unit module (such as one constructed based on a convolution layer or a neural network) is used to capture single-directional transmission information of word vector information respectively, to obtain a hidden layer state, so as to realize the extraction of forward and backward semantic features, which is not limited in the embodiment of the present application.

[0131] It should be noted that the front-back bidirectional data splicing of the first sub-text processing information by using the second detection sub-model (such as a unit module constructed based on splicing operation) can be that the second detection sub-model is used to splice context information in forward and reverse directions, to realize complete and rich representation of forward and reverse information of the text sequence, which is not limited in the embodiment of the present application.

[0132] It should be noted that the sequence labeling of the second sub-text processing information by using the third detection sub-model (such as one constructed based on a softmax activation function) is to project the second sub-text processing information to a numerical interval of a text label with explicit meaning expression, to further identify whether a text word is a sensitive word, which is not limited in the embodiment of the present application.

[0133] It can be seen that the sensitive information detection method described in the embodiment of the present application is beneficial to improving the accuracy and efficiency of sensitive information recognition, and improving the accuracy and comprehensiveness of sensitive information recognition.

[0134] In another optional embodiment, the target text detection information is processed for optimization, to obtain target detection result information, including:

[0135] The target text detection information is processed by a union set to obtain optimized text detection information;

[0136] In the optimization time range, it is detected whether the user input text modification information is received, and detection result information is obtained;

[0137] When the detection result information is yes, the intersection information of the optimized text detection information and the detection result information is removed from the optimized text detection information to obtain target detection result information;

[0138] When the detection result information is no, the optimized text detection information is determined as the target detection result information.

[0139] It should be noted that the above union set processing of the target text detection information is to take the union set of the sensitive word information detected based on the rule-based sensitive word detection and the target information detection model, so as to ensure that no potential sensitive information is missed, and the embodiments of the present application are not limited.

[0140] It should be noted that the above optimization time range is a time within 1 time unit after the optimized text detection information is obtained, and further, the time unit can be one of 3 minutes, 5 minutes or 10 minutes, and the embodiments of the present application are not limited. Further, the time space given by 1 optimization time range, that is, enough modification checking time is left for the user, and long waiting time does not occur, so as to ensure the further optimization accuracy of the sensitive word and the detection efficiency, and the embodiments of the present application are not limited.

[0141] It should be noted that the above text modification information represents the non-sensitive word selected by the user, that is, the text word that needs to be removed from the target text detection information, so as to further accurately optimize the user-acceptable text representation, avoid the false sensitive word judgment caused by too strict sensitive word detection, and further improve the detection accuracy and efficiency of the sensitive word, and the embodiments of the present application are not limited.

[0142] It can be seen that the sensitive information detection method described in the embodiments of the present application is beneficial to improve the accuracy and efficiency of sensitive information recognition, and improve the accuracy and comprehensiveness of sensitive information recognition.

[0143] Embodiment two

[0144] Please refer to Figure 3 , Figure 3 is a structural schematic diagram of a sensitive information detection system disclosed by the embodiments of the present application. Among them, Figure 3 The system described can be applied to a management system, such as a local server or a cloud server for management, and the embodiments of the present application are not limited. As Figure 3 shown, the system can include:

[0145] The acquisition module 201 is configured to acquire text information to be detected.

[0146] The second processing module 202 is configured to perform detection processing on the text information to be detected to obtain target text detection information; the target text detection information includes first target detection information, and / or second target detection information, and / or third target detection information; the first target detection information includes L first target detection word information; the second target detection information includes M second target detection word information; and the third target detection information includes N third target detection word information.

[0147] The third processing module 203 is configured to perform optimization processing on the target text detection information to obtain target detection result information.

[0148] It can be seen that the sensitive information detection system described in the embodiments has the advantages of Figure 3 The sensitive information detection system described in the embodiments is beneficial to improving the accuracy and efficiency of sensitive information recognition, and improving the accuracy and comprehensiveness of sensitive information recognition.

[0149] In another optional embodiment, as shown in Figure 3 The detection processing on the text information to be detected to obtain the target text detection information includes:

[0150] The detection type information is obtained in response to a selection operation of a user.

[0151] The type judgment result is obtained by judging whether the detection type information is the first detection type.

[0152] When the type judgment result is yes, the first target detection information and the second target detection information are obtained by performing rule-based detection processing on the text information to be detected.

