Multi-industry data grading processing method based on dynamic access control
The method addresses static data classification issues by employing dynamic feature extraction and real-time access control, enhancing data security and flexibility through user attribute and environment integration.
Patent Information
- Application Number
- CN202510388001.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-31
- Publication Date
- 2025-07-15
AI Technical Summary
In the prior art, the data hierarchical processing method is fixed, and the access control lacks dynamic adaptability, resulting in insufficient data security and inflexible access permission management.
By obtaining the data to be classified in the target data space, local timing data segmentation and dynamic feature capture are performed, sensitive identification data are obtained, and dynamic permission verification is carried out in combination with the access user's attributes and environment information, authorizing operations and proof storage through blockchain.
It realizes refined hierarchy based on the dynamic characteristics of data, improves the flexibility of data security and access management, and ensures dynamic adaptability of access control.
Smart Images
Figure CN120316792A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data processing, and particularly to a multi-industry data classification processing method based on dynamic access control. Background Art
[0002] In the context of the current rapid development of digitalization, the amount of data in various industries has grown exponentially, and different types of data have different sensitivities and security requirements. Traditional data classification processing methods usually rely on static rules and are difficult to adapt to complex and changeable business scenarios, resulting in limitations in data security and access management. In addition, existing access control mechanisms often rely on fixed permission policies and are difficult to adjust dynamically, unable to effectively respond to changes in the access environment, and there are problems such as data leakage or access restrictions. Therefore, how to implement a data classification processing method based on dynamic access control to improve the flexibility of access management while ensuring data security has become an important research direction in the field of current data security management. Summary of the Invention
[0003] This application provides a multi-industry data classification processing method based on dynamic access control, which solves the technical problems in the prior art that the data classification processing method is fixed and the access control lacks dynamic adaptability, resulting in insufficient data security and inflexible access permission management.
[0004] This application provides a multi-industry data classification processing method based on dynamic access control, and the method includes:
[0005] Obtain K to-be-classified data in the target data space, traverse the K to-be-classified data and perform local time-series data segmentation according to K segmentation scales to obtain K sets of to-be-classified local time-series data, where K is a positive integer; traverse the K sets of to-be-classified local time-series data to perform dynamic feature capture, obtain K dynamic features of the to-be-classified data for data classification, and obtain K sensitivity identification data; collect the user attribute information and access environment information of the access user in the target data space, and perform dynamic access permission verification in combination with the K sensitivity identification data. If the verification passes, authorize the access user to operate on the sensitivity identification data accessed, and record the operation behavior on the blockchain.
[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:
[0007] First, obtain K data to be classified in the target data space, traverse the K data to be classified, and perform local time-series data segmentation according to K segmentation scales to obtain K sets of local time-series data to be classified, where K is a positive integer. Then, traverse the K sets of local time-series data to be classified to perform dynamic feature capture, obtain dynamic features of the K data to be classified for data classification, and obtain K sensitivity identification data. Finally, collect the user attribute information and access environment information of the access users in the target data space, and perform dynamic verification of access permissions in combination with the K sensitivity identification data. If the verification passes, authorize the access users to operate on the sensitive identification data of the access, and record the operation behavior through the blockchain. This solves the technical problems in the prior art that the data classification processing method is fixed, the access control lacks dynamic adaptability, resulting in insufficient data security and inflexible access permission management, and achieves the technical effect of performing refined classification according to the dynamic features of the data, and realizing dynamic access control in combination with user attributes and access environment, improving data security and access management flexibility. BRIEF DESCRIPTION OF THE DRAWINGS
[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.
[0009] Figure 1 Schematic flow chart of a multi-industry data classification processing method based on dynamic access control provided by an embodiment of the present application;
[0010] Figure 2 Schematic flow chart of obtaining K sensitivity identification data in the multi-industry data classification processing method based on dynamic access control provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0011] By providing a multi-industry data classification processing method based on dynamic access control, the present application solves the technical problems in the prior art that the data classification processing method is fixed, the access control lacks dynamic adaptability, resulting in insufficient data security and inflexible access permission management.
[0012] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts belong to the scope of protection of the present application.
