Data security management method based on user classification

Abstract

The present invention provides a data security management method based on user grading and classification, which belongs to the field of data security technology. The method includes obtaining the main business type and the secondary business type of the current social media platform, and collecting basic data of the current social media users based on the main business type and the secondary business type; performing cluster analysis on the basic data, obtaining first data according to the cluster analysis result, performing data security identification on the first data, and obtaining second data; inputting the second data into a machine learning model, setting grading rules for the machine learning model, grading users, and formulating a data protection strategy according to the grading results; setting data update time to monitor, evaluate and optimize the basic data, so as to realize effective management and utilization of social media user data, help improve user satisfaction, protect user data security, and effectively improve the service quality and operation efficiency of the platform.

Description

Technical Field

[0001] The present invention relates to the field of data security technology, and in particular to a data security management method based on user classification. Background Art

[0002] With the popularity of social media platforms and the explosive growth in the number of users, a large amount of user data has emerged on the platforms. These data have extremely high commercial value, but also have problems such as privacy leakage and data security. How to effectively use these data and provide customized services while protecting user privacy has become an important issue facing social media platforms. At the same time, the behaviors and needs of users on social media platforms are diverse, and a single service cannot satisfy all users.

[0003] Therefore, the present invention provides a data security management method based on user classification. Summary of the invention

[0004] The present invention provides a data security management method based on user grading and classification, which is used to collect basic data of users by obtaining the main business type and secondary business type of the current social media platform, perform cluster analysis on technical data and perform data security identification on the cluster analysis results to obtain second data, classify users according to the second data input by a machine learning model, and then set the data update time, so as to realize effective management and utilization of social media user data, help to improve user satisfaction, protect user data security, and effectively improve the service quality and operation efficiency of the platform.

[0005] The present invention provides a data security management method based on user classification, including:

[0006] Step 1: Obtain the main business type and secondary business type of the current social media platform, and collect basic data on current social media users based on the main business type and secondary business type;

[0007] Step 2: performing cluster analysis on the basic data, obtaining first data according to the cluster analysis result, performing data security identification on the first data, and obtaining second data;

[0008] Step 3: Input the second data into the machine learning model, set classification rules for the machine learning model, classify users, and formulate a data protection strategy based on the classification results;

[0009] Step 4: Set data update time to monitor, evaluate and optimize basic data.

[0010] The present invention provides a data security management method based on user classification and classification, which obtains the main business type and the secondary business type of the current social media platform, and collects basic data of the current social media users based on the main business type and the secondary business type, including:

[0011] Determine the main and secondary business types of each social media platform based on the main and secondary functions of all current social media platforms;

[0012] Extracting first relevant descriptions of all main business types to construct a first feature vector set and extracting second relevant descriptions of all secondary business types to construct a second feature vector set;

[0013] Performing a first similarity calculation on the first feature vector set, and at the same time, performing a second similarity calculation on the second feature vector set;

[0014] Extracting a first type exceeding a first set value according to a first similarity calculation result;

[0015] extracting a second type exceeding a second set value according to a second similarity calculation result;

[0016] Performing intersection processing on the first type and the second type to determine an intersection part and a non-intersection part;

[0017] Merge and collect basic data of the types involved in the intersection;

[0018] The basic data of the types involved in the non-intersection part are collected separately.

[0019] The present invention provides a data security management method based on user classification, performing cluster analysis on the basic data, obtaining first data according to the cluster analysis result, performing data security identification on the first data, and obtaining second data, including:

[0020] Divide user behaviors into primary activities and secondary activities based on the collected data, and classify users based on the activity classification results;

[0021] Perform cluster analysis on the basic data collected within a period of time based on the data type, use the cluster analysis result as the first data, and use the maximum and second largest values ​​in the cluster analysis results of each data type within the period of time as the daily data of the corresponding data type;

[0022] Extracting keywords from the daily data, and expanding the daily data range in the first data according to the keywords;

[0023] Performing data security identification on the first data according to the data security rules and the expanded daily data;

[0024] The second data is obtained by combining the security identification result with the user classification result corresponding to the first data.

[0025] The present invention provides a data security management method based on user classification and classification, inputs the second data into a machine learning model, sets classification rules for the machine learning model, classifies users, and formulates a data protection strategy based on the classification results, including:

[0026] A security screening structure throughout the entire life cycle of the data is constructed based on the machine learning model and the second data, and a classification rule is set for the machine learning model to screen the structured data and unstructured data in the second data. According to the screening results and the preset classification rules, the primary security level corresponding to each sub-data in the second data is determined;

[0027] Identify the sensitive data in the second data using the sensitive data automatic identification mechanism, and determine the sensitivity category and corresponding level of each sub-data in the second data according to the data classification and grading method;

[0028] Adjusting the primary security level of the corresponding sub-data according to the sensitive categories and corresponding levels to obtain a final security level;

[0029] Set weights for users’ publishing and participation behaviors, and determine the user’s security level based on the weight setting results and the final security level;

[0030] The users are classified based on their security levels, and protection strategies are set according to the classification results. The protection strategies include: basic security, architecture security, data security and application security.