[0153] The third target detection information is obtained by performing detection processing on the text information to be detected by using a target information detection model; the target information detection model includes a first detection model, a second detection model, and a third detection model.

[0154] When the type judgment result is no, the first target detection information and the second target detection information are obtained by performing rule-based detection processing on the text information to be detected.

[0155] It can be seen that the sensitive information detection system described in the embodiments has the advantages of Figure 3 The sensitive information detection system described in the embodiments is beneficial to improving the accuracy and efficiency of sensitive information recognition, and improving the accuracy and comprehensiveness of sensitive information recognition.

[0156] In yet another optional embodiment, as shown in Figure 3 The rule-based detection processing on the text information to be detected to obtain the first target detection information and the second target detection information includes:

[0157] The regular expression is used to detect the text information to be detected, and first target detection information is obtained.

[0158] The sensitive information matching detection is performed on the text information to be detected, and second target detection information is obtained.

[0159] It can be seen that the described sensitive information detection system is beneficial to improving the accuracy and efficiency of sensitive information identification, and improving the accuracy and comprehensiveness of sensitive information identification. Figure 3 The described sensitive information detection system is beneficial to improving the accuracy and efficiency of sensitive information identification, and improving the accuracy and comprehensiveness of sensitive information identification.

[0160] In another optional embodiment, as shown in Figure 3 The sensitive information matching detection is performed on the text information to be detected, and second target detection information is obtained, including:

[0161] The text information to be detected is segmented, and first processing text information is obtained; the first processing text information includes a plurality of first text word information;

[0162] For any first text word information in the first processing text information, the first text word information and the sensitive word symbol vector information in the sensitive word symbol information are calculated and processed by using the association calculation model, and the word symbol association value information corresponding to the first text word information is obtained; the word symbol association value information includes a plurality of word symbol association values;

[0163] The association calculation model is:

[0164]

[0165] The CFGLZ represents the word symbol association value; the WBC and the YFC represent the first text word information and the sensitive word symbol vector information respectively; the qz1 and the qz2 represent the first weight coefficient and the second weight information respectively;

[0166] It is judged whether there is a word symbol association value greater than or equal to the association threshold value in the word symbol association value information, and an association value judgment result is obtained;

[0167] When the association value judgment result is yes, the first text word information is determined as a second target detection word information;

[0168] When the association value judgment result is no, the detection process corresponding to the first text word information is ended.

[0169] It can be seen that the described sensitive information detection system is beneficial to improving the accuracy and efficiency of sensitive information identification, and improving the accuracy and comprehensiveness of sensitive information identification. Figure 3 The described sensitive information detection system is beneficial to improving the accuracy and efficiency of sensitive information identification, and improving the accuracy and comprehensiveness of sensitive information identification.

[0170] In another optional embodiment, as shown in Figure 3As shown, the target information detection model is used to detect the text information to be detected, to obtain third target detection information, including:

[0171] The text information to be detected is pre-processed to obtain first text processing information;

[0172] The first text processing information is processed by a first detection model to obtain second text processing information;

[0173] The second text processing information is processed by a second detection model to obtain third text processing information; the second detection model includes a first detection sub-model, a second detection sub-model, and a third detection sub-model;

[0174] The third text processing information is processed by a third detection model to obtain third target detection information.

[0175] As can be seen, the sensitive information detection system described Figure 3 The sensitive information detection system described is beneficial to improving the accuracy and efficiency of sensitive information recognition, and improving the accuracy and comprehensiveness of sensitive information recognition.

[0176] In another optional embodiment, as shown Figure 3 The second text processing information is processed by a second detection model to obtain third text processing information, including:

[0177] The second text processing information is processed by a first detection sub-model to obtain first sub-text processing information; the first sub-text processing information represents the bidirectional semantic features of the text;

[0178] The first sub-text processing information is processed by a second detection sub-model to obtain second sub-text processing information;

[0179] The second sub-text processing information is processed by a third detection sub-model to obtain third text processing information.

[0180] As can be seen, the sensitive information detection system described Figure 3 The sensitive information detection system described is beneficial to improving the accuracy and efficiency of sensitive information recognition, and improving the accuracy and comprehensiveness of sensitive information recognition.