[0013] It should be noted that the terms "include" and "have" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units need not be limited to those clearly listed steps or units, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products, or devices.
[0014] Examples, such as Figure 1 As shown, the embodiments of the present application provide a multi-industry data classification processing method based on dynamic access control. Among them, the method includes:
[0015] Obtain K to-be-classified data in the target data space, traverse the K to-be-classified data, and perform local time-series data segmentation according to K segmentation scales to obtain K sets of to-be-classified local time-series data, where K is a positive integer.
[0016] The target data space refers to a specific environment for storing, managing, and processing data. The target data space includes multiple entities such as data providers, users, service providers, and regulators, and relies on technologies such as blockchain, privacy computing, and secure evidence storage to ensure the security, credibility, and circulation of data.
[0017] In the target data space, first, determine the set of data to be processed, which includes K to-be-classified data. Each to-be-classified data has time-series characteristics, such as sensor data, user behavior logs, etc. Then, for each of the K to-be-classified data, determine K segmentation scales, and each segmentation scale corresponds to one to-be-classified data. Then, traverse the K to-be-classified data, and perform local time-series data segmentation according to the preset K segmentation scales. Specifically, extract time-series data from the K to-be-classified data respectively, divide multiple time windows according to the K segmentation scales, and each window forms a local time-series segment. After segmentation, obtain K sets of to-be-classified local time-series data, and each set reflects the local time-series characteristics of the original data.
[0018] Furthermore, obtaining K to-be-classified data in the target data space, traversing the K to-be-classified data, and performing local time-series data segmentation according to K segmentation scales to obtain K sets of to-be-classified local time-series data includes:
[0019] Extract indicators from the K to-be-classified data according to the preset set of segmentation indicators to obtain K sets of data segmentation indicators; use a segmentation scale recognition network to recognize the K sets of data segmentation indicators to obtain the K segmentation scales.
[0020] Furthermore, the preset set of segmentation indicators includes data window length, data activity, and data access frequency.
[0021] The preset segmentation indicator set includes data window length, data activity and data access frequency, where data window length refers to the time span of the data, that is, the number of data points contained in a window; data activity measures the frequency of data changes in the time dimension, which can usually be obtained by calculating the number of data changes or the rate of change; data access frequency refers to the number of times the data is read or used, reflecting the popularity of the data.
[0022] For the K data to be classified, each data to be classified is traversed, and indicators are extracted based on the preset segmentation indicator set; for each data to be classified, its corresponding segmentation indicator set is calculated including data window length, data activity and data access frequency to form K data segmentation indicator sets.
[0023] The segmentation scale recognition network is used to identify the K extracted data segmentation index sets to determine the optimal segmentation scale for each data sample; the segmentation scale recognition network calculates the optimal segmentation scale through historical data analysis, pattern recognition and adaptive learning, combined with the data access characteristics, time series characteristics and business needs, so that the data can retain key features after segmentation without causing information redundancy or loss. For each data to be classified, the segmentation scale recognition network inputs its segmentation index set and outputs the corresponding segmentation scale.
[0024] Optionally, the segmentation scale recognition network is composed of multiple computing units, including a normalization layer, a feature extraction layer, a scale prediction layer, etc.; the normalization layer first standardizes the input data to ensure that data indicators of different scales have the same distribution to improve the computing stability; then, the feature extraction layer performs deep feature extraction on the input segmentation indicators through model structures such as convolutional neural networks (CNN) and recurrent neural networks (RNN) to mine the association between data activity, access patterns and window features; then, the extracted feature vector is input to the scale prediction layer, which uses regression analysis or classification prediction methods to calculate the segmentation scale suitable for the current data. The segmentation scale recognition network can be trained with historical data to learn the optimal segmentation scale selection strategy.
[0025] Traverse K local time series data sets to be classified to capture dynamic features, obtain K dynamic features of the data to be classified for data classification, and obtain K sensitivity identification data.
[0026] By traversing K sets of local time series data to be classified, dynamic features are captured, thereby extracting dynamic features that can reflect the sensitivity and importance of the data, and obtaining K dynamic features of the data to be classified; further, data classification is performed on the K dynamic features of the data to be classified to obtain K sensitivity identification data.
