Campus network security emergency management method and system based on public opinion monitoring

Abstract

The application provides a campus network security emergency management method and system based on public opinion monitoring, relates to the technical field of data management, and comprises the following steps: performing data screening on campus network information data to obtain public opinion data; preprocessing the public opinion data to output denoised text content, publishing time sequence fragments and public opinion scene identification; calculating a text basic risk value and a propagation characteristic compensation value according to the denoised text content and the publishing time sequence fragments; obtaining a corresponding single public opinion risk benchmark value according to the text basic risk value and the propagation characteristic compensation value; slidingly aggregating the single public opinion risk benchmark values under the same public opinion scene identification with a fixed time window, calculating the in-window peak risk and the cross-scene diffusion degree of the corresponding fixed time window, and further generating a time period public opinion heat value, and predicting the public opinion outbreak probability of a future time period based on a long short-term memory network, and executing corresponding emergency processing instructions in a time period before the public opinion outbreak probability exceeds a preset threshold.

Description

Technical Field

[0001] This invention relates to the field of data management technology, specifically to a campus network security emergency management method and system based on public opinion monitoring. Background Technology

[0002] With the increasing networking of campuses, campus network information is characterized by rapid dissemination, wide coverage, and complex content. Various campus public opinion incidents occur frequently, and if they are not detected and effectively controlled in a timely manner, they can easily trigger campus network security risks, affecting the normal teaching and living order on campus. Currently, campus network security management suffers from insufficient targeting and limited screening precision in the collection of public opinion data, making it difficult to accurately capture public opinion information with security risks. At the same time, the processing of public opinion data often remains superficial, lacking accurate quantification of public opinion risks and in-depth analysis of dissemination characteristics. This makes it impossible to effectively predict the trend of public opinion outbreaks, resulting in delayed emergency response and difficulty in quickly responding to campus network security risks caused by various public opinion incidents. Existing management methods can no longer meet the actual needs of campus network security emergency management.

[0003] Therefore, we now provide a campus network security emergency management method and system based on public opinion monitoring. Summary of the Invention

[0004] To address the aforementioned technical problems, the present invention aims to provide a campus network security emergency management method and system based on public opinion monitoring.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a campus network security emergency management system based on public opinion monitoring, the system comprising: a data acquisition module, a data preprocessing module, a data processing module, a data analysis module, and an emergency response module; The data acquisition module is used to collect campus network information data from the corresponding campus network information sharing platform through the API interface; The data filtering module is used to filter campus network information data to obtain public opinion data; The data preprocessing module preprocesses public opinion data and outputs denoised text content, release time sequence fragments, and public opinion scene identifiers. The data processing module is used to input the denoised text content into the lightweight semantic coding model and obtain the basic risk value of the text based on the campus risk dictionary; calculate the short-term propagation growth rate based on the release time sequence segment, and then obtain the propagation feature compensation value; and obtain the single public opinion risk benchmark value of the corresponding public opinion data according to the basic risk value of the text and the propagation feature compensation value. The data analysis module is used to aggregate single public opinion risk benchmark values ​​under the same public opinion scenario identifier in a fixed time window, and calculate the peak risk within the corresponding fixed time window and the cross-scenario diffusion degree, thereby generating the public opinion heat value for the time period. The emergency response module, based on the public opinion heat value of the time period and the prediction of the probability of public opinion outbreak in future time periods based on the long short-term memory network, executes the corresponding emergency response instructions in the time period before the probability of public opinion outbreak exceeds the preset threshold.

[0006] Furthermore, the process of filtering campus network information data to obtain public opinion data includes: The data filtering includes primary data filtering and secondary data filtering; wherein, primary data filtering is to perform preliminary filtering on the campus network information data, including sensitive keyword filtering, sensitive keyword homophone filtering, and sensitive keyword letter abbreviation filtering, thereby obtaining preliminary filtered data; Secondary data filtering involves semantic feature filtering of the initially filtered data to obtain public opinion data.

