A method and system for separating permissions between C-end data input and B-end data query in an integrated oilfield-level mineral rights and reserves platform.

By constructing an isolated architecture between C-end and B-end and multi-dimensional anomaly detection in the oilfield-level mineral rights and reserves data platform, combined with blockchain traceability tags and expert review feedback closed loop, the problems of data privacy leakage and high false alarm rate have been solved, and the deep integration and adaptive control of data security and quality have been achieved.

CN123116325APending Publication Date: 2026-11-17KARAMAY HONGYOU SOFTWARE
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611285240.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-24
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing oilfield-level mineral rights and reserves data management platforms face high risks of data privacy breaches and difficulties in ensuring data quality. In particular, the high false alarm rate of traditional role-based access control and AI models leads to a disconnect between data security and quality control.

Method used

An isolated architecture is constructed between the C-end and B-end, and one-way asynchronous communication is carried out through independent message queues. A multi-dimensional anomaly detection engine and blockchain traceability tags are used to calculate data credibility factors and dynamically allocate permissions. The model self-evolution is achieved by combining expert review feedback closed loop.

Benefits of technology

It achieves physical and logical isolation between the C-end and the B-end, reduces the risk of data leakage, improves the accuracy and adaptability of data quality control, reduces the false alarm rate, and ensures the deep integration of data security and quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116325A_ABST
    Figure CN123116325A_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for separating permissions between C-end data input and B-end data query on an integrated platform for oilfield-level mineral rights and reserves. The method constructs an isolated architecture for the C-end input channel and the B-end query channel through independent message queues, prohibiting the B-end from accessing the physical storage address of the data. On the C-end, a multi-dimensional anomaly detection engine combined with spatiotemporal context is used to detect the input data, calculate the data credibility factor, and divide it into safe zones, blocking zones, or gray zones. Data in the gray zone is labeled as "suspected anomaly" and stored in the database with low credibility. On the B-end, a multi-dimensional dynamic permission allocation mechanism reads the blockchain traceability tag and performs adaptive desensitization processing based on the data credibility factor and user attributes. This invention achieves dynamic linkage between data quality control and access permissions, effectively solving the problems of internal data leakage risks and data loss caused by AI detection false alarms, ensuring the privacy and availability of oilfield reserve data.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of data processing and information security technology in the petroleum industry, and in particular to a method and system for separating permissions for data input by C-end and data query by B-end in an integrated platform for oilfield-level mineral rights and reserves. Background Technology

[0002] With the advancement of digital oilfield construction, oilfield-level mineral reserve data (such as proven reserves, production data, and well logging curves) has become a core asset for enterprises. This data is typically uploaded by frontline operators via the C-end (data entry terminal) and analyzed and used for decision-making by management or researchers via the B-end (query terminal).

[0003] However, existing data management platforms face two major challenges:

[0004] High risk of data privacy breaches: Traditional platforms often employ role-based static access control (RBAC). Once account permissions are stolen or abused by internal personnel, sensitive data can be obtained in bulk. Furthermore, C-end and B-end users often share databases or connect via simple APIs, allowing attackers to exploit vulnerabilities to directly connect to the database and steal data, leading to serious data breaches.

[0005] Data quality is difficult to guarantee and false alarms are common: Oilfield data comes from complex sources and human input errors are frequent. Although automated anomaly detection algorithms (such as Isolation Forest and LSTM) have been introduced for screening, AI models often exhibit false alarms (i.e., misjudging real but special data, such as outliers from high-yield wells, as erroneous data and blocking their entry). This not only leads to the loss of valuable data but also increases the burden of manual review. Furthermore, the models lack a self-evolution mechanism, making it difficult for the false alarm rate to decrease over time.

[0006] Therefore, there is an urgent need for an intelligent permission separation method that can achieve physical and logical isolation between C-end and B-end, dynamically adjust access permissions based on data quality, and has self-evolution capabilities. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for separating permissions between data entered by the C-end and data queried by the B-end in an integrated platform for oilfield-level mineral rights and reserves, so as to solve the problems of high risk of data privacy leakage, disconnect between data quality control and access control, and high false alarm rate of anomaly detection in the prior art.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0009] In a first aspect, the present invention provides a method for separating permissions between data entered by the C-end and data queried by the B-end in an integrated platform for oilfield-level mineral rights and reserves, comprising the following steps:

[0010] Construct an isolated architecture for the C-end data entry channel and the B-end data query channel. The isolated architecture uses an independent message queue for one-way asynchronous communication and configures access control policies to prevent the B-end from obtaining the physical storage address of the data.

[0011] In the C-end data entry channel, a multi-dimensional anomaly detection engine is used to detect the entered data. The detection includes dynamic pattern detection based on spatiotemporal context. The data credibility factor is calculated based on the detection results, and the data is divided into safe zone, blocked zone or gray zone according to the anomaly score.

[0012] For data in the gray area, it is allowed to be entered into the database with a lower initial data credibility factor and labeled as "suspected anomaly - pending verification", generating a blockchain traceability label containing the data credibility factor and label;

[0013] In the B-end data query channel, query requests are received, and blockchain traceability tags are read through a multi-dimensional dynamic permission allocation mechanism; real-time decisions are made based on data credibility factors and user attributes to determine access permissions and de-identification strategy levels;

[0014] Perform adaptive data masking; if the queried object has a "suspected anomaly" label, include a confidence level indicator in the returned data.

[0015] Construct a closed loop of expert review and feedback: Generate training samples based on the experts' approval or rejection instructions for grayscale data, and perform incremental training on the multi-dimensional anomaly detection engine to adjust the detection threshold.

[0016] Preferably, dynamic pattern detection based on spatiotemporal context includes:

[0017] Spatial outlier detection: Extract the geographical location of the entered data, use the K-Nearest Neighbor (KNN) algorithm to retrieve historical neighboring well data within a preset radius, and calculate the characteristic deviation between the current data and the neighboring well data group;

[0018] Time series anomaly detection: Construct a Long Short-Term Memory (LSTM) network prediction model for the monitored data and calculate the reconstruction error between the actual entered values ​​and the predicted values;

[0019] If the feature deviation or reconstruction error exceeds the preset threshold, it is judged as abnormal and a corresponding abnormal score is generated.