[0027] Furthermore, if Figure 2As shown, traverse K sets of local time-series data to be classified for dynamic feature capture, obtain K dynamic features of the data to be classified for data classification, and obtain K sensitive identification data, including:
[0028] Respectively perform feature capture on the K sets of local time-series data to be classified to obtain K sets of local time-series data feature sets to be classified; perform dynamic feature interaction on the K sets of local time-series data feature sets to be classified in chronological order to obtain K globally interactive local time-series data features to be classified; use a dynamic normalizer to perform interaction analysis on the K globally interactive local time-series data features to be classified and the K sets of local time-series data feature sets to be classified respectively to obtain the K dynamic features of the data to be classified; use a data classifier to identify the K dynamic features of the data to be classified, and label the K sets of data to be classified based on the identification results to obtain the K sensitive identification data.
[0029] For the K sets of local time-series data to be classified, data feature extraction is performed through a neural network layer to obtain K sets of local time-series data feature sets to be classified. Specifically, each set of local time-series data to be classified consists of data points at multiple time steps and is input into the first layer of the neural network, which is usually a one-dimensional convolutional layer, for extracting local time-series patterns. In the one-dimensional convolutional layer, multiple convolutional kernels slide in the time dimension to extract features of the numerical relationships within the local window, obtaining short-term dependence features. At the same time, the dimension is reduced through a pooling layer (Max Pooling or Average Pooling) to extract the main trend information. For example, after inputting a time series, the convolutional operation calculates the weighted sum within different windows and extracts the maximum value or the mean value in the pooling layer, enabling the model to focus on local important features. Then, the extracted local features pass through a multi-layer fully connected network (MLP) for non-linear transformation and feature combination to further extract high-dimensional features. Among them, the MLP enhances the feature expression ability through activation functions (such as ReLU, Tanh) to extract complex patterns of the data. Finally, after passing through the feature extraction layer of the neural network, the K sets of local time-series data to be classified are converted into K sets of local time-series data feature sets to be classified, and these feature sets will be used for subsequent feature interaction analysis, normalization processing, and data classification.
[0030] Perform dynamic feature interaction on K sets of local time-series data features to be classified in chronological order to obtain global features. Specifically, calculate the cosine similarity between two adjacent sets of local time-series data features to be classified to obtain globally interacting local time-series data features to be classified. On this basis, use a dynamic normalizer to normalize the K globally interacting local time-series data features to be classified and the K sets of local time-series data features to be classified to ensure data alignment at different scales. The normalization method uses mean normalization, and after normalization, K sets of dynamic features of the data to be classified are obtained. Subsequently, use a data classifier to perform sensitivity identification on the K sets of dynamic features of the data to be classified. The data classifier can be classified based on rules, such as setting a threshold. If the mean value is greater than a certain set value, it is determined as highly sensitive data. If the mean value is within a certain range, it is determined as moderately sensitive data. Otherwise, it is low-sensitive data. It can also be based on machine learning methods, training models through classification algorithms such as support vector machines (SVMs) and random forests, and predicting the data sensitivity level based on historical data. Finally, the data to be classified is assigned a sensitivity level (such as low sensitivity, medium sensitivity, high sensitivity) to form K sensitivity-identified data.
[0031] Furthermore, perform dynamic feature interaction on the K sets of local time-series data features to be classified respectively in chronological order to obtain K globally interacting local time-series data features to be classified, including:
[0032] Extract K first local time-series data features to be classified and K second local time-series data features to be classified without replacement from the K sets of local time-series data features to be classified in chronological order from front to back; perform dynamic feature interaction on the K first local time-series data features to be classified and the K second local time-series data features to be classified to obtain K first interacting local time-series data features to be classified; extract K third local time-series data features to be classified without replacement from the K sets of local time-series data features to be classified in chronological order from front to back, and perform dynamic feature interaction on them with the K first interacting local time-series data features to be classified to obtain K second interacting local time-series data features to be classified; continue to perform dynamic feature interaction on the remaining local time-series data features to be classified in the K sets of local time-series data features to be classified based on the K second interacting local time-series data features to be classified until the interaction is completed to obtain the K globally interacting local time-series data features to be classified.