[0031] The present invention provides a data security management method based on user classification and classification, and the primary security level of the corresponding sub-data is adjusted according to the sensitive category and the corresponding level, and the process of obtaining the final security level also includes:

[0032] Perform user privilege management and control based on the screening results. Privilege management and control includes privileged user scenario management, privileged user account discovery, privileged user account integrated management, privileged user account access control, privileged user account automatic password change, and privileged session monitoring.

[0033] Through the usage scenarios and corresponding usage rules of privileged users, a corresponding approval and authorization mechanism is established to record in detail the access steps, operation logs and privileged operations of privileged users' accounts, and a post-inspection mechanism is provided for the approval and authorization mechanism to conduct scenario control of privileged users on the screening results;

[0034] Establish a privilege protection mechanism, reasonably allocate authorized access to user data, establish a privilege management platform, and determine the unique access entrance for all privileged users based on the privilege management platform;

[0035] Based on threat control commands and resource access control, abnormal behavior identification and interception are performed on privileged users with the only access entrance;

[0036] Update the monitoring capabilities of privileged data access and data violations based on abnormal behaviors, and make secondary adjustments to the primary security level.

[0037] The present invention provides a data security management method based on user classification and classification, which sets weights for user publishing and participation behaviors, and determines the user's security level based on the weight setting results combined with the final security level, including:

[0038] Through machine learning, we analyze the emotional contagion of users’ posting and participation behaviors, calculate the proportion of emotions in the sub-data corresponding to users’ posting and participation behaviors every day, and set the threshold for the proportion of negative emotions;

[0039] Predicting the user's subsequent emotional trend according to the negative emotion ratio threshold, and setting weights for access rights of different data types for the corresponding user's subsequent emotional trend;

[0040] The weight setting result is added to the final security level of the corresponding sub-data to obtain the user data security level, and the user's security level is weighted and combined with the data level of daily data in the corresponding period of time.

[0041] The present invention provides a data security management method based on user classification and classification, which sets the data update time to monitor, evaluate and optimize the basic data, including:

[0042] Establish a data security view of the second data to monitor user behavior in the basic data at the data update time, and combine the user's multi-dimensional related data to intelligently predict possible abnormal behavior of the user;

[0043] Adjust the pre-set update time according to the prediction results, collect and monitor user behavior data according to the adjusted time, evaluate the accuracy of the current user classification based on the user behavior data on the user side, and optimize the user classification standards based on the evaluation results.

[0044] The present invention provides a data security management method based on user classification and classification, which performs emotional infection analysis on users through machine learning methods according to their publishing and participation behaviors, calculates the proportion of emotions in the sub-data corresponding to the users' publishing and participation behaviors every day, and sets a threshold for the proportion of negative emotions, including:

[0045] The content influence range is obtained based on the content popularity corresponding to the user's publishing and participation behaviors, and the emotional influence analysis of the content influence range is performed based on machine learning;

[0046] Determine the proportion of users' emotions in each time period of the day based on the emotion lag, determine the time when negative emotions are generated, and make negative adjustments to the time and degree of current negative emotions based on the time and degree of the user's corresponding negative emotions in history;

[0047] According to the adjusted proportion of negative emotions and the emotional connection between negative emotions, a proportion threshold of each negative emotion and a quantity threshold of negative emotions are set.

[0048] Compared with the prior art, the beneficial effects of the present application are as follows: basic data of users is collected by acquiring the main business type and secondary business type of the current social media platform, cluster analysis is performed on the technical data and data security identification is performed on the cluster analysis results to obtain second data, users are classified according to the second data input by the machine learning model, and then the data update time is set to achieve effective management and utilization of social media user data, which helps to improve user satisfaction, protect user data security, and effectively improve the service quality and operational efficiency of the platform.

[0049] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0050] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0052] Figure 1 The present invention is a flowchart of a data security management method based on user classification and classification in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0054] Embodiment 1:

[0055] The embodiment of the present invention provides a data security management method based on user classification, such as Figure 1 As shown, including:

[0056] Step 1: Obtain the main business type and secondary business type of the current social media platform, and collect basic data on current social media users based on the main business type and secondary business type;

[0057] Step 2: performing cluster analysis on the basic data, obtaining first data according to the cluster analysis result, performing data security identification on the first data, and obtaining second data;

[0058] Step 3: Input the second data into the machine learning model, set classification rules for the machine learning model, classify users, and formulate a data protection strategy based on the classification results;

[0059] Step 4: Set data update time to monitor, evaluate and optimize basic data.

[0060] In this embodiment, basic data collection is performed by constructing a feature vector based on the main business type and the secondary business type of the social media platform to divide the range, and then the range is subjected to intersection and union processing. Based on the processing results, all aspects of the basic data to be collected are determined to effectively classify users on the social media platform and collect their behavioral data.

[0061] In this embodiment, the second data is obtained by performing data security identification on the first data based on the data security rules in combination with the expanded daily data, and the second data is obtained by combining the security identification result with the user classification result corresponding to the first data.