[0020] Preferably, the execution steps of the expert review and feedback closed loop include:

[0021] When the data is in the grayscale area, an expert review task is generated and pushed to the expert terminal;

[0022] Receive confirmation instructions from experts regarding the review task. If the instruction is "approved", the original detection result is determined to be a false alarm, the corresponding data is marked as "true value", and it is added to the training set as a negative sample.

[0023] If the instruction is "reject", the original detection result is determined to be a true anomaly, and the corresponding data and its features are added to the training set as positive samples;

[0024] The training set is used to fine-tune or incrementally learn the model parameters of the multi-dimensional anomaly detection engine.

[0025] Preferably, the data credibility factor is calculated using a weighted fusion algorithm:

[0026] Where T is the data reliability factor. For compliance scoring, For consistency scoring based on an anomaly detection engine, Rate the source;

[0027] The weighting coefficients are dynamically adjusted based on the input scenario. When an untrusted network is detected as the input environment, the weighting is increased. weight .

[0028] Preferably, the method further includes a false alarm suppression step based on external context:

[0029] When a specific type of anomaly is detected, the system automatically queries the associated job logs or equipment status;

[0030] If the work log shows that there were work operations in the corresponding time period, or the equipment status shows a malfunction, then the anomaly is determined to be a non-geological data anomaly, and the anomaly alarm is automatically suppressed while maintaining the normal data reliability factor.

[0031] Preferably, the multi-dimensional dynamic permission allocation mechanism executes the following logic during decision-making:

[0032] If the data credibility factor is in the high credibility range and the user role permissions match, then plaintext or low-level de-identified data can be returned.

[0033] If the data confidence factor is in the gray range, the data is allowed to be returned, but the "data is disputed" confidence level prompt will be forcibly displayed on the query results page, and the weight of the data in the statistical aggregation will be limited.

[0034] If the data credibility factor is in the low credibility range, the highest level of de-identification strategy will be enforced, and the return of numerical plaintext will be prohibited.

[0035] Preferably, the independent message queue uses encrypted topics for data transmission, and only the writing process of the C-end data entry channel is allowed to write message bodies containing data digests to the encrypted topics, and only the consumption process of the B-end data query channel is allowed to read the message bodies, thereby realizing unidirectional data flow on the link.

[0036] Preferably, the method supports multi-tenant data isolation based on geological spatial topology:

[0037] Logical isolation boundaries are constructed based on the spatial topological relationships of geological structural units;

[0038] When a query request involves data across different geological tectonic units, the merged cross-domain data is only allowed to be returned if the user's role permissions include "whole basin level" authorization; otherwise, only data from a single tectonic unit is returned.

[0039] Preferably, the adaptive desensitization process specifically includes:

[0040] Based on the determined desensitization strategy level, a dynamic replacement algorithm is used for sensitive reserve values, or a random perturbation algorithm is used for geological coordinate data.

[0041] For data in the grayscale range, while preserving the original numerical visualization, the confidence interval range of the data distribution is overlaid to alert users to the potential uncertainty of the data.

[0042] Secondly, this invention provides an integrated platform for oilfield-level mineral rights and reserves, based on a permission separation system for data input by the C-end and data query by the B-end, comprising:

[0043] The C-end data entry channel is equipped with a multi-dimensional anomaly detection engine and a grayscale classification module, which are used to perform anomaly detection based on spatiotemporal context and calculate data credibility factors.

[0044] The B-end data query channel is configured with a multi-dimensional dynamic permission allocation mechanism and a data de-identification processor.

[0045] The expert review and feedback module is configured to display grayscale data and collect expert correction instructions, which are then converted into training samples and fed back to the multi-dimensional anomaly detection engine.

[0046] An isolated communication bus is used to connect the C-end and the B-end to implement one-way asynchronous communication and physical address hiding.

[0047] The beneficial effects of this invention are as follows: Architecture-level security isolation: One-way communication between the C-end and the B-end is realized through independent message queues and encrypted topics, which physically and logically cuts off the path for the B-end to directly access the C-end database or physical storage, thus eliminating data breaches and unauthorized access from the root of the architecture.

[0048] Quality-Driven Access Control: This innovative approach introduces a "data credibility factor," directly mapping data quality assessment results from the consumer (C-end) to access permission levels for businesses (B-end). The more credible the data, the more open the permissions; the more suspicious the data, the stricter the control. This achieves a deep integration of data security and data quality.

[0049] To address the "data black hole" caused by false alarms: The concept of a "grayscale zone" is introduced. Data that appears abnormal but is not obvious is not directly rejected; instead, it is tagged, stored in the database, and displayed at a lower level. This preserves the data while highlighting the risks, effectively preventing the loss of high-value geological data due to AI false alarms.

[0050] Model self-evolution capability: An expert review and feedback loop is established, using expert corrections (approval / rejection) of grayscale data as training samples to continuously incrementally train the anomaly detection model. As the system runs longer, the model's understanding of the geological patterns of specific oilfields deepens, the false alarm rate decreases significantly, and the system becomes increasingly intelligent. Attached Figure Description

[0051] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a schematic diagram of the overall architecture of the permission separation method for C-end data entry and B-end data query provided in an embodiment of the present invention.

[0053] Figure 2 This is a flowchart illustrating the data flow and access control process for C-end data entry and B-end data query provided in an embodiment of the present invention.

[0054] Figure 3 This is a flowchart illustrating the data flow and access control process for C-end data entry and B-end data query provided in an embodiment of the present invention.

[0055] Figure 4 This is a schematic diagram illustrating the principle of the dynamic desensitization strategy for B-end provided in an embodiment of the present invention.