[0033] For K sets of local time-series data features to be classified, dynamic feature interaction is performed in chronological order to construct global interaction features, so as to obtain K sets of global interaction local time-series data features to be classified. First, K first local time-series data features to be classified and K second local time-series data features to be classified are extracted without replacement from each feature set in chronological order from the front to the back. Subsequently, a dynamic feature interaction mechanism is adopted to fuse the K first local time-series data features to be classified and the K second local time-series data features to be classified, resulting in K first interactive local time-series data features to be classified; then, K third local time-series data features to be classified are continuously extracted without replacement from the K sets of local time-series data features to be classified in chronological order, and dynamic feature interaction is performed with the K first interactive local time-series data features to be classified, obtaining K second interactive local time-series data features to be classified. This process continues iteratively, each time extracting new features from the remaining sets of local time-series data features to be classified and fusing them with the currently interacted features until dynamic feature interaction is completed for all K sets of local time-series data features to be classified, finally obtaining K sets of global interaction local time-series data features to be classified.
[0034] Furthermore, performing dynamic feature interaction on the K first local time-series data features to be classified and the K second local time-series data features to be classified to obtain K first interactive local time-series data features to be classified includes:
[0035] Calculating the similarity between the K first local time-series data features to be classified and the K second local time-series data features to be classified to obtain K first interaction similarity sets; constructing a dynamic feature interaction matrix based on the K first interaction similarity sets to generate a first dynamic feature interaction matrix, and performing a convolution operation on the first dynamic feature interaction matrix and the K second local time-series data features to obtain the K first interactive local time-series data features to be classified.
[0036] Calculating the similarity between the K first local time-series data features to be classified and the K second local time-series data features to be classified. The calculation of similarity can adopt various methods, such as cosine similarity, Euclidean distance, Manhattan distance, etc. After the calculation, K first interaction similarity sets are obtained, and each set contains the similarity values between the corresponding feature pairs. Based on the K first interaction similarity sets, a dynamic feature interaction matrix is constructed to form a first dynamic feature interaction matrix. Specifically, the K first interaction similarity sets are normalized using the Softmax formula, that is, the exponential operation is taken on each similarity value and then normalized, and then the data after Softmax normalization is gradually filled into an initially empty matrix to obtain the first dynamic feature interaction matrix.
[0037] After obtaining the normalized interaction matrix, it is used as a weight matrix to perform feature interaction with K second to-be-classified local temporal data features. The weighted features are calculated through matrix multiplication, and then a one-dimensional convolution process is further performed on the interaction matrix using a convolutional neural network (CNN). A trainable convolutional kernel is used to extract local temporal features to enhance the ability to represent the temporal relationship of features. During the convolution operation, the interaction matrix serves as the input tensor and undergoes feature mapping with the set convolutional kernel. The convolution calculation uses a linear transformation plus a bias and is mapped through the non-linear activation function ReLU to ensure the non-linear expression ability of the features and avoid the problem of gradient disappearance. Finally, the matrix multiplication is performed between the convolved interaction feature matrix and the second to-be-classified local temporal data features to generate K first interaction to-be-classified local temporal data features. This feature set can fully integrate the data features of different time series and provide accurate feature representations for subsequent dynamic feature interaction and data classification.
[0038] Furthermore, the dynamic normalizer is constructed based on the dynamic normalization function, where the dynamic normalization function is:
[0039] where, is the dynamic feature of the data to be classified, A nc is the global interaction to-be-classified local temporal data feature, x i is the i-th to-be-classified local temporal data feature in a set of to-be-classified local temporal data features, w i is the weight of the i-th to-be-classified local temporal data feature, ∈ is a parameter less than or equal to 1 for numerical stability, γ and β are preset transformation parameters, both are positive integers less than or equal to 1, and I is the total number of features.
[0040] In the process of constructing the dynamic normalizer based on the dynamic normalization function, first, it is necessary to perform normalization processing on the global interaction to-be-classified local temporal data feature A nc to ensure that the features at different time steps are interactively analyzed on the same scale, and improve the stability and expression ability of feature fusion. The normalization process depends on the dynamic normalization function, and its core idea is to calculate the weighted mean and standardized variance in the set of to-be-classified local temporal data features to achieve dynamic normalization transformation of the data.