[0062] In this embodiment, the grading rules are determined based on data sensitivity, data usage, data access rights and legal and regulatory requirements, such as Level 1: non-sensitive data, which can be publicly accessed; Level 2: low-sensitivity data, which can only be accessed by authorized users or systems; Level 3: medium-sensitivity data, which contains certain sensitive information, such as personal contact information, and requires stricter access control; Level 4: highly sensitive data, such as ID numbers, medical records, etc., which can only be accessed by users or systems with specific permissions.

[0063] In this embodiment, monitoring and evaluation data is to monitor user behavior in basic data in real time through data security view, and predict abnormal behavior that may occur to users in combination with user multi-dimensional related data, and adjust the time of data collection and monitoring according to the abnormal behavior prediction results.

[0064] In this embodiment, the data protection strategy includes setting different data access and usage permissions for users at each level according to the user level; encrypting the user's sensitive data, and the recipient needs a decryption key to view the data content; encrypting the user's sensitive data, and the recipient needs a decryption key to view the data content; clearly defining the management methods for each stage of data from generation to destruction to protect the entire life cycle of data; formulating emergency response plans and repair methods for possible data security issues; and setting corresponding privileged user handling methods.

[0065] The working principle and beneficial effects of the above technical solution are: basic data of users is collected by obtaining the main business type and secondary business type of the current social media platform, cluster analysis is performed on the technical data and data security identification is performed on the cluster analysis results to obtain second data, users are classified according to the second data input by the machine learning model, and then the data update time is set to achieve effective management and utilization of social media user data, which helps to improve user satisfaction, protect user data security, and effectively improve the service quality and operational efficiency of the platform.

[0066] Embodiment 2:

[0067] The embodiment of the present invention provides a data security management method based on user classification and classification, which obtains the main business type and the secondary business type of the current social media platform, and collects basic data of the current social media users based on the main business type and the secondary business type, including:

[0068] Determine the main and secondary business types of each social media platform based on the main and secondary functions of all current social media platforms;

[0069] Extracting first relevant descriptions of all main business types to construct a first feature vector set and extracting second relevant descriptions of all secondary business types to construct a second feature vector set;

[0070] Performing a first similarity calculation on the first feature vector set, and at the same time, performing a second similarity calculation on the second feature vector set;

[0071] Extracting a first type exceeding a first set value according to a first similarity calculation result;

[0072] extracting a second type exceeding a second set value according to a second similarity calculation result;

[0073] Performing intersection processing on the first type and the second type to determine an intersection part and a non-intersection part;

[0074] Merge and collect basic data of the types involved in the intersection;

[0075] The basic data of the types involved in the non-intersection part are collected separately.

[0076] In this embodiment, the first feature vector set includes media interaction type, content publishing form, user interaction mode, content positioning and management, and user group positioning.

[0077] In this embodiment, the second feature vector set includes social media data analysis tools, advertising push content, e-commerce, plug-in applications, paid membership services, developer tools and community modules.

[0078] In this embodiment, the first similarity calculation is to calculate the cosine similarity between the first feature vectors corresponding to each social media platform. ,in, is the first similarity, and represents the first eigenvector and jth eigenvector of n-dimension, T represents the transposition operation, It means that two vectors are exactly the same. It means that the two vectors are completely unrelated. Indicates that the two vectors are completely opposite.

[0079] In this embodiment, the first relevant description is text or sentences that describe in detail the main functions or services of the social media platform. For example, for Xiaohongshu, whose main function is picture sharing, the first relevant description of its main business type is that users can share and discover new life photos, and can also edit and process photos through the filter function. For WeChat, whose main function is message communication, the first relevant description of its main business type is that users can send and receive text messages, voice, pictures and videos in real time.

[0080] In this embodiment, the second related description is a more specific or detailed description of the ancillary functions or services of the social media platform. For example, for Douyin, the main service type is video sharing, and the corresponding second related description is that users can upload and share videos, conduct real-time online live broadcasts, subscribe to updates from other users or channels, etc. When video comments and likes are regarded as secondary service types, the corresponding second related description is that users can comment on the video, like or reply to other users' comments, and participate in community interaction.

[0081] In this embodiment, the first type is a type obtained by merging main business types with high cosine similarity after calculating cosine similarity of the main business types.

[0082] In this embodiment, the second type is a type obtained by merging sub-camp types with high cosine similarity after calculating cosine similarity of the sub-camp types.

[0083] The working principle and beneficial effects of the above technical solution are: constructing feature vectors through the main business type and secondary business type of the social media platform to divide the range, performing intersection and union processing on the range, and determining all aspects of collecting basic data based on the processing results, effectively classifying users on the social media platform, and collecting and analyzing their behavioral data, thereby providing social media platforms with a deeper understanding of user behavior, and facilitating the formulation of targeted operation strategies and improving user satisfaction and engagement.