[0056] Figure 5 This is a logical diagram of the abnormal behavior detection module provided in an embodiment of the present invention. Detailed Implementation

[0057] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0058] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0059] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0060] Reference Figures 1-5 This is one embodiment of the present invention, which provides a method for separating permissions between data entered by the C-end and data queried by the B-end in an integrated platform for oilfield-level mineral rights and reserves, including the following steps:

[0061] Construct an isolated architecture for the C-end data entry channel and the B-end data query channel. The isolated architecture uses an independent message queue for one-way asynchronous communication and configures access control policies to prevent the B-end from obtaining the physical storage address of the data.

[0062] The specific steps are as follows: First, construct a channel architecture with both physical and logical isolation. Firstly, independently deploy C-end data entry services and B-end data query services within the system, severing the direct database connection between the two. Then, deploy a dedicated, independent message queue cluster between the two as the sole data relay bus. This message queue employs an encrypted Topic mechanism to ensure complete decoupling between the C-end and B-end at the underlying storage level, allowing logical interaction solely through this bus.

[0063] The second step is to configure a unidirectional asynchronous communication strategy based on the principle of least privilege. A strict access control list is set up in the message queue management interface, granting only the C-end entry process permission to write messages to a specified encrypted topic, and only the B-end query process permission to read messages from that topic. This establishes a unidirectional data flow from the C-end to the B-end, with all data transmission using an asynchronous, non-blocking mode, avoiding the risk of the B-end directly accessing the C-end database through a synchronous interface.

[0064] The third step involves implementing mechanisms to hide and block the physical storage address of data. Before the data is written to the message queue on the C-side, the data payload is encapsulated to completely remove any metadata information containing the physical storage path, database connection string, or server IP address. Combined with access control policies, the architecture prohibits B-side processes from initiating any address requests pointing to the physical storage layer, ensuring that the B-side can only consume encapsulated summary data or de-identified data, thereby eliminating the possibility of obtaining the original storage address through reverse engineering or direct database dumping on the B-side.

[0065] In the C-end data entry channel, a multi-dimensional anomaly detection engine is used to detect the entered data. The detection includes dynamic pattern detection based on spatiotemporal context. The data credibility factor is calculated based on the detection results, and the data is divided into safe zone, blocked zone or gray zone according to the anomaly score.

[0066] This step specifically involves: In the first stage, a multi-dimensional anomaly detection engine is used to perform dual feature detection on the entered data. First, a benchmark check is performed using a rule engine to verify the data's format integrity, numerical and physical limits, and logical dependencies between fields, generating a compliance score. Subsequently, the engine initiates dynamic pattern detection based on spatiotemporal context, extracts geographical location information from the data, and calls the K-Nearest Neighbors (KNN) algorithm to retrieve historical neighboring well data within a preset radius. It then calculates the characteristic deviation between the current data and the neighboring well group to identify spatial outliers. Simultaneously, for the monitoring data, a Long Short-Term Memory (LSTM) network is used to calculate the reconstruction error between the actual recorded value and the predicted value to identify time-series anomalies. Finally, the aforementioned anomaly scores are converted into a consistency score. .

[0067] The second stage involves performing weighted fusion calculations to determine the data credibility factor. The system combines authentication information from the data entry point with the environmental context to assess the entry personnel's qualification level, device security status, and network environment, generating a source credibility score. Subsequently, the system dynamically adjusts the weighting coefficients based on the security level of the current input scenario. For example, when an untrusted network environment is detected, the weight W2 of the consistency score is increased, and finally the formula is used:

[0068] Calculate the final data reliability factor T.

[0069] In the third stage, the system implements partitioning and labeling based on the calculated data credibility factor. The system compares the data credibility factor T with preset security and blocking thresholds: if T is higher than the security threshold, the data is determined to belong to the safe zone, is entered into the database normally, and assigned a high credibility level; if T is lower than the blocking threshold, the data is determined to belong to the blocking zone, submission is directly rejected, and the user is forced to make corrections; if T is between the security and blocking thresholds, the data is determined to belong to the gray zone, and the system allows it to enter the database with a lower initial credibility factor, while automatically labeling it "suspected anomaly - pending review." This label is then bound to the data subject to generate a blockchain traceability label, triggering subsequent expert review processes and B-end downgrade access strategies.

[0070] For data in the gray area, it is allowed to be entered into the database with a lower initial data credibility factor and labeled as "suspected anomaly - pending verification", generating a blockchain traceability label containing the data credibility factor and label;

[0071] Specifically, this step involves the following steps: First, the system performs a threshold determination on the calculated data credibility factor T to confirm that its value falls within a preset grayscale range (i.e., between the blocking threshold and the safety threshold). Once the data is determined to be grayscale data, the system will bypass the forced blocking mechanism, retain the lower calculation result as the data's "initial data credibility factor," and construct a temporary record object in memory containing the original data and the factor. Simultaneously, a "suspected anomaly - pending verification" status flag will be appended to this object.

[0072] Secondly, the system uses a hash algorithm to extract a unique data digest hash value from the key business fields of the entered data (such as well location coordinates, reserve values, etc.), and packages the hash value, the initial data credibility factor T, the "suspected anomaly - pending review" label, and the entry timestamp into a transaction data structure that conforms to the blockchain interface standard.

[0073] Subsequently, the system calls the smart contract interface of the underlying blockchain service to upload the encapsulated transaction data structure to the blockchain for evidence storage. After the blockchain network nodes complete consensus verification, a digital certificate containing the block height and transaction hash is generated, which serves as an immutable blockchain traceability label.

[0074] Finally, the system writes the original data into the business database, and at the same time, it associates and stores the generated blockchain traceability tags, initial data credibility factors, and abnormal status tags into the corresponding record indexes, completing the data entry operation with low credibility and warning tags, and triggering an asynchronous message notification to the audit terminal to enter the expert review process.

[0075] In the B-end data query channel, query requests are received, and blockchain traceability tags are read through a multi-dimensional dynamic permission allocation mechanism; real-time decisions are made based on data credibility factors and user attributes to determine access permissions and de-identification strategy levels;

[0076] Specifically, this step involves the following steps: First, the B-end data query channel receives and parses the query request submitted by the user. Then, the user's contextual user attributes are extracted through the identity authentication module. These user attributes include the user's role ID, department, security level, and operating environment information, which serve as one of the input variables for permission decisions.