[0041] Specifically, first calculate the weighted mean, that is where, x i is the i-th to-be-classified local temporal data feature in a set of to-be-classified local temporal data features, w i is the weight of the i-th to-be-classified local temporal data feature, and I represents the total number of features. Subsequently, calculate the sum of squared errors after removing the mean and obtain the standard deviation. The calculation method of the standard deviation is Among them, the denominator I is averaged to make the variance estimation more stable. At the same time, a numerically stable parameter ∈ is introduced to prevent calculation problems caused by a zero denominator. The normalized eigenvalue is then multiplied element-wise with the global interaction to-be-classified local time-series data feature A nc Perform element-wise multiplication to preserve the information of the original features while enhancing the dynamics of the data. Finally, perform a linear transformation through the scaling parameter γ and the offset parameter β to make the distribution of the normalized data more flexible and adaptable to the changes of different data features.
[0042] Collect the user attribute information and access environment information of the access user in the target data space, and perform dynamic verification of the access permission in combination with the K sensitive identification data. If the verification passes, authorize the access user to operate on the accessed sensitive identification data, and record the operation behavior on the blockchain.
[0043] Collect the user attribute information and access environment information of the access user in the target data space, where the user attribute information includes but is not limited to user identity, role, historical access records, etc., and the access environment information involves context data such as access device, network environment, geographical location, and access time. Subsequently, associate the collected information with the K sensitive identification data, and perform real-time evaluation of the access request based on the dynamic permission management policy. This evaluation process includes calculating the matching degree between the sensitive data and the user attributes, evaluating the credibility of the access environment, and analyzing the consistency of historical behaviors to ensure the legality and security of the access request. If it is verified that the user has the access permission, the system authorizes the access user to perform corresponding operations on the target sensitive identification data, such as reading, modifying, or sharing, etc., and at the same time monitors and records the access behavior. In addition, to ensure the secure traceability of the access operation, the system stores all authorized access and operation behaviors on the blockchain technology to achieve the immutability and traceability of the access records. The blockchain storage process includes calculating the hash of the operation behavior to generate a unique identification value, and writing the access log into the distributed ledger through a smart contract to ensure the transparency and data security of the access process.
[0044] Furthermore, collecting the user attribute information and access environment information of the access user in the target data space, and performing dynamic verification of the access permission in combination with the K sensitive identification data, includes:
[0045] Extract the target access data of the access user and match it with the K sensitive identification data to obtain the target sensitive identification data; construct a first verification vector based on the user attribute information, access environment information, and target sensitive identification data; use a permission validator to verify the first verification vector to obtain a verification result, where the verification result includes verification passed and verification failed.
[0046] During the dynamic verification of access rights, first extract the target access data from the request of the accessing user and match it with K sensitive identification data to determine the target sensitive identification data involved in the user request. Subsequently, based on this target sensitive identification data, combine the user attribute information and access environment information to construct a first verification vector, where the user attribute information includes user identity, permission level, historical access records, etc., and the access environment information involves factors such as access device, network security status, access time, etc. Next, use the permission validator to verify the first verification vector. The permission validator adopts a verification mechanism based on policy rules, machine learning models, or access control lists to perform feature extraction and matching calculations on the verification vector to determine whether the user meets the permission requirements for accessing the target sensitive identification data. Finally, the permission validator outputs the verification result. If the verification passes, the user is allowed to perform the corresponding access operation; otherwise, access is denied, and a security response mechanism is triggered, such as recording the access log, issuing a security alert, or requiring additional identity authentication, to ensure the security of sensitive data and access compliance.
[0047] Furthermore, when the verification result is not passed, generate an access warning instruction for data security warning.