[0084] Embodiment 3:

[0085] The embodiment of the present invention provides a data security management method based on user classification, performing cluster analysis on the basic data, obtaining first data according to the cluster analysis result, performing data security identification on the first data, and obtaining second data, including:

[0086] Divide user behaviors into primary activities and secondary activities based on the collected data, and classify users based on the activity classification results;

[0087] Perform cluster analysis on the basic data collected within a period of time based on the data type, use the cluster analysis result as the first data, and use the maximum and second largest values ​​in the cluster analysis results of each data type within the period of time as the daily data of the corresponding data type;

[0088] Extracting keywords from the daily data, and expanding the daily data range in the first data according to the keywords;

[0089] Performing data security identification on the first data according to the data security rules and the expanded daily data;

[0090] The second data is obtained by combining the security identification result with the user classification result corresponding to the first data.

[0091] In this embodiment, users are classified in a vector machine model according to their main activities and secondary activities, that is, activities are used as features to construct activity feature vectors, and the feature vectors are trained using a vector machine model to ultimately obtain the classification of the corresponding users.

[0092] In this embodiment, the activity division results include that primary activities are the dominant behaviors of users on social media platforms, such as posting status updates, sharing videos, etc. These behaviors usually occupy most of the user's online activities and reflect the user's main usage habits and purposes of the social media platform; secondary activities are auxiliary behaviors of users on social media platforms, such as commenting on others' posts, giving "like" or "dislike" feedback, sharing others' content, etc.

[0093] In this embodiment, the user classification results include content consumption users, content creation users, socially active users, business users and potential users.

[0094] In this embodiment, text analysis is performed on daily data to determine keywords, words with a cosine similarity of 0.95 or more with the keywords are matched from a vocabulary to expand the keywords, and the expanded keywords are matched with the first data to expand the daily data.

[0095] In this embodiment, the maximum value in the cluster analysis result is the most frequent behavior of the user within a period of time. For example, if a cluster analysis is performed on the user's behavior on a social media platform, the maximum value corresponds to browsing news, and the second largest value corresponds to posting comments. This means that during this period of time, the user spent the most time browsing news, followed by posting comments, which is used as daily data.

[0096] In this embodiment, data security identification includes abnormal behavior detection, data leakage detection and malware detection.

[0097] The working principle and beneficial effects of the above technical solution are: by dividing the collected data into activities, classifying users according to the division results, performing cluster analysis on the basic data, determining daily data, and securely identifying the cluster analysis results to obtain the second data, deep identification and classification of user behaviors, cluster analysis and anomaly identification of data, and protection of data security are achieved, providing social media platforms with efficient, secure, and accurate data processing and service methods.

[0098] Embodiment 4:

[0099] The embodiment of the present invention provides a data security management method based on user classification and classification, inputs the second data into a machine learning model, sets classification rules for the machine learning model, classifies users, and formulates a data protection strategy based on the classification results, including:

[0100] A security screening structure throughout the entire life cycle of the data is constructed based on the machine learning model and the second data, and a classification rule is set for the machine learning model to screen the structured data and unstructured data in the second data. According to the screening results and the preset classification rules, the primary security level corresponding to each sub-data in the second data is determined;

[0101] Identify the sensitive data in the second data using the sensitive data automatic identification mechanism, and determine the sensitivity category and corresponding level of each sub-data in the second data according to the data classification and grading method;

[0102] Adjusting the primary security level of the corresponding sub-data according to the sensitive categories and corresponding levels to obtain a final security level;

[0103] Set weights for users’ publishing and participation behaviors, and determine the user’s security level based on the weight setting results and the final security level;

[0104] The users are classified based on their security levels, and protection strategies are set according to the classification results. The protection strategies include: basic security, architecture security, data security and application security.

[0105] In this embodiment, the primary security level is based on classification rules formulated according to data access frequency, data type, and data content. The data filtered by the rules is reviewed with the preset classification rules, and a primary security level is assigned to each data. For example, data that is frequently viewed by users and contains sensitive information may be marked as a higher security level.

[0106] In this embodiment, sensitive data generally includes personal identity information, contact information, personal financial information, personal health information, legal information, education and work information, network information and other sensitive information involving personal privacy or prone to abuse or fraud. The corresponding levels are divided according to the risks that each type of information may bring. For example, information involving direct personal financial interests is set as high-level sensitive information; information involving basic personal identity information is set as medium-level sensitive information; information involving part of contact information, etc. is set as low-level sensitive information.

[0107] In this embodiment, the security screening structure is a set of mechanisms or tools that can identify and filter out potential high-risk or non-compliant data, including data reading, data preprocessing, machine learning model processing, security level assessment and data filtering.

[0108] In this embodiment, the final security level adjustment process is to use the matched sensitive categories and corresponding levels to adjust the security level of each sub-data. For data that is ultimately assessed to be at a high-risk level, the data is processed according to business needs and relevant regulations, such as data isolation, encrypted storage, deletion, etc. For example, if a sub-data is identified as containing personally identifiable information, its security level will be adjusted to the highest level.