[0077] Secondly, based on the query request index, the system locates the target data, calls the blockchain query interface, and obtains the blockchain traceability tag associated with the data. After verifying the integrity of the tag, the system parses the tag content to extract the key indicators generated in the previous steps, namely the data credibility factor T and the anomaly type tag carried by the data (such as "suspected anomaly - pending review", "logical conflict", etc.).

[0078] Subsequently, a multi-dimensional dynamic permission allocation mechanism is activated, inputting user attributes and data credibility factors into a pre-set decision matrix for real-time calculation. The system makes judgments based on rules: if the data credibility factor T is below a preset threshold or the anomaly type label shows high risk, permission is restricted; if the data credibility factor T is in the gray range, permission determination requires additional warning conditions; if the data credibility factor T is in the high credibility range and the user role matches, permission is granted.

[0079] Finally, based on the decision results, the final access permissions and de-identification strategy level for this query are determined. The system explicitly specifies the strategy level; for example, a "high-level de-identification" strategy (prohibiting the return of plaintext) is enforced for low-confidence data, a "plaintext with warning" strategy is enforced for grayscale data, and a "standard de-identification" strategy is enforced for high-confidence data. The strategy instruction is then sent to the data de-identification processor to perform subsequent data rendering actions.

[0080] Perform adaptive data masking; if the queried object has a "suspected anomaly" label, include a confidence level indicator in the returned data.

[0081] This step specifically involves: First, performing field-level data masking transformations matching the strategy level. The data masking processor processes sensitive fields (such as reserve values ​​and geological coordinates) in the query results field by field, based on the masking strategy level determined in the preceding decisions. For high-level masking requirements, numerical range generalization or coordinate random perturbation algorithms are used to hide the true values ​​and prevent inversion; for medium- and low-level requirements, masking or partial hiding algorithms (such as partially masking phone numbers or precise values) are used to generate a masked data view that conforms to current security standards.

[0082] The second step is to identify anomaly tags and generate confidence level prompts. The system parses the blockchain traceability tags to determine if there is a "suspected anomaly - pending verification" marker. If this marker is detected, the system automatically generates confidence level prompt text according to a preset rule template (e.g., "Data is disputed, confidence level: medium" or "This data has not been approved by experts and is for reference only"). This prompt text is designed to be highlighted in a specific color or displayed as an icon in the user interface to clearly distinguish it from normal data.

[0083] The third step is to encapsulate the data view and output the response. The system logically binds and encapsulates the anonymized data view with the generated confidence level hints to construct the final response data packet. If the data is confirmed to have a "suspected anomaly" tag, a confidence level hint field is explicitly injected into the data packet returned to the B-end frontend to ensure that users can receive real-time warnings about data quality while browsing the anonymized data; if the data does not have this tag, only the clean anonymized data is returned, completing the query response.

[0084] Construct a closed loop of expert review and feedback: Generate training samples based on the experts' approval or rejection instructions for grayscale data, and perform incremental training on the multi-dimensional anomaly detection engine to adjust the detection threshold.

[0085] The specific steps are as follows: First, the background review module extracts data records labeled "suspected anomaly - pending review" according to the preset scanning strategy, generates an expert review task, and pushes the task to the expert terminal. The expert views the original data values, spatiotemporal context features (such as comparison with adjacent wells, historical trends) and the basis for the system's judgment of anomalies on the terminal. Based on geological knowledge and business experience, the expert makes a judgment. If the data is confirmed to be true and valid, an "approval" instruction is issued. If the data is confirmed to be an input error, a "rejection" instruction is issued.

[0086] Secondly, the system receives expert instructions and transforms the data into training samples. The specific logic is as follows: when the "approval" instruction is received, the system determines that the original detection result is a false alarm and marks the feature vector of the data as a "negative sample" (i.e., a non-abnormal sample); when the "rejection" instruction is received, the system determines that the original detection result is correct and marks the feature vector of the data as a "positive sample" (i.e., an abnormal sample); the system stores the marked samples along with their feature vectors (including spatial bias, reconstruction error, etc.) into the feedback training dataset.

[0087] Finally, the system initiates the model self-evolution process, using the updated feedback training dataset to perform incremental training or fine-tuning on the machine learning models (such as Long Short-Term Memory Network LSTM or K-Nearest Neighbors algorithm KNN) in the multi-dimensional anomaly detection engine; the model parameters are updated through the backpropagation algorithm, the decision boundary and detection threshold for anomaly judgment are dynamically adjusted, and the trained new model parameters are loaded into the detection engine in real time, thereby achieving continuous correction and improvement of anomaly detection accuracy.

[0088] Preferably, dynamic pattern detection based on spatiotemporal context includes:

[0089] Spatial outlier detection: Extract the geographical location of the entered data, use the K-Nearest Neighbor (KNN) algorithm to retrieve historical neighboring well data within a preset radius, and calculate the characteristic deviation between the current data and the neighboring well data group;

[0090] Time series anomaly detection: Construct a Long Short-Term Memory (LSTM) network prediction model for the monitored data and calculate the reconstruction error between the actual entered values ​​and the predicted values;

[0091] If the feature deviation or reconstruction error exceeds the preset threshold, it is judged as abnormal and a corresponding abnormal score is generated.

[0092] Specifically, dynamic pattern detection based on spatiotemporal context is mainly implemented through the following process, referring to... Figure 3 .

[0093] First, spatial outlier detection is performed. The system extracts the geographic coordinates from the data to be entered, and uses these coordinates as the center to retrieve historical neighboring well data within a preset radius from the entire geological database. The K-Nearest Neighbors (KNN) algorithm is used to calculate the spatial distance between the current data point and neighboring wells, and the K nearest wells are selected as the reference neighboring well set. The system reads the corresponding geological attribute parameters (such as porosity, permeability, etc.) from this set, calculates the statistical mean and variance of the neighboring well data, and substitutes the attribute values ​​of the currently entered data into the distribution model to calculate its characteristic deviation (e.g., Z-score deviation) from the neighboring well group.