[0048] When the verification result is not passed, the system will trigger a security warning mechanism and generate an access warning instruction for data security warning. First, after the permission validator detects insufficient permissions of the accessing user or an abnormal access environment, it immediately constructs warning information, which includes the identity identification of the accessing user, access time, access device, network environment, target access data, and the specific reason for the failed verification. Subsequently, the system evaluates the risk level of the access behavior according to the preset security policy and determines the corresponding security response measures. For low-risk access failure situations, such as access denial due to insufficient user permissions, the system can return a permission insufficient prompt to the user and record the log for subsequent auditing. For medium and high-risk access failure situations, such as abnormal device access, multiple failed attempts within a short period of time, or illegal access involving highly sensitive data, the system will send an access warning instruction to the security management platform to notify the security administrator for manual review. At the same time, it can trigger automatic security defense policies, such as temporarily blocking the user account, restricting access to related devices, enhancing the access authentication method (such as multi-factor authentication), or writing access exception records to the blockchain deposit system, to ensure data security and system stability.
[0049] Furthermore, when the verification passes, obtain the historical verification key set and the current random verification key of the accessing user; determine whether the coincidence rate of the random verification key and the historical verification key set exceeds the preset coincidence rate threshold. If so, obtain a key regeneration instruction.
[0050] When the verification passes, the system first extracts the historical verification key set of the accessing user from the user's historical access records, and simultaneously generates a current random verification key. Subsequently, the system performs a matching calculation on the current random verification key and the historical verification key set, calculates the coincidence rate between the two, and compares it with a preset coincidence rate threshold. If the calculation result shows that the coincidence rate exceeds the threshold, it indicates that the randomness of the current key may be insufficient and there are certain security risks. Therefore, the system will trigger the key regeneration mechanism, generate a key regeneration instruction, and notify the key management module to regenerate a new random verification key to ensure the uniqueness and security of the key, and prevent potential replay attacks or key leakage risks. At the same time, the system can also record this key regeneration operation in the blockchain evidence storage system to ensure the traceability and transparency of key management.
[0051] In summary, the embodiments of the present application have at least the following technical effects:
[0052] First, obtain K data to be classified in the target data space, traverse the K data to be classified, and perform local time series data segmentation according to K segmentation scales to obtain K sets of local time series data to be classified, where K is a positive integer. Then, traverse the K sets of local time series data to be classified to perform dynamic feature capture, obtain the dynamic features of the K data to be classified for data grading, and obtain K sensitivity identification data. Finally, collect the user attribute information and access environment information of the accessing user in the target data space, and perform dynamic access verification in combination with the K sensitivity identification data. If the verification passes, authorize the accessing user to operate on the sensitive identification data accessed, and record the operation behavior on the blockchain. This solves the technical problems in the prior art that the data grading processing method is fixed and the access control lacks dynamic adaptability, resulting in insufficient data security and inflexible access right management, and achieves the technical effect of performing refined grading according to the dynamic features of the data, and realizing dynamic access control in combination with user attributes and access environment, improving data security and access management flexibility.
[0053] It should be noted that the above sequence of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above describes specific embodiments of this specification. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0054] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.
[0055] This specification and the drawings are merely exemplary illustrations of the present application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications therein.
Claims
1. A multi-industry data classification and processing method based on dynamic access control, characterized in that, The method includes: Obtaining K data to be classified in the target data space, traversing the K data to be classified, and performing local time-series data segmentation according to K segmentation scales to obtain K sets of local time-series data to be classified, where K is a positive integer; Traversing the K sets of local time-series data to be classified for dynamic feature capture, obtaining dynamic features of the K data to be classified for data grading, and obtaining K sensitivity identification data; Collecting user attribute information and access environment information of the access user in the target data space, and performing dynamic verification of access permissions in combination with the K sensitivity identification data. If the verification is passed, the access user is authorized to operate on the sensitive identification data being accessed, and the operation behavior is stored on the blockchain.
2. The multi-industry data classification processing method based on dynamic access control according to claim 1, wherein, Obtaining K data to be classified in the target data space, traversing the K data to be classified, and performing local time-series data segmentation according to K segmentation scales to obtain K sets of local time-series data to be classified, including: Extracting indicators from the K data to be classified according to a preset set of segmentation indicators to obtain K sets of data segmentation indicators; Using a segmentation scale recognition network to recognize the K sets of data segmentation indicators to obtain the K segmentation scales.
3. The multi-industry data classification processing method based on dynamic access control according to claim 2, characterized in that The preset set of segmentation indicators includes data window length, data activity, and data access frequency.