[0007] Furthermore, the process of primary and secondary data filtering includes: A data filtering process includes setting up a sensitive keyword filter set, a sensitive keyword homophone filter set, and a sensitive keyword letter abbreviation filter set; The collected campus network information data set is traversed, and the text matching degree of each piece of campus network information data in the sensitive keyword filtering set, the sensitive keyword homophone filtering set, and the sensitive keyword letter abbreviation filtering set is calculated; if the text matching degree is greater than or equal to the standard matching degree threshold, the preliminary filtered data form the preliminary filtered dataset. Extract the semantic feature vectors corresponding to the preliminary screening data from the preliminary screening dataset, and calculate the corresponding public opinion confidence score based on the semantic feature vectors; set a public opinion judgment threshold, and if the public opinion confidence score of the corresponding preliminary screening data is greater than the public opinion judgment threshold, it is recorded as public opinion data, otherwise it is recorded as normal data.

[0008] Furthermore, the process of preprocessing public opinion data to output denoised text content, release time sequence fragments, and public opinion scene identifiers includes: Based on the chronological order of the input public opinion data, a text cleaning operation is performed to remove HTML tags, special characters, and stop words, generating standardized denoised text content. Extract timestamp information from the denoised text content, construct a time series vector, and discretize the time series vector according to a preset time granularity to generate a release time sequence segment; Extract the keyword feature vectors from the denoised text content and input them into the campus public opinion scene classifier. Calculate the probability distribution of each preset campus public opinion scene and select the scene corresponding to the highest probability distribution as the public opinion scene identifier.

[0009] Furthermore, the process of inputting the denoised text content into a lightweight semantic coding model and obtaining the basic risk value of the text based on a campus risk dictionary includes: The denoised text content is input into a lightweight semantic coding model, which outputs a text semantic embedding vector. The denoised text content is then segmented to obtain a word sequence. Based on a campus risk dictionary and an indicator function, the word-level risk weight of the word sequence is calculated. The campus risk dictionary includes risk words, risk word embedding vectors of the risk words, and corresponding preset risk weights. Based on the text semantic embedding vector and the risk word embedding vector, and based on the maximum value of the cosine similarity function, the semantic similarity gain coefficient is calculated; The basic risk value of the text is calculated based on the word-level risk weight, the text semantic embedding vector, and the semantic similarity gain coefficient, and based on the Sigmoid function.

[0010] Furthermore, the process of obtaining the propagation feature compensation value includes: Based on the aforementioned release time sequence segment, the number of public opinion releases within a unit time granularity is counted to generate a propagation count sequence; Calculate the first-order difference between the release volume at the current time granularity and the previous time granularity, denoted as the short-term propagation growth rate; normalize the short-term propagation growth rate, and calculate the corresponding propagation characteristic compensation value based on the linear rectification function and the attenuation factor; Based on the aforementioned dissemination characteristic compensation value and the basic text risk value, a single public opinion risk benchmark value is obtained.

[0011] Furthermore, the process of calculating the peak risk within a fixed time window and the cross-scenario diffusion degree, and then generating the public opinion heat value for a given period, includes: Set a fixed time window length; aggregate all single public opinion risk benchmark values ​​belonging to the same public opinion scenario identifier within the fixed time window, calculate the peak risk within the window, and normalize it; The number of public opinion scene identifiers associated with the public opinion data within a fixed time window is counted, and then the cross-scene diffusion degree is calculated; the cross-scene diffusion degree and the peak risk within the window are weighted and summed to obtain the public opinion heat value for the time period.

[0012] Furthermore, the process of predicting the probability of a future public opinion outburst based on the public opinion heat value of the aforementioned time period and using a long short-term memory network includes: Construct a sequence of public opinion heat values ​​for historical periods and input it into a trained long short-term memory network; output the latent feature vector for future periods through the hidden layer state propagation of the long short-term memory network. The hidden feature vector is input into a fully connected regression layer, and the probability of a public opinion outbreak in the future is calculated using the Softmax function. Execute corresponding emergency response instructions in the period before the probability of a public opinion outbreak exceeds a preset threshold.