[0094] Next, time series anomaly detection is performed. For monitoring data with time-series characteristics, the system uses a pre-trained Long Short-Term Memory (LSTM) prediction model for processing. The system uses the input time as the cutoff point and extracts a preset length of time-series data (e.g., daily oil production over the past 30 days) as the input feature vector, which is then fed into the LSTM model to calculate the theoretical predicted value for that time point. Subsequently, the system calculates the difference between the actual input value and the model's predicted value to obtain the reconstruction error.

[0095] Finally, the system compares the aforementioned feature deviations and reconstruction errors with their respective preset safety thresholds. If any indicator exceeds the threshold, the data is determined to be abnormal, and a standardized anomaly score is generated based on the magnitude of the exceedance. This score serves as an important parameter for subsequent calculations of the data credibility factor.

[0096] Preferably, the execution steps of the expert review and feedback closed loop include:

[0097] When the data is in the grayscale area, an expert review task is generated and pushed to the expert terminal;

[0098] Receive confirmation instructions from experts regarding the review task. If the instruction is "approved", the original detection result is determined to be a false alarm, the corresponding data is marked as "true value", and it is added to the training set as a negative sample.

[0099] If the instruction is "reject", the original detection result is determined to be a true anomaly, and the corresponding data and its features are added to the training set as positive samples;

[0100] The training set is used to fine-tune or incrementally learn the model parameters of the multi-dimensional anomaly detection engine.

[0101] The specific implementation method of the expert review and feedback closed loop is as follows, refer to Figure 4 .

[0102] First, the system automatically triggers an expert review process for data falling into the grayscale zone. The backend service extracts the detailed attributes of the data (such as well location, value, and entry time) and the preliminary diagnostic information output by the multi-dimensional anomaly detection engine (such as spatial deviation and time reconstruction error), encapsulates it into a review task, and sends it in real time to the authorized expert's terminal device (such as a mobile APP or PC workbench) through a message push interface, reminding the expert to intervene manually.

[0103] Secondly, the system receives and parses the confirmation instructions returned by experts for the review task. If the expert confirms that the data is true and valid (for example, a high-yield well has abnormal values ​​due to the adoption of a new process), the expert clicks the "Approve" instruction. At this time, the system determines that the original detection result is a false alarm, marks the data as "true value," and adds its feature vector as a negative sample (i.e., a non-abnormal sample) to the preset training dataset to correct the model's cognitive bias towards specific complex geological patterns. Conversely, if the expert confirms that the data has input errors or quality problems, the expert clicks the "Reject" instruction. The system determines that the original detection result is accurate, adds the data and its abnormal features as a positive sample (i.e., an abnormal sample) to the training dataset to strengthen the model's ability to identify errors.

[0104] Finally, the system uses the updated training set to train the multi-dimensional anomaly detection engine. For detection models based on neural networks (such as LSTM), incremental learning or parameter fine-tuning strategies are employed, using backpropagation with new samples to adjust network weights; for distance-based algorithms (such as KNN), the feature sample library is dynamically updated. Through this closed-loop process, the system can continuously optimize the detection threshold and decision boundary, thereby continuously reducing the false alarm rate and improving detection accuracy.

[0105] Preferably, the data credibility factor is calculated using a weighted fusion algorithm:

[0106] Where T is the data reliability factor. For compliance scoring, For consistency scoring based on an anomaly detection engine, Rate the source;

[0107] The weighting coefficients are dynamically adjusted based on the input scenario. When an untrusted network is detected as the input environment, the weighting is increased. The weight W2.

[0108] Specifically, the calculation of the data credibility factor T uses a weighted fusion algorithm, and its specific implementation process is as follows (refer to...). Figure 2 .

[0109] First, the system obtains scores across three dimensions through parallel computation. Then, a rules engine is used to perform format and logic validation on the entered data, generating a compliance score. Its value depends on whether the data meets the preset requirements for integrity, type, and physical limits; spatiotemporal analysis is performed using the aforementioned multi-dimensional anomaly detection engine to generate a consistency score. This score reflects the degree to which the data conforms to historical geological patterns; simultaneously, based on the identity authentication and environmental awareness modules, it assesses the security of the input terminal, personnel qualifications, and network status, generating a source score. .

[0110] Secondly, the system dynamically adjusts the weighting coefficients based on the current data entry scenario. In a typical intranet security environment, the system assigns balanced initial weights. However, when the system's environment awareness module detects that the input environment is an untrusted network (such as public IP access, unknown devices, or high-risk areas), it triggers dynamic adjustment logic, significantly increasing the weight of the consistency score.

[0111] W2 (e.g., increased from 0.4 to 0.6), and the weights of other dimensions decreased accordingly.

[0112] Finally, the system substitutes the obtained scores and adjusted weights into the formula:

[0113] A weighted sum is performed to calculate the final data credibility factor T. This mechanism ensures that in untrusted environments where security is uncontrollable, the judgment logic will rely more heavily on the consistency characteristics of the data itself, thereby rigorously identifying potentially risky data.

[0114] Preferably, the method further includes a false alarm suppression step based on external context:

[0115] When a specific type of anomaly is detected, the system automatically queries the associated job logs or equipment status;

[0116] If the work log shows that there were work operations in the corresponding time period, or the equipment status shows a malfunction, then the anomaly is determined to be a non-geological data anomaly, and the anomaly alarm is automatically suppressed while maintaining the normal data reliability factor.

[0117] Specifically, the false alarm suppression step based on external context is implemented through the following process, referring to... Figure 3 .

[0118] First, during the operation of the multi-dimensional anomaly detection engine, once a specific type of abnormal signal such as a sudden drop in production or abnormal pressure fluctuation is detected, the system pauses the regular anomaly judgment process and automatically starts the false alarm suppression check module.