[0109] In this embodiment, the weight setting process is to measure and formulate standards based on the user's publishing behavior and participation behavior, and give different weights according to the importance of each standard. For example, if it is believed that publishing behavior can better reflect a user's behavioral habits than participation behavior, publishing behavior can be given a higher weight. Then, for each user, based on the performance of his or her publishing behavior and participation behavior, combined with the set weights, a behavior score is calculated. For example, if the data involving the user is often judged to be high risk, the security score of this user needs to be lowered. Finally, the behavior score is adjusted according to the user's final security level, and then a threshold is set to determine the user's security level to determine the final security level.

[0110] In this embodiment, basic security refers to basic security measures that should be implemented for all users. For example, all users are required to perform password protection and two-step verification with sufficient strength, conduct security education and training regularly, and guide users to avoid disclosing sensitive information. Architecture security refers to architecture-level security, which is usually aimed at the system itself. For example, using a layered security architecture, users with different permissions can only access data and resources at the corresponding level, using encryption and firewall technology to protect data transmission and storage, setting up a log and monitoring system, tracking and recording the behavior of all users, and timely discovering and preventing potential security threats. Data security is a protection measure for the data itself. For example, data generated by users with high risk levels may require a higher level of protection; adding backup and recovery mechanisms to prevent data loss, and for data containing sensitive information, anonymization and de-identification processing needs to be considered. Application security refers to other application-level security measures. For example, for users with low risk levels, they may be allowed to use certain functions, such as posting comments and sharing content; while for users with high risk levels, some or all of their functions may be restricted, and it may even be decided to remove them from the system.

[0111] The working principle and beneficial effects of the above technical solution are: by scientifically and reasonably classifying and grading data, setting user security levels, and orderly managing data access according to user permissions, data security can be effectively improved, data leakage, tampering, damage, etc. can be prevented, and unauthorized personnel can be prevented from accessing sensitive data, thereby ensuring the safe use of data.

[0112] Embodiment 5:

[0113] The embodiment of the present invention provides a data security management method for user classification and classification, wherein the primary security level of the corresponding sub-data is adjusted according to the sensitive category and the corresponding level, and the process of obtaining the final security level also includes:

[0114] Perform user privilege management and control based on the screening results. Privilege management and control includes privileged user scenario management, privileged user account discovery, privileged user account integrated management, privileged user account access control, privileged user account automatic password change, and privileged session monitoring.

[0115] Through the usage scenarios and corresponding usage rules of privileged users, a corresponding approval and authorization mechanism is established to record in detail the access steps, operation logs and privileged operations of privileged users' accounts, and a post-inspection mechanism is provided for the approval and authorization mechanism to conduct scenario control of privileged users on the screening results;

[0116] Establish a privilege protection mechanism, reasonably allocate authorized access to user data, establish a privilege management platform, and determine the unique access entrance for all privileged users based on the privilege management platform;

[0117] Based on threat control commands and resource access control, abnormal behavior identification and interception are performed on privileged users with the only access entrance;

[0118] Update the monitoring capabilities of privileged data access and data violations based on abnormal behaviors, and make secondary adjustments to the primary security level.

[0119] In this embodiment, privilege management refers to the management and monitoring of users with high-level permissions on the social platform.

[0120] In this embodiment, privileged user scenario control is a management method that defines user permissions based on the user's operations or behaviors on the social platform. For example, for Facebook management and maintenance personnel, they can access and manage user data and system settings, and the corresponding permissions are limited to their necessary work scope, such as system maintenance, user complaint handling, etc.

[0121] In this embodiment, privileged user account discovery refers to the periodic automatic discovery and identification of all privileged user accounts by a system or tool. For example, Facebook needs to regularly check and audit all privileged accounts to ensure that they are not abused or used improperly. Privileged user account integrated management is usually a process of centralized management and monitoring of privileged user accounts that access sensitive data and key system resources, including system administrators, developers, advanced users, and other users who can access sensitive data and system resources. For example, Facebook has a unified privileged account management system that can centrally manage all privileged accounts, including permission allocation, password changes, account freezing and deletion, etc. Privileged user account access control is to Access control for privileged users is based on a series of behaviors such as user identity authentication, authorization, and monitoring of account usage. For example, Facebook has strict access control over privileged accounts. For example, multiple authentications are required to log into privileged accounts. At the same time, login records of privileged accounts are strictly monitored. Automatic password change for privileged user accounts means that in the privileged user management system, the system automatically and regularly changes the passwords of privileged accounts without manual intervention, thereby ensuring the security of privileged account passwords. Privileged session monitoring means that when a privileged user uses his or her privileged account to perform operations, the system can record and monitor all activities of the privileged session, ensuring that the behavior of the privileged user is within the scope of compliance and security, and when a security incident occurs, it can be traced back to the specific operation steps.

[0122] In this embodiment, the approval and authorization mechanism is a security measure used to control the access and operations of privileged users. In the approval and authorization mechanism, each operation of the privileged user needs to be authorized in advance, otherwise it will be denied execution. For example, a privileged user may be allowed to perform certain operations within a specific time period, but may not be allowed at other times, or some important system changes may require the joint authorization of multiple privileged users.