[0013] A second aspect of the present invention also provides a campus network security emergency management method based on public opinion monitoring, the method comprising: Through API interfaces, collect campus network information data from the corresponding campus network information sharing platform; filter the campus network information data to obtain public opinion data; Preprocess the public opinion data to output denoised text content, release time sequence fragments, and public opinion scene identifiers; The denoised text content is input into a lightweight semantic coding model, and a basic risk value is obtained based on a campus risk dictionary. Based on the release time sequence, the short-term propagation growth rate is calculated, and then the propagation feature compensation value is obtained. Based on the basic risk value of the text and the propagation feature compensation value, the single public opinion risk benchmark value of the corresponding public opinion data is obtained. By using a fixed time window to aggregate the benchmark value of a single public opinion risk under the same public opinion scenario identifier, and calculating the peak risk within the corresponding fixed time window and the cross-scenario diffusion, a public opinion heat value for a given period is generated. Based on the public opinion heat value of the specified time period, and based on the prediction of the probability of public opinion outbreak in future time periods using a long short-term memory network, the corresponding emergency handling instructions are executed in the time period before the probability of public opinion outbreak exceeds a preset threshold.

[0014] Compared with existing technologies, the beneficial effects of this invention are as follows: This system and method, by constructing a complete public opinion monitoring and emergency management system, realizes closed-loop management of the entire process of campus network public opinion from collection, screening, preprocessing to risk analysis and emergency response, effectively improving the accuracy and efficiency of campus network public opinion monitoring. Through multi-dimensional data screening, it can accurately identify campus public opinion data, avoid interference from irrelevant information, and provide reliable data support for subsequent risk analysis. By combining a lightweight semantic coding model with a campus risk dictionary, it achieves accurate quantification of public opinion risks, and further enhances the scientific nature of risk assessment by combining propagation characteristic compensation values. Through fixed-time-window aggregation analysis and long short-term memory network prediction, it can predict the trend of public opinion outbreaks in advance, allowing sufficient time for emergency response. By promptly executing emergency response instructions, it effectively curbs the spread of public opinion, reduces campus network security risks, ensures the safety and stability of the campus network environment, and maintains normal campus order. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0016] Figure 1 This is a schematic diagram of the modules of a campus network security emergency management system based on public opinion monitoring.

[0017] Figure 2 This is a schematic diagram illustrating the steps of a campus network security emergency management method based on public opinion monitoring.

[0018] Figure 3 This is a flowchart for determining public opinion.

[0019] Figure 4 This is a flowchart for emergency response. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0021] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0022] Example 1: like Figure 1 As shown, the campus network security emergency management system based on public opinion monitoring includes: a data acquisition module, a data preprocessing module, a data processing module, a data analysis module, and an emergency response module. The data acquisition module is used to collect campus network information data from the corresponding campus network information sharing platform through the API interface.

[0023] In this embodiment, the data acquisition module collects raw information data from the campus network information sharing platform via API interfaces (such as RESTful APIs), including but not limited to campus forums, Weibo campus topics, or WeChat official account comment sections. The acquisition frequency is set to once every 5 minutes, with each acquisition window containing incremental data from the most recent 5 minutes.

[0024] The data filtering module is used to filter campus network information data to obtain public opinion data.

[0025] like Figure 3 As shown in this embodiment, the specific process of filtering campus network information data to obtain public opinion data includes: It should be noted that since the collected campus network information data is massive, treating all the data as public opinion data would result in a huge amount of computation in subsequent data processing. Therefore, the data processing volume is reduced by filtering first. Furthermore, how to accurately filter out the corresponding public opinion data is also a consideration of this invention. Thus, a two-stage filtering approach is adopted to solve this technical problem.

[0026] The data filtering includes primary data filtering and secondary data filtering; wherein, primary data filtering is to perform preliminary filtering on the campus network information data, including sensitive keyword filtering, sensitive keyword homophone filtering, and sensitive keyword letter abbreviation filtering, thereby obtaining preliminary filtered data; Secondary data filtering involves semantic feature filtering of the initially filtered data to obtain public opinion data.