[0119] Secondly, the system utilizes data association technology to automatically initiate query requests to the operation management system and equipment IoT platform via the data bus based on the device ID, well location coordinates, and timestamp corresponding to abnormal data. The system retrieves and obtains relevant operation logs (such as well workover, fracturing, and well opening / closing operations) and equipment health status (such as sensor failures, network interruptions, and power outage records) within the specific time period in real time.

[0120] Subsequently, the system performs logical comparison and attribution analysis on the acquired external context information. If the operation log shows that there were planned operations affecting data fluctuations during that time period, or if the equipment status monitoring shows that there are hardware failures that cause data distortion, the system will clearly determine that the abnormal feature is not a true reflection of the geological reservoir conditions, but belongs to non-geological data anomalies (i.e., "pseudo-anomalies").

[0121] Finally, the system implements a suppression strategy, automatically overriding or ignoring anomaly alarms generated by the original anomaly detection engine. When calculating the data reliability factor, the system does not reduce the score of the data, maintaining its normal reliability factor and ensuring that the entered data can smoothly enter subsequent processes. This step effectively eliminates the interference of daily operations and equipment malfunctions on the geological data anomaly detection model, reducing the system's false alarm rate.

[0122] Preferably, the multi-dimensional dynamic permission allocation mechanism executes the following logic during decision-making:

[0123] If the data credibility factor is in the high credibility range and the user role permissions match, then plaintext or low-level de-identified data can be returned.

[0124] If the data confidence factor is in the gray range, the data is allowed to be returned, but the "data is disputed" confidence level prompt will be forcibly displayed on the query results page, and the weight of the data in the statistical aggregation will be limited.

[0125] If the data credibility factor is in the low credibility range, the highest level of de-identification strategy will be enforced, and the return of numerical plaintext will be prohibited.

[0126] Specifically, the multi-dimensional dynamic permission allocation mechanism executes the following logic during decision-making, referring to... Figure 5 :

[0127] The first step involves processing data in the high-confidence zone. The system first determines whether the data's confidence factor falls within the high-confidence range and verifies the query user's role and permissions. If the data has high confidence and the user's role has the corresponding permissions, the system determines it as compliant access, allowing the execution of low-level anonymization strategies (such as partial masking) or directly returning plaintext data, ensuring that the user can obtain high-precision information for business analysis.

[0128] The second step involves processing data in the grayscale region. If the data confidence factor falls within the grayscale range, the system adopts a conditional release strategy. Simultaneously with returning the data, the system forcibly loads and highlights a "Data is disputed" confidence level label in the query results interface. Furthermore, the system automatically modifies the weight parameters of this data in the background statistical aggregation task, significantly reducing its influence in total calculations or trend analysis, preventing potential anomalies from interfering with the overall assessment.

[0129] The third step involves processing data in the low-confidence zone. If the data confidence factor falls within the low-confidence range, the system triggers the highest level of security protection mechanisms. The system enforces the highest level of data anonymization, completely masking or generalizing numeric fields to an extremely wide fuzzy range, strictly prohibiting the return of any form of plaintext numeric data, ensuring that potentially low-quality or abnormal data will not mislead users.

[0130] Preferably, the independent message queue uses encrypted topics for data transmission, and only the writing process of the C-end data entry channel is allowed to write message bodies containing data digests to the encrypted topics, and only the consumption process of the B-end data query channel is allowed to read the message bodies, thereby realizing unidirectional data flow on the link.

[0131] Specifically, this step achieves unidirectional data flow by configuring an independent message queue and strict access control policies, as described in the following reference. Figure 1 .

[0132] First, the system creates a dedicated encrypted topic in the message queue cluster, enables Advanced Encryption Standard (AES) to encrypt the transmitted message body, and configures Transport Layer Security (TLS) protocol to ensure the confidentiality and integrity of data during transmission over the public network or across networks, preventing eavesdropping or man-in-the-middle attacks.

[0133] Secondly, strict write access control is implemented. The system only authorizes the write process (i.e., the producer) within the C-end data entry channel to access the write interface of the encrypted Topic. After completing data entry and blockchain uploading, the process automatically extracts a message body containing a data digest (such as data hash value, business primary key, and timestamp) and pushes it to the Topic. Access control lists (ACLs) strictly prohibit any other service, terminal, or process within the system from writing data to this Topic, preventing unauthorized injection.

[0134] Secondly, strict read access control is implemented. The system only authorizes the consumer process (i.e., the consumer) within the B-end data query channel to access the read interface of this encrypted Topic. This consumer process only has the permission to subscribe to and pull messages, used to synchronize data status in real time to update the B-end query cache or index. Similarly, through ACL policies, this consumer process is prohibited from performing any write, modification, or rollback operations on the Topic.

[0135] Finally, through the aforementioned unique "C-end write, B-end read" permission binding and encrypted channel mechanism, the system constructs an irreversible unidirectional data flow path in logic and links, ensuring that the data consumption behavior of the query end cannot reversely affect or pollute the original data of the input end, thus guaranteeing the uniqueness and purity of the data source.

[0136] Preferably, the method supports multi-tenant data isolation based on geological spatial topology:

[0137] Logical isolation boundaries are constructed based on the spatial topological relationships of geological structural units;

[0138] When a query request involves data across different geological tectonic units, the merged cross-domain data is only allowed to be returned if the user's role permissions include "whole basin level" authorization; otherwise, only data from a single tectonic unit is returned.

[0139] Specifically, the method supports the following implementation process for multi-tenant data isolation based on geospatial topology, as follows: (Refer to...) Figure 2 .

[0140] First, the system pre-constructs logical isolation boundaries based on the spatial topological relationships of geological structural units, maps each structural unit to an independent tenant space, and records its topological adjacency and containment relationships.

[0141] When a query request arrives from the B-end, the system parses the spatial scope of the request to determine whether it involves data spanning different geological structural units. If it involves only a single structural unit, the system directly performs the query and permission verification based on the tenant boundary corresponding to that unit. If a cross-domain query across structural units is detected, the system immediately verifies whether the user's role permissions include "whole basin level" authorization. Only when the user has this level of authorization is the fused cross-domain data allowed to be returned and aggregated for statistics; otherwise, the query scope is forcibly limited to a single structural unit, only data from that unit is returned, and cross-domain fusion is rejected in the response.