[0123] In this embodiment, the privilege protection mechanism includes strict access control, password protection policy, auditing and monitoring, principle of least privilege, account management and disaster recovery plan.

[0124] The working principle and beneficial effects of the above technical solution are: by defining the usage scenarios and corresponding usage rules of privileged users, establishing an approval and authorization mechanism, making detailed records of privileged user activities, and reviewing the operations performed, it is ensured that only authorized users can access the data, and setting a unique access entrance for all privileged users through the privilege management platform, and identifying and intercepting their abnormal behaviors, and updating the monitoring capabilities of data access and data violations controlled by privileged users based on these abnormal behaviors, preventing security issues such as data leakage and tampering, strengthening the authority management and control of privileged users, avoiding privileged users from abusing their authority or unauthorized authority access, and ensuring the security and integrity of data.

[0125] Embodiment 6:

[0126] The embodiment of the present invention provides a data security management method based on user classification and classification, which sets weights for user publishing and participation behaviors, and determines the user's security level based on the weight setting result combined with the final security level, including:

[0127] Through machine learning, we analyze the emotional contagion of users’ posting and participation behaviors, calculate the proportion of emotions in the sub-data corresponding to users’ posting and participation behaviors every day, and set the threshold for the proportion of negative emotions;

[0128] Predicting the user's subsequent emotional trend according to the negative emotion ratio threshold, and setting weights for access rights of different data types for the corresponding user's subsequent emotional trend;

[0129] The weight setting result is added to the final security level of the corresponding sub-data to obtain the user data security level, and the user's security level is weighted and combined with the data level of daily data in the corresponding period of time.

[0130] In this embodiment, emotional contagion analysis is a method for analyzing emotional changes in user posts and participation behaviors using machine learning technology. Through text mining technology, such as sentiment analysis, the user's emotional state, such as happiness, sadness, anger, etc., is extracted from the user's posts and participation behaviors. Then, the proportion of specific emotions (such as negative emotions) in the user's posts and participation behaviors in a specific time period, such as a day, is calculated.

[0131] In this embodiment, once the proportion of negative emotions exceeds a preset threshold, the system will predict that the user's emotions may continue to be negative in the future. Based on this prediction, the system will set weights for the user's data access rights in subsequent time periods to protect the security of users and social platforms. For example, if a user's negative emotions account for a high proportion and it is predicted that he may continue to post negative and potentially controversial content in the following period, the system may reduce his permission to post comments to prevent potential negative impacts.

[0132] The working principle and beneficial effects of the above technical solution are: by conducting emotional contagion analysis on user behavior, predicting the user's emotional state, using the user emotion analysis results for setting data permissions, determining the security level of privileged users and the security level of users, to a certain extent prevent malicious operations or improper behaviors caused by user emotional problems, and achieve meticulous and in-depth user management and powerful data protection.

[0133] Embodiment 7:

[0134] The embodiment of the present invention provides a data security management method based on user classification and classification, which sets the data update time to monitor, evaluate and optimize the basic data, including:

[0135] Establish a data security view of the second data to monitor user behavior in the basic data at the data update time, and combine the user's multi-dimensional related data to intelligently predict possible abnormal behavior of the user;

[0136] Adjust the pre-set update time according to the prediction results, collect and monitor user behavior data according to the adjusted time, evaluate the accuracy of the current user classification based on the user behavior data on the user side, and optimize the user classification standards based on the evaluation results.

[0137] In this embodiment, a convolutional neural network is used to analyze the text and image information posted by users on social media.

[0138] In this embodiment, time adjustment means that if the prediction results show that a user may generate abnormal behavior in a specific time period (such as excessively frequent login attempts, posting false news or misleading information, etc.), then the platform may decide to update and monitor the user's behavior data more frequently during this time period.

[0139] In this embodiment, multi-dimensional correlation data means that in user behavior analysis, not only one dimension of data is focused on, but multiple dimensions of data of the user are combined for analysis, and the dimensions include login behavior data, user personal information, user behavior data, user device information and user social network data.

[0140] In this embodiment, the data security view is the steps of data collection, cleaning, conversion, analysis and visualization. For example, it may include processing the original data, extracting key features and patterns, and then presenting them in the form of charts, reports or other forms. The information corresponding to the data security view includes the time, frequency and location of user login behavior, user activities after registration, the trajectory of user operation behavior, the interaction between users and other users, and any other information that helps to identify abnormal behavior.

[0141] The working principle and beneficial effects of the above technical solution are: real-time monitoring of user behavior in basic data through data security view, and prediction of abnormal behavior that may occur to users in combination with user multi-dimensional related data, and adjustment of data collection and monitoring time according to abnormal behavior prediction results, to achieve refined and intelligent user behavior prediction and management, enhance the ability to prevent network security threats, and ensure data security.