[0027] Furthermore, since many sensitive words in online slang use euphemisms, including but not limited to puns and abbreviations, similar issues exist in campus network information data. Therefore, in this embodiment, a preliminary screening of campus network information data can be achieved through a single data filtering step. After the first data filtering, some non-public opinion data is often retained, such as "Will you go, right?", where "go" sounds like "pass away." This data is retained after the first data filtering but is not considered public opinion data. Based on this, further filtering is needed. In this embodiment, a second data filtering step is performed to screen the data based on semantic features to determine whether it is public opinion data.

[0028] In this embodiment, the specific processes of the primary and secondary data filtering include: A data filtering process includes setting a set of sensitive keywords for filtering. Sensitive keyword homophone filtering set and a set of sensitive keyword abbreviation filters ; The campus network information data set was collected through traversal. Calculate each piece of campus network information data In the sensitive keyword filtering set Sensitive keyword homophone filtering set and a set of sensitive keyword abbreviation filters Text matching degree ; Text matching degree ;in, For string matching functions, if Returns 1 if the hit occurs, otherwise returns 0. Preset standard matching threshold If the text matching degree Greater than or equal to the standard matching threshold The initial screening data constitutes the initial screening dataset. ; Extract the preliminary filtered dataset The corresponding preliminary screening data semantic feature vector According to the semantic feature vector This forms the corresponding semantic feature vector sequence. And calculate the corresponding public opinion confidence score. ; Among them, the confidence score of public opinion ;in This represents a sequence of semantic feature vectors representing public opinion, extracted using a pre-trained BERT model.

[0029] Set public opinion judgment threshold If the public opinion confidence score corresponds to the initial screening data Greater than the public opinion judgment threshold If the data is positive, it is recorded as public opinion data; otherwise, it is recorded as normal data.

[0030] The data preprocessing module preprocesses public opinion data and outputs denoised text content, release time sequence fragments, and public opinion scene identifiers.

[0031] It should be noted that by performing multi-level data filtering on campus network information data, the preprocessing pressure on the data preprocessing module is reduced. Only the retained public opinion data needs to be preprocessed, while the rest of the data is discarded, which greatly improves data processing efficiency.

[0032] In this embodiment, the specific process of preprocessing public opinion data and outputting denoised text content, release time sequence fragments, and public opinion scene identifiers includes: Perform text cleaning operations on the input public opinion data in chronological order, remove HTML tags, special characters (such as @, #, emojis, etc.) and stop words (such as "de", "le"), and generate standardized denoised text content; Extract the timestamp information of the denoised text content , construct a time series vector , and according to the preset time granularity , perform discretization processing on the time series vector to generate a release time sequence segment ; where represents the release quantity of the nth time granularity; represents the nth moment.

[0033] Extract the keyword feature vector of the denoised text content, input it into the campus public opinion scene classifier, and calculate the probability distribution of each preset campus public opinion scene , select the scene corresponding to the maximum probability distribution as the public opinion scene identifier.

[0034] It should be noted that the campus public opinion scene classifier adopts a pre-trained TextCNN natural language processing model; the campus public opinion scenes include but are not limited to academic integrity, campus logistics services, campus safety, teacher ethics and conduct, and student physical and mental health, etc.

[0035] A data processing module, configured to input the denoised text content into a lightweight semantic encoding model, and obtain a text basic risk value according to a campus risk dictionary; calculate a short-term propagation growth rate based on the release time sequence segment, and further obtain a propagation feature compensation value; obtain the single public opinion risk reference value of the corresponding public opinion data according to the text basic risk value and the propagation feature compensation value.

[0036] In this embodiment, the specific process of inputting the denoised text content into the lightweight semantic encoding model and obtaining the text basic risk value according to the campus risk dictionary includes: Input the denoised text content into a lightweight semantic encoding model (such as DistilBERT), and output a text semantic embedding vector ; perform word segmentation processing on the denoised text content to obtain a word sequence ; according to the campus risk dictionary , and based on the indicator function, calculate the word-level risk weight of this word sequence ; where, is the indicator function, when the word hits the campus risk dictionary it takes 1, otherwise it takes 0.

[0037] Wherein, the campus risk dictionary includes risk words The risk word embedding vector of this risk word and the corresponding preset risk weights ; Based on the text semantic embedding vector and risk word embedding vectors Based on the maximum value of the cosine similarity function, the semantic similarity gain coefficient is calculated. ;in, This is the cosine similarity function.