[0142] Furthermore, during query execution, the system automatically injects filtering conditions and permission policies based on topological boundaries to ensure that unauthorized users cannot bypass isolation through combined queries. This mechanism achieves multi-tenant data isolation and fine-grained authorization across the geological spatial topology dimension.

[0143] Preferably, the adaptive desensitization process specifically includes:

[0144] Based on the determined desensitization strategy level, a dynamic replacement algorithm is used for sensitive reserve values, or a random perturbation algorithm is used for geological coordinate data.

[0145] For data in the grayscale range, while preserving the original numerical visualization, the confidence interval range of the data distribution is overlaid to alert users to the potential uncertainty of the data.

[0146] Specifically, adaptive desensitization is implemented in the following manner, referring to... Figure 5 .

[0147] First, the system invokes the corresponding processing algorithm based on the determined desensitization strategy level. For sensitive reserve values ​​(such as proven geological reserves), the system employs a dynamic replacement algorithm. Based on preset fuzzy rules or data distribution characteristics, the original precise values ​​are replaced with surrogate values ​​or interval symbols that retain order-of-magnitude characteristics. For example, precise reserve values ​​are replaced with fuzzy interval values ​​or calculated random fuzzy values, thereby preventing the acquisition of real trade secrets through numerical inversion.

[0148] Secondly, for geological coordinate data (such as well location latitude and longitude), the system employs a random perturbation algorithm. Using the original coordinate point as the center, the perturbation radius is determined according to the strategy level. Within this radius, random offsets following a specific distribution (such as uniform or Gaussian distribution) are generated. The offset coordinate points are then returned as visualization data, achieving location privacy protection while preserving macroscopic geographical distribution characteristics.

[0149] Finally, for data in the grayscale range, the system adopts a "preserved value overlay interval" strategy. While retaining the original numerical visualization (such as line chart points or list values), the system calculates its statistical error range based on the data's confidence index and overlays this confidence interval range into the query results on the user interface (e.g., through error bars, background shading, or numerical range labels). This approach, while ensuring data readability, intuitively alerts users to the potential uncertainty of the data and the need for caution when using it.

[0150] Secondly, this invention provides an integrated platform for oilfield-level mineral rights and reserves, based on a permission separation system for data input by the C-end and data query by the B-end, comprising:

[0151] The C-end data entry channel is equipped with a multi-dimensional anomaly detection engine and a grayscale classification module, which are used to perform anomaly detection based on spatiotemporal context and calculate data credibility factors.

[0152] The B-end data query channel is configured with a multi-dimensional dynamic permission allocation mechanism and a data de-identification processor.

[0153] The expert review and feedback module is configured to display grayscale data and collect expert correction instructions, which are then converted into training samples and fed back to the multi-dimensional anomaly detection engine.

[0154] An isolated communication bus is used to connect the C-end and the B-end to implement one-way asynchronous communication and physical address hiding.

[0155] Specifically, this integrated oilfield-level mineral rights and reserves platform, based on a permission separation system for C-end data entry and B-end data querying, mainly consists of the following four collaborative parts, as referenced. Figure 1 .

[0156] The C-end data entry channel serves as the source entry point for data, internally integrating a multi-dimensional anomaly detection engine and a grayscale classification module. This channel is responsible for receiving raw data in real time, performing dynamic anomaly detection operations based on spatiotemporal context, calculating the data credibility factor based on the degree of deviation of the detection results, and automatically classifying the data into different states such as high credibility, grayscale, and anomaly using the grayscale classification module, thus completing the preliminary quality assessment and classification of the data.

[0157] The B2B data query channel, serving as the data service outlet, is equipped with a multi-dimensional dynamic permission allocation mechanism and a data masking processor. When responding to user query requests, the channel dynamically determines data access strategies based on the aforementioned data credibility factors and the user's role permissions. The data masking processor then performs tiered masking (such as masking or perturbation) on sensitive fields to ensure that data output complies with security standards and the principle of least privilege.

[0158] The expert review feedback module forms the system's self-learning closed loop and is configured specifically to receive and display data in the grayscale range. This module provides an interactive interface for review experts, collects their approval or rejection correction instructions for grayscale data, and transforms these instructions into structured training samples, which are then fed back to the multi-dimensional anomaly detection engine on the consumer side to guide the incremental training of the model and the dynamic adjustment of the detection threshold.

[0159] The isolated communication bus is the sole link connecting the C-end and the B-end, responsible for implementing unidirectional asynchronous communication and physical address hiding. Through independent message queues and other middleware technologies, this bus only allows the C-end to push data summaries or status updates unidirectionally to the B-end, physically isolating the B-end from directly accessing the C-end's underlying database. This achieves strict separation of permissions for data entry and querying, as well as data source protection at the architectural level.

[0160] In summary, by constructing an architecture that strictly separates C-end input from B-end querying, and combining it with multi-dimensional intelligent control strategies, this invention brings the following significant beneficial effects:

[0161] First, it significantly enhances data security and privacy protection capabilities. Through the unidirectional asynchronous communication of the isolated communication bus and the physical address hiding mechanism, the direct access path from the B-end to the C-end underlying data is completely cut off. Combined with the highest level of desensitization strategy and multi-tenant isolation based on geological spatial topology, it effectively prevents the leakage of sensitive reserve data and unauthorized access, ensuring the absolute security of core data assets.

[0162] Secondly, it achieves efficient self-evolution and accurate detection of data quality. Utilizing a spatiotemporal context-based anomaly detection engine, combined with an expert review and feedback loop, the system can continuously learn from expert approval and rejection instructions, dynamically optimizing detection thresholds and model parameters. This not only significantly reduces the false alarm rate caused by operational interference or equipment failure, but also enables the detection model to continuously adapt to complex geological environmental changes, ensuring the accuracy of the data stored.