[0181] In another optional embodiment, as shown Figure 4 The target text detection information is processed to obtain target detection result information, including:

[0182] The target text detection information is processed to obtain optimized text detection information;

[0183] Detecting whether the user inputted text modification information in an optimization time range, obtaining detection result information;

[0184] When the detection result information is yes, removing intersection information of the optimization text detection information and the detection result information from the optimization text detection information, obtaining target detection result information;

[0185] When the detection result information is no, determining the optimization text detection information as the target detection result information.

[0186] It can be seen that the implementation Figure 4 The sensitive information detection system described is beneficial to improving the accuracy and efficiency of sensitive information identification, and improving the accuracy and comprehensiveness of sensitive information identification.

[0187] Embodiment three

[0188] Please refer to Figure 4 , Figure 4 is another structure diagram of a sensitive information detection system disclosed by the embodiment of the application. Among them, ​ The system described can be applied to a management system, such as a local server or a cloud server for management, etc., and the embodiment of the application is not limited. As ​ As shown in the figure, the system can include:

[0189] The memory 301 stores executable program codes;

[0190] The processor 302 is coupled to the memory 301;

[0191] The processor 302 calls the executable program codes stored in the memory 301, and is used to execute the steps in the sensitive information detection method described in embodiment one.

[0192] Embodiment four

[0193] The embodiment of the application discloses a computer readable storage medium which stores a computer program for electronic data exchange, wherein the computer program causes the computer to execute the steps in the sensitive information detection method described in embodiment one.

[0194] Embodiment five

[0195] The embodiment of the application discloses a computer program product, which includes a non-transitory computer readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute the steps in the sensitive information detection method described in embodiment one.

[0196] The system embodiments described above are only illustrative, wherein the modules illustrated as separate components can or can not be physically separated, and the components illustrated as modules can or can not be physical modules, i.e., can be located in one place or distributed to multiple network modules. Part or all of the modules can be selected to achieve the purposes of the embodiments according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0197] Through the specific description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be realized by means of software and necessary general hardware platform, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of software products, and the computer software products can be stored in a computer readable storage medium, including a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disk storage, a magnetic disk storage, a magnetic tape storage, or any other computer readable medium that can be used to carry or store data.

[0198] Finally, it should be noted that: the sensitive information detection method and system disclosed by the embodiments of the present application are only the preferred embodiments of the present application, and are used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that; it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A sensitive information detection method, characterized by, The method comprises: obtaining to-be-detected text information; detecting the to-be-detected text information to obtain target text detection information; the target text detection information comprises first target detection information, and / or second target detection information, and / or third target detection information; the first target detection information comprises L first target detection word information; the second target detection information comprises M second target detection word information; the third target detection information comprises N third target detection word information; the first target detection information is a sensitive word detected from the to-be-detected text information based on a regular expression; the second target detection information is a sensitive word detected from the to-be-detected text information through sensitive information matching detection; and the third target detection information is a sensitive word detected from third text processing information based on a third detection model; optimizing the target text detection information to obtain target detection result information.

2. The sensitive information detection method of claim 1, wherein, The method comprises: obtaining detection type information in response to a selection operation of a user; judging whether the detection type information is a first detection type to obtain a type judgment result; when the type judgment result is yes, performing rule-based detection on the to-be-detected text information to obtain the first target detection information and the second target detection information; performing detection on the to-be-detected text information by using a target information detection model to obtain the third target detection information; the target information detection model comprises a first detection model, a second detection model, and a third detection model; the first detection model is implemented based on a large model or a BERT model, mainly undertakes the task of deep semantic understanding and feature extraction of sensitive word text, deeply mines the semantic representation of words and sentences, generates high-quality word vectors containing multi-dimensional information such as semantics and syntax by bidirectional encoding of input sensitive word text, and fully considers the context information before and after the word; the second detection model accurately captures the information flow from the beginning to the end when processing the sensitive word text sequence in sensitive information recognition; the third detection model considers the correlation between sensitive word labels, ensures the effective mapping of the dependency relationship between sensitive word labels by the way of sequence labeling, and assigns an accurate label to each input sequence element in the sequence labeling link of sensitive information recognition, and there is usually a specific constraint relationship between these labels; when the type judgment result is no, performing rule-based detection on the to-be-detected text information to obtain the first target detection information and the second target detection information.

3. The sensitive information detection method of claim 2, wherein, The method comprises: performing rule-based detection on the to-be-detected text information to obtain the first target detection information and the second target detection information. Detecting the to-be-detected text information by using a regular expression to obtain the first target detection information; Performing sensitive information matching detection on the to-be-detected text information to obtain the second target detection information.