[0038] Based on the word-level risk weights, text semantic embedding vectors, and semantic similarity gain coefficients, and using the Sigmoid function, the basic text risk value is calculated. ;in , This is the training adjustment coefficient; For the bias term, the output range is... .

[0039] In this embodiment, the specific process of obtaining the propagation feature compensation value includes: Based on the aforementioned release timeline segment Count the number of public opinion posts within a unit of time granularity to generate a propagation count sequence. ; Calculate the first difference between the release volume at the current time granularity and the previous time granularity, denoted as the short-term propagation speed. The short-term propagation speedup is normalized, and the corresponding propagation characteristic compensation value is calculated based on the linear rectification function and the attenuation factor. ;in It is a linear rectification function used to filter out noise interference caused by negative growth; It is the attenuation factor; This represents the normalized short-term propagation rate.

[0040] Based on the propagation characteristic compensation value With the aforementioned textual base risk value Obtain a single public opinion risk benchmark value. ;in The propagation compensation coefficient is 0.3.

[0041] The data analysis module is used to aggregate single public opinion risk benchmark values ​​under the same public opinion scenario identifier in a fixed time window, and calculate the peak risk within the corresponding fixed time window and the cross-scenario diffusion degree, thereby generating the public opinion heat value for the time period.

[0042] In this embodiment, the specific process of calculating the peak risk within a fixed time window and the cross-scenario diffusion degree, and then generating the public opinion heat value for a given time period, includes: Set a fixed time window length ; in the fixed time window It aggregates all individual public opinion risk benchmark values ​​belonging to the same public opinion scenario identifier and calculates the peak risk within the window. and normalize it; Statistics within a fixed time window The number of types of public opinion scene identifiers associated with the public opinion data mentioned above Then calculate the cross-scene diffusion degree. ,in For fixed time windows The total number of categories of public opinion scene identifiers associated with the corresponding public opinion data is normalized; the cross-scene diffusion degree is then analyzed. and peak risk within the window We perform a weighted summation to obtain the public opinion heat value for the time period. ,in , These are the corresponding weighting coefficients, generally , .

[0043] The emergency response module, based on the public opinion heat value of the time period and the prediction of the probability of public opinion outbreak in future time periods based on the long short-term memory network, executes the corresponding emergency response instructions in the time period before the probability of public opinion outbreak exceeds the preset threshold.

[0044] like Figure 4 As shown, in this embodiment, the specific process of predicting the probability of a future public opinion outbreak based on the public opinion heat value of the specified time period and using a long short-term memory network includes: Construct a series of public opinion heat values ​​for historical periods The input is then fed into a pre-trained Long Short-Term Memory (LSTM) network; through the hidden layer state propagation of the LSM network, the hidden feature vectors for future time periods are output. ; The hidden feature vector Input a fully connected regression layer and use the Softmax function to calculate the probability of a future public opinion outbreak. ;in The weight matrix of the fully connected layer represents the hidden features learned by the Long Short-Term Memory network. Unnormalized scores mapped to each category; As a bias term, it provides a shift, allowing the model to have non-zero outputs and improve fitting ability even without relying on input features.

[0045] Preset threshold for public opinion outbreak ; When the probability of a public opinion outbreak Greater than the threshold for public opinion outbreak In such cases, emergency instructions are automatically issued before the next time period begins. For example, if a public opinion crisis related to student safety is predicted to erupt, the system will automatically trigger actions including sending a warning email to the academic affairs office and temporarily freezing anonymous posting privileges on related topics.