[0163] Third, it balances data sharing efficiency with usage risks. By introducing a data credibility factor and a multi-dimensional dynamic permission allocation mechanism, this invention can flexibly adjust the anonymization strategy (such as dynamic replacement and random perturbation) and display method (such as confidence interval prompts) according to data quality and user roles. In particular, the special processing for grayscale data avoids the excessive discarding of valid data while fully highlighting potential uncertainties, thereby maximizing the business value and utilization of data while ensuring data security.

[0164] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for separating the rights of C-end input data and B-end query data based on an oilfield-level mineral right reserve integration platform, characterized in that, Includes the following steps: An isolation architecture is constructed between the C-end data entry channel and the B-end data query channel. The isolation architecture uses an independent message queue for one-way asynchronous communication and is configured with access control policies to prevent the B-end from obtaining the physical storage address of the data. In the C-end data input channel, a multi-dimensional anomaly detection engine is used to detect the input data. The detection includes dynamic pattern detection based on spatiotemporal context. The data credibility factor is calculated based on the detection results, and the data is divided into a safe zone, a blocking zone, or a gray zone according to the anomaly score. For data in the gray area, it is allowed to be entered into the database with a lower initial data credibility factor and labeled "suspected anomaly - pending verification" to generate a blockchain traceability label containing the data credibility factor and label; In the B-end data query channel, query requests are received, and the blockchain traceability tag is read through a multi-dimensional dynamic permission allocation mechanism; real-time decisions are made based on the data credibility factor and user attributes to determine access permissions and desensitization strategy levels. Perform adaptive data masking; if the queried object has a "suspected anomaly" label, include a confidence level indicator in the returned data. Constructing an expert review and feedback closed loop: Based on the experts' approval or rejection instructions for the grayscale data, training samples are generated, and the multi-dimensional anomaly detection engine is incrementally trained to adjust the detection threshold.

2. The method of claim 1, wherein, The spatiotemporal context-based dynamic pattern detection includes: Spatial outlier detection: Extract the geographical location of the entered data, use the K-Nearest Neighbor (KNN) algorithm to retrieve historical neighboring well data within a preset radius, and calculate the characteristic deviation between the current data and the neighboring well data group; Time series anomaly detection: Construct a Long Short-Term Memory (LSTM) network prediction model for the monitored data and calculate the reconstruction error between the actual entered values ​​and the predicted values; If the feature deviation or reconstruction error exceeds a preset threshold, it is determined to be abnormal, and a corresponding abnormal score is generated.

3. The method of claim 1, wherein, The execution steps of the expert review and feedback closed loop include: When the data is in the grayscale area, an expert review task is generated and pushed to the expert terminal; Receive confirmation instructions from experts regarding the review task. If the instruction is "approved", the original detection result is determined to be a false alarm, the corresponding data is marked as "true value", and it is added to the training set as a negative sample. If the instruction is "reject", the original detection result is determined to be a true anomaly, and the corresponding data and its features are added to the training set as positive samples; The training set is used to fine-tune or incrementally learn the model parameters of the multi-dimensional anomaly detection engine.

4. The method of claim 1, wherein, The data credibility factor is calculated using a weighted fusion algorithm: ; wherein, T is a data credibility factor, is a compliance score, is a consistency score based on the anomaly detection engine, is a source score; the weight coefficient is dynamically adjusted according to the entry scene, and when it is detected that the entry environment is an untrusted network, the weight W2 of is increased.

5. The method of claim 1, wherein, The method further includes a false alarm suppression step based on external context: When a specific type of anomaly is detected, the system automatically queries the associated job logs or equipment status; If the work log shows that there are work operations in the corresponding time period, or the equipment status shows a malfunction, then the abnormality is determined to be a non-geological data abnormality, and the abnormal alarm is automatically suppressed while maintaining the normal data reliability factor.

6. The method of claim 1, wherein, The multi-dimensional dynamic permission allocation mechanism executes the following logic during decision-making: If the data credibility factor is in the high credibility range and the user role permissions match, then plaintext or low-level de-identified data can be returned. If the data confidence factor is in the grayscale range, the data is allowed to be returned, but the "data is disputed" confidence level prompt will be forcibly displayed on the query results page, and the weight of the data in the statistical aggregation will be limited. If the data credibility factor is in the low credibility range, the highest level of de-identification strategy will be enforced, and the return of numerical plaintext will be prohibited.

7. The method of claim 1, wherein, The independent message queue uses encrypted topics for data transmission, and only the writing process of the C-end data entry channel is allowed to write message bodies containing data digests to the encrypted topics, and only the consumption process of the B-end data query channel is allowed to read the message bodies, thereby realizing unidirectional data flow on the link.

8. The method of claim 1, wherein, The method supports multi-tenant data isolation based on geospatial topology: Logical isolation boundaries are constructed based on the spatial topological relationships of geological structural units; When a query request involves data across different geological tectonic units, the merged cross-domain data is only allowed to be returned if the user's role permissions include "whole basin level" authorization; otherwise, only data from a single tectonic unit is returned.

9. The method of claim 6, wherein, The adaptive desensitization process specifically includes: Based on the determined desensitization strategy level, a dynamic replacement algorithm is used for sensitive reserve values, or a random perturbation algorithm is used for geological coordinate data. For data in the grayscale range, while preserving the original numerical visualization, the confidence interval range of the data distribution is overlaid to alert users to the potential uncertainty of the data.

10. An oilfield-level mineral right reserve integration platform based on C-end entry data and B-end query data permission separation system, characterized in that, include: The C-end data entry channel is equipped with a multi-dimensional anomaly detection engine and a grayscale classification module, which are used to perform anomaly detection based on spatiotemporal context and calculate data credibility factors. The B-end data query channel is configured with a multi-dimensional dynamic permission allocation mechanism and a data de-identification processor. The expert review and feedback module is configured to display grayscale data and collect expert correction instructions, and convert the correction instructions into training samples and feed them back to the multi-dimensional anomaly detection engine. An isolated communication bus is used to connect the C-end and the B-end to implement one-way asynchronous communication and physical address hiding.