[0142] Embodiment 8:

[0143] The embodiment of the present invention provides a data security management method based on user classification and classification, which performs emotional infection analysis on users through machine learning methods according to their publishing and participation behaviors, calculates the proportion of emotions in the sub-data corresponding to the users' publishing and participation behaviors every day, and sets a threshold for the proportion of negative emotions, including:

[0144] The content influence range is obtained based on the content popularity corresponding to the user's publishing and participation behaviors, and the emotional influence analysis of the content influence range is performed based on machine learning;

[0145] Determine the proportion of users' emotions in each time period of the day based on the emotion lag, determine the time when negative emotions are generated, and make negative adjustments to the time and degree of current negative emotions based on the time and degree of the user's corresponding negative emotions in history;

[0146] According to the adjusted proportion of negative emotions and the emotional connection between negative emotions, a proportion threshold of each negative emotion and a quantity threshold of negative emotions are set.

[0147] In this embodiment, the popularity of content refers to the degree of user attention to specific content or the common response of users of the social media platform caused by it, which is quantified by factors such as the number of user views, comments, shares, and likes.

[0148] In this embodiment, emotional contagion analysis uses machine learning and other data analysis methods to annotate and understand the user's emotional state from content and user interactions, and to quantify and simulate user emotions. For example, by analyzing the user's response to the content (such as comments, likes, shares, etc.), as well as the popularity of these contents, it is possible to infer the degree of impact of the content on the user's emotions and the possible trend of emotional changes that may result.

[0149] In this embodiment, the emotion proportion is used to describe the proportion of various emotions (such as happiness, sadness, anger, fear, etc.) expressed by the user during a certain period of time by analyzing the user's emotional response to certain content during the period of time. For example, the user's emotions during a certain period of time may be 80% happiness, 10% sadness, 5% anger and 5% fear. The overall trend of the user's emotions when facing certain content or in a certain situation is determined based on the emotion proportion, and subsequent emotion management or content recommendation operations are performed.

[0150] In this embodiment:

[0151] ;

[0152] in, represents the negative sentiment ratio of the i-th user, Indicates the coefficient of the change trend of the number of friends of the i-th user within time t; represents the proportion of negative content that the i-th user participates in; represents the proportion of negative content posted by the i-th user; A calculation factor indicating the user's engagement with negative content; A calculation factor indicating the probability of a user posting negative content; represents the number of likes of negative content by the i-th user; represents the number of negative comments posted by the i-th user on negative content; represents the number of negative contents posted by the i-th user; represents the number of negative behaviors of the i-th user; represents the proportion of content in which the i-th user participates in active content; represents the proportion of positive content posted by the i-th user; A calculation factor indicating the user's engagement with active content; A calculation factor indicating the positive content posted by the user; represents the number of likes of positive content by the i-th user; represents the number of comments posted by the i-th user on positive content; represents the number of positive contents posted by the i-th user; represents the number of positive behaviors of the i-th user; It represents the cumulative emotion factor of users' likes and comments on negative content, with a value range of (0, 0.06). For example, within one day, user a unfollows all the friends they follow each other, the change trend coefficient is 1, the proportion of content in which user a participates in negative content is 0.75, the proportion of content in which user a posts negative content is 0.05, the proportion of content in which user a posts positive content is 0.1, and the proportion of content in which user a participates in positive content is 0.1, then the negative emotion proportion is calculated to be 0.76.

[0153] In this embodiment, the user's emotional state, including positive, negative and neutral, is identified based on the user's comments, messages, ratings and other behaviors according to the sentiment analysis algorithm.

[0154] The working principle and beneficial effects of the above technical solution are: determining the scope of influence of the content by the popularity of the content involved by users, and then performing emotional infection analysis through machine learning to determine the proportion of users' emotions, predict and identify the time and degree of negative emotions, and adjust the proportion of negative emotions according to the user's historical behavior. Then set the proportion threshold and quantity threshold of each negative emotion, and use this as a reference standard for emotion management, which helps to adjust the user's access level in time, and push positive content to users and restrict access to negative content according to the proportion of negative emotions, thereby realizing refined and dynamic user emotion management.

[0155] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A data security management method based on user classification, characterized in that: include: Step 1: Obtain the main business type and secondary business type of the current social media platform, and collect basic data on current social media users based on the main business type and secondary business type; Step 2: performing cluster analysis on the basic data, obtaining first data according to the cluster analysis result, performing data security identification on the first data, and obtaining second data; Step 3: Input the second data into the machine learning model, set classification rules for the machine learning model, classify users, and formulate a data protection strategy based on the classification results; Step 4: Set the data update time to monitor, evaluate and optimize the basic data; The second data is input into the machine learning model, a classification rule is set for the machine learning model, the users are classified, and a data protection strategy is formulated according to the classification results, including: A security screening structure throughout the entire life cycle of the data is constructed based on the machine learning model and the second data, and a classification rule is set for the machine learning model to screen the structured data and unstructured data in the second data. According to the screening results and the preset classification rules, the primary security level corresponding to each sub-data in the second data is determined; Identify the sensitive data in the second data using the sensitive data automatic identification mechanism, and determine the sensitivity category and corresponding level of each sub-data in the second data according to the data classification and grading method; Adjusting the primary security level of the corresponding sub-data according to the sensitive categories and corresponding levels to obtain a final security level; Set weights for users’ publishing and participation behaviors, and determine the user’s security level based on the weight setting results and the final security level; The users are classified based on their security levels, and protection strategies are set according to the classification results. The protection strategies include: basic security, architecture security, data security and application security.