[0046] Example 2: like Figure 2 As shown, a campus network security emergency management method based on public opinion monitoring includes: Through API interfaces, collect campus network information data from the corresponding campus network information sharing platform; filter the campus network information data to obtain public opinion data; Preprocess the public opinion data to output denoised text content, release time sequence fragments, and public opinion scene identifiers; The denoised text content is input into a lightweight semantic coding model, and a basic risk value is obtained based on a campus risk dictionary. Based on the release time sequence, the short-term propagation growth rate is calculated, and then the propagation feature compensation value is obtained. Based on the basic risk value of the text and the propagation feature compensation value, the single public opinion risk benchmark value of the corresponding public opinion data is obtained. By using a fixed time window to aggregate the benchmark value of a single public opinion risk under the same public opinion scenario identifier, and calculating the peak risk within the corresponding fixed time window and the cross-scenario diffusion, a public opinion heat value for a given period is generated. Based on the public opinion heat value for the specified time period, and using a long short-term memory network to predict the probability of a public opinion outbreak in future time periods, corresponding emergency handling instructions are executed in the time period before the probability of a public opinion outbreak exceeds a preset threshold. Optionally, in this embodiment, those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. This program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a disk, or an optical disk, etc.

[0047] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0048] If the integrated units in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in the aforementioned computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause one or more electronic devices to execute all or part of the steps of the methods described in the various embodiments of this application.

[0049] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0050] In the several embodiments provided in this application, it should be understood that the disclosed application can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of units or modules may be electrical or other forms.

[0051] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0052] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0053] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A campus network security emergency management system based on public opinion monitoring, characterized in that: The system includes: a data acquisition module, a data filtering module, a data preprocessing module, a data processing module, a data analysis module, and an emergency response module; The data acquisition module is used to collect campus network information data from the corresponding campus network information sharing platform through the API interface; The data filtering module is used to filter campus network information data to obtain public opinion data; The data preprocessing module preprocesses public opinion data and outputs denoised text content, release time sequence fragments, and public opinion scene identifiers. The data processing module is used to input the denoised text content into the lightweight semantic coding model and obtain the basic risk value of the text based on the campus risk dictionary; calculate the short-term propagation growth rate based on the release time sequence segment, and then obtain the propagation feature compensation value; and obtain the single public opinion risk benchmark value of the corresponding public opinion data according to the basic risk value of the text and the propagation feature compensation value. The data analysis module is used to aggregate single public opinion risk benchmark values ​​under the same public opinion scenario identifier in a fixed time window, and calculate the peak risk within the corresponding fixed time window and the cross-scenario diffusion degree, thereby generating the public opinion heat value for the time period. The emergency response module, based on the public opinion heat value of the time period and the prediction of the probability of public opinion outbreak in future time periods based on the long short-term memory network, executes the corresponding emergency response instructions in the time period before the probability of public opinion outbreak exceeds the preset threshold. The process of inputting denoised text content into a lightweight semantic coding model and obtaining the basic risk value of the text based on a campus risk dictionary includes: The denoised text content is input into a lightweight semantic coding model, which outputs a text semantic embedding vector. The denoised text content is then segmented to obtain a word sequence. Based on a campus risk dictionary and an indicator function, the word-level risk weight of the word sequence is calculated. The campus risk dictionary includes risk words, risk word embedding vectors of the risk words, and corresponding preset risk weights. Based on the text semantic embedding vector and the risk word embedding vector, and based on the maximum value of the cosine similarity function, the semantic similarity gain coefficient is calculated; Based on the word-level risk weights, text semantic embedding vectors, and semantic similarity gain coefficients, and using the Sigmoid function, the basic risk value of the text is calculated. The process of obtaining propagation feature compensation values ​​includes: Based on the aforementioned release time sequence segment, the number of public opinion releases within a unit time granularity is counted to generate a propagation count sequence; Calculate the first-order difference between the release volume at the current time granularity and the previous time granularity, denoted as the short-term propagation growth rate; normalize the short-term propagation growth rate, and calculate the corresponding propagation characteristic compensation value based on the linear rectification function and the attenuation factor; Based on the aforementioned dissemination characteristic compensation value and the basic text risk value, a single public opinion risk benchmark value is obtained; The process of calculating the peak risk within a fixed time window and the cross-scenario diffusion degree, and then generating the public opinion heat value for a given period, includes: Set a fixed time window length; aggregate all single public opinion risk benchmark values ​​belonging to the same public opinion scenario identifier within the fixed time window, calculate the peak risk within the window, and normalize it; The number of public opinion scene identifiers associated with the public opinion data within a fixed time window is counted, and then the cross-scene diffusion degree is calculated; the cross-scene diffusion degree and the peak risk within the window are weighted and summed to obtain the public opinion heat value for the time period.