4. The method of claim 3, wherein, The sensitive information matching detection on the to-be-detected text information to obtain the second target detection information comprises: Performing word segmentation on the to-be-detected text information to obtain first processed text information; the first processed text information comprises a plurality of first text word information; For any first text word information in the first processed text information, performing calculation processing on the first text word information and sensitive word symbol vector information in sensitive word symbol information by using an association calculation model to obtain word symbol association value information corresponding to the first text word information; the word symbol association value information comprises a plurality of word symbol association values; The association calculation model is: wherein, CFGLZ represents the word symbol association value; WBC and YFC represent the first text word information and the sensitive word symbol vector information respectively; qz1 and qz2 represent a first weight coefficient and a second weight information respectively; determining whether there is a word symbol association value greater than or equal to an association threshold value in the word symbol association value information to obtain an association value judgment result; when the association value judgment result is yes, the first text word information is determined as a second target detection word information; when the association value judgment result is no, the detection process corresponding to the first text word information is ended.

5. The method of claim 2, wherein, The detection processing on the to-be-detected text information by using a target information detection model to obtain the third target detection information comprises: performing text preprocessing on the to-be-detected text information to obtain first text processing information; performing bidirectional encoding processing on the first text processing information by using the first detection model to obtain second text processing information; performing bidirectional feature extraction processing on the second text processing information by using the second detection model to obtain third text processing information; the second detection model comprises a first detection submodel, a second detection submodel and a third detection submodel; the first detection submodel performs bidirectional semantic feature extraction on the second text processing information by using a bidirectional text information capture unit module to capture single-directional transmission information of word vector information to obtain a hidden layer state, thereby realizing forward and backward semantic feature extraction, wherein the first detection submodel is constructed based on a convolution layer or a neural network; the second detection submodel performs front and back bidirectional data splicing processing on the first subtext processing information by using the second detection submodel to perform forward and reverse splicing on context information, to realize complete and rich representation of forward and reverse information of a text sequence, wherein the second detection submodel is a unit module constructed based on a splicing operation; the third detection submodel performs sequence labeling processing on the second subtext processing information by projecting the second subtext processing information to a numerical interval of a text label with explicit meaning expression, and then identifying whether a text word is a sensitive word; the third detection submodel is constructed based on a softmax activation function; The third detection model is used for detecting and identifying the third text processing information, to obtain the third target detection information.

6. The method of claim 5, wherein, The second detection model is used for bidirectional feature extraction processing on the second text processing information, to obtain third text processing information, including: The first detection sub-model is used for bidirectional semantic feature extraction on the second text processing information, to obtain first sub-text processing information; the first sub-text processing information represents bidirectional semantic features of forward propagation and backward propagation in the text; The second detection sub-model is used for front-back bidirectional data splicing processing on the first sub-text processing information, to obtain second sub-text processing information; The third detection sub-model is used for sequence labeling processing on the second sub-text processing information, to obtain third text processing information.

7. The method of claim 1, wherein, The target text detection information is optimized, to obtain target detection result information, including: The target text detection information is processed by set union, to obtain optimized text detection information; It is detected whether user input text modification information is received within an optimization time range, to obtain detection result information; When the detection result information is yes, the intersection information of the optimized text detection information and the detection result information is removed from the optimized text detection information, to obtain target detection result information; When the detection result information is no, the optimized text detection information is determined as the target detection result information.

8. A sensitive information detection system characterized by, The system includes: An acquisition module is configured to acquire text information to be detected. A second processing module is configured to detect the text information to be detected, to obtain target text detection information; the target text detection information includes first target detection information, and / or second target detection information, and / or third target detection information; the first target detection information includes L first target detection word information; the second target detection information includes M second target detection word information; the third target detection information includes N third target detection word information; the first target detection information is sensitive words detected based on a regular expression on the text information to be detected; the second target detection information is sensitive words detected by sensitive information matching detection on the text information to be detected; and the third target detection information is sensitive words detected based on a third detection model on third text processing information. A third processing module is configured to optimize the target text detection information, to obtain target detection result information.

9. A sensitive information detection system characterized by, The system includes: A memory storing executable program codes; A processor coupled with the memory; The processor invokes the executable program codes stored in the memory, to execute the sensitive information detection method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions, which are invoked to execute the sensitive information detection method according to any one of claims 1-7.

Images