2. The data security management method based on user classification according to claim 1 is characterized in that: Obtain the main business type and secondary business type of the current social media platform, and collect basic data on the current social media users based on the main business type and secondary business type, including: Determine the main and secondary business types of each social media platform based on the main and secondary functions of all current social media platforms; Extracting first relevant descriptions of all main business types to construct a first feature vector set and extracting second relevant descriptions of all secondary business types to construct a second feature vector set; Performing a first similarity calculation on the first feature vector set, and at the same time, performing a second similarity calculation on the second feature vector set; Extracting a first type exceeding a first set value according to a first similarity calculation result; extracting a second type exceeding a second set value according to a second similarity calculation result; Performing intersection processing on the first type and the second type to determine an intersection part and a non-intersection part; Merge and collect basic data of the types involved in the intersection; The basic data of the types involved in the non-intersection part are collected separately.

3. The data security management method based on user classification according to claim 1 is characterized in that: Performing cluster analysis on the basic data, obtaining first data according to the cluster analysis result, and performing data security identification on the first data to obtain second data, including: Divide user behaviors into primary activities and secondary activities based on the collected data, and classify users based on the activity classification results; Perform cluster analysis on the basic data collected within a period of time based on the data type, use the cluster analysis result as the first data, and use the maximum and second largest values ​​in the cluster analysis results of each data type within the period of time as the daily data of the corresponding data type; Extracting keywords from the daily data, and expanding the daily data range in the first data according to the keywords; Performing data security identification on the first data according to the data security rules and the expanded daily data; The second data is obtained by combining the security identification result with the user classification result corresponding to the first data.

4. The data security management method based on user classification according to claim 1 is characterized in that: The process of adjusting the primary security level of the corresponding sub-data according to the sensitive categories and corresponding levels to obtain the final security level also includes: Perform user privilege management and control based on the screening results. Privilege management and control includes privileged user scenario management, privileged user account discovery, privileged user account integrated management, privileged user account access control, privileged user account automatic password change, and privileged session monitoring. Through the usage scenarios and corresponding usage rules of privileged users, a corresponding approval and authorization mechanism is established to record in detail the access steps, operation logs and privileged operations of privileged users' accounts, and a post-inspection mechanism is provided for the approval and authorization mechanism to conduct scenario control of privileged users on the screening results; Establish a privilege protection mechanism, reasonably allocate authorized access to user data, establish a privilege management platform, and determine the unique access entrance for all privileged users based on the privilege management platform; Based on threat control commands and resource access control, abnormal behavior identification and interception are performed on privileged users with the only access entrance; Update the monitoring capabilities of privileged data access and data violations based on abnormal behaviors, and make secondary adjustments to the primary security level.

5. The data security management method based on user classification according to claim 1 is characterized in that: Weights are set for users’ publishing and participating behaviors, and the user’s security level is determined based on the weight setting results and the final security level, including: Through machine learning, we analyze the emotional contagion of users’ posting and participation behaviors, calculate the proportion of emotions in the sub-data corresponding to users’ posting and participation behaviors every day, and set the threshold for the proportion of negative emotions; Predicting the user's subsequent emotional trend according to the negative emotion ratio threshold, and setting weights for access rights of different data types for the corresponding user's subsequent emotional trend; The weight setting result is added to the final security level of the corresponding sub-data to obtain the user data security level, and the user's security level is weighted and combined with the data level of daily data in the corresponding period of time.

6. The data security management method based on user classification according to claim 1 is characterized in that: Set data update time to monitor, evaluate and optimize basic data, including: Establish a data security view of the second data to monitor the user behavior in the basic data at the data update time, and combine the user's multi-dimensional related data to intelligently predict the abnormal behavior of the user; Adjust the pre-set update time according to the prediction results, collect and monitor user behavior data according to the adjusted time, evaluate the accuracy of the current user classification based on the user behavior data on the user side, and optimize the user classification standards based on the evaluation results.

7. The data security management method based on user classification according to claim 5 is characterized in that: Calculate the proportion of emotions in the sub-data corresponding to the user's posting and participation behaviors every day, and set the threshold for the proportion of negative emotions, including: The content influence range is obtained based on the content popularity corresponding to the user's publishing and participation behaviors, and the emotional influence analysis of the content influence range is performed based on machine learning; Determine the proportion of users' emotions in each time period of the day based on the emotion lag, determine the time when negative emotions are generated, and make negative adjustments to the time and degree of current negative emotions based on the time and degree of the user's corresponding negative emotions in history; According to the adjusted proportion of negative emotions and the emotional connection between negative emotions, a proportion threshold of each negative emotion and a quantity threshold of negative emotions are set.

Images