2. The campus network security emergency management system based on public opinion monitoring according to claim 1, characterized in that, The process of filtering campus network information data to obtain public opinion data includes: The data filtering includes primary data filtering and secondary data filtering; wherein, primary data filtering is to perform preliminary filtering on the campus network information data, including sensitive keyword filtering, sensitive keyword homophone filtering, and sensitive keyword letter abbreviation filtering, thereby obtaining preliminary filtered data; Secondary data filtering involves semantic feature filtering of the initially filtered data to obtain public opinion data.

3. The campus network security emergency management system based on public opinion monitoring according to claim 2, characterized in that, The process of primary and secondary data filtering includes: A data filtering process includes setting up a sensitive keyword filter set, a sensitive keyword homophone filter set, and a sensitive keyword letter abbreviation filter set; The collected campus network information data set is traversed, and the text matching degree of each piece of campus network information data in the sensitive keyword filtering set, the sensitive keyword homophone filtering set, and the sensitive keyword letter abbreviation filtering set is calculated; if the text matching degree is greater than or equal to the standard matching degree threshold, the preliminary filtered data form the preliminary filtered dataset. Extract the semantic feature vectors corresponding to the preliminary screening data from the preliminary screening dataset, and calculate the corresponding public opinion confidence score based on the semantic feature vectors; set a public opinion judgment threshold, and if the public opinion confidence score of the corresponding preliminary screening data is greater than the public opinion judgment threshold, it is recorded as public opinion data, otherwise it is recorded as normal data.

4. The campus network security emergency management system based on public opinion monitoring according to claim 3, characterized in that, The process of preprocessing public opinion data to output denoised text content, release time sequence fragments, and public opinion scene identifiers includes: Based on the chronological order of the input public opinion data, a text cleaning operation is performed to remove HTML tags, special characters, and stop words, generating standardized denoised text content. Extract timestamp information from the denoised text content, construct a time series vector, and discretize the time series vector according to a preset time granularity to generate a release time sequence segment; Extract the keyword feature vectors from the denoised text content and input them into the campus public opinion scene classifier. Calculate the probability distribution of each preset campus public opinion scene and select the scene corresponding to the highest probability distribution as the public opinion scene identifier.

5. The campus network security emergency management system based on public opinion monitoring according to claim 4, characterized in that, The process of predicting the probability of a future public opinion outburst based on the public opinion heat value of the specified time period and using a long short-term memory network includes: Construct a sequence of public opinion heat values ​​for historical periods and input it into a trained long short-term memory network; output the latent feature vector for future periods through the hidden layer state propagation of the long short-term memory network. The hidden feature vector is input into a fully connected regression layer, and the probability of a public opinion outbreak in the future period is calculated using the Softmax function. Execute corresponding emergency response instructions in the period before the probability of a public opinion outbreak exceeds a preset threshold.

6. A campus network security emergency management method based on public opinion monitoring, implementing the campus network security emergency management system based on public opinion monitoring as described in any one of claims 1 to 5, characterized in that, The method includes: Through API interfaces, collect campus network information data from the corresponding campus network information sharing platform; filter the campus network information data to obtain public opinion data; Preprocess the public opinion data to output denoised text content, release time sequence fragments, and public opinion scene identifiers; The denoised text content is input into a lightweight semantic coding model, and a basic risk value is obtained based on a campus risk dictionary. Based on the release time sequence, the short-term propagation growth rate is calculated, and then the propagation feature compensation value is obtained. Based on the basic risk value of the text and the propagation feature compensation value, the single public opinion risk benchmark value of the corresponding public opinion data is obtained. By using a fixed time window to aggregate the benchmark value of a single public opinion risk under the same public opinion scenario identifier, and calculating the peak risk within the corresponding fixed time window and the cross-scenario diffusion, a public opinion heat value for a given period is generated. Based on the public opinion heat value of the specified time period, and based on the prediction of the probability of public opinion outbreak in future time periods using a long short-term memory network, the corresponding emergency handling instructions are executed in the time period before the probability of public opinion outbreak exceeds a preset threshold.

Images