4. The multi-industry data classification processing method based on dynamic access control according to claim 1, characterized in that, Traversing the K sets of local time-series data to be classified for dynamic feature capture, obtaining dynamic features of the K data to be classified for data grading, and obtaining K sensitivity identification data, including: Performing feature capture on the K sets of local time-series data to be classified respectively to obtain K sets of local time-series data features to be classified; Performing dynamic feature interaction on the K sets of local time-series data features to be classified in chronological order to obtain K globally interactive local time-series data features to be classified; Using a dynamic normalizer to perform interaction analysis on the K globally interactive local time-series data features to be classified and the K sets of local time-series data features to be classified respectively to obtain the dynamic features of the K data to be classified; Using a data grader to recognize the dynamic features of the K data to be classified, and identifying the K data to be classified based on the recognition result to obtain the K sensitivity identification data.
5. The multi-industry data classification and processing method based on dynamic access control according to claim 4, characterized in that, Performing dynamic feature interaction on the K sets of local time-series data features to be classified in chronological order to obtain K globally interactive local time-series data features to be classified, including: Respectively extracting K first local time-series data features to be classified and K second local time-series data features to be classified from the K sets of local time-series data features to be classified in chronological order without replacement; Performing dynamic feature interaction on the K first local time-series data features to be classified and the K second local time-series data features to be classified to obtain K first interactive local time-series data features to be classified; Respectively extracting K third local time-series data features to be classified from the K sets of local time-series data features to be classified in chronological order without replacement, and performing dynamic feature interaction with the K first interactive local time-series data features to be classified to obtain K second interactive local time-series data features to be classified; Continue to perform dynamic feature interaction on the remaining to-be-classified local temporal data features in the K to-be-classified local temporal data feature sets based on the K second to-be-classified local temporal data features with feature interaction until the interaction is completed, and obtain the K globally interacted to-be-classified local temporal data features.
6. The multi-industry data grading and processing method based on dynamic access control according to claim 5, characterized in that Perform dynamic feature interaction on the K first to-be-classified local temporal data features and the K second to-be-classified local temporal data features to obtain K first interacted to-be-classified local temporal data features, including: Calculate the similarity between the K first to-be-classified local temporal data features and the K second to-be-classified local temporal data features to obtain K first interaction similarity sets; Construct a dynamic feature interaction matrix based on the K first interaction similarity sets, generate a first dynamic feature interaction matrix, and perform a convolution operation on the first dynamic feature interaction matrix and the K second to-be-classified local temporal data features to obtain the K first interacted to-be-classified local temporal data features.
7. The multi-industry data classification and processing method based on dynamic access control according to claim 4, wherein Including: Construct the dynamic normalizer based on a dynamic normalization function, where the dynamic normalization function is: Among them, is the dynamic feature of the data to be classified, A nc is the global interaction feature of the local time series data to be classified, x i is the i-th local time series data feature to be classified in a set of local time series data features to be classified, w i is the weight of the i-th local time series data feature to be classified, ∈ is a parameter less than or equal to 1 for numerical stability, γ and β are preset transformation parameters, both are positive integers less than or equal to 1, and I is the total number of features.
8. The multi-industry data classification and processing method based on dynamic access control according to claim 1, characterized in that Collect the user attribute information and access environment information of the accessing user in the target data space, and perform dynamic access permission verification in combination with the K sensitivity identification data, including: Extract the target access data of the accessing user and match it with the K sensitivity identification data to obtain target sensitivity identification data; Construct a first verification vector based on the user attribute information, access environment information, and target sensitivity identification data; Use a permission verifier to verify the first verification vector to obtain a verification result, where the verification result includes verification passed and verification failed.
9. The multi-industry data classification processing method based on dynamic access control according to claim 8, characterized in that, When the verification result is verification failed, generate an access warning instruction for data security warning.
10. The multi-industry data classification and processing method based on dynamic access control according to claim 1, wherein, It also includes: When the verification is passed, obtain the historical verification key set and the current random verification key of the accessing user; Judge whether the coincidence rate of the random verification key and the historical verification key set exceeds a preset coincidence rate threshold. If so, obtain a key regeneration instruction.