Method and system for multi-dimensional data view generation with dynamic policy support and security access control

Abstract

The application relates to the technical field of electric digital data processing and access control, and discloses a multi-dimensional data view generation and security access control method and system supporting dynamic strategies, which comprises the following steps: obtaining an access request containing a business role and an environment parameter, generating an access permission mask bitmap indicating field-level access permission through strategy judgment logic; pushing the access permission mask bitmap to a storage engine index layer, locking a physical address meeting the permission through bit logic operation, and generating an address mapping table; identifying a restricted field when encapsulating a view descriptor, and redirecting a read pointer pointing to an original address of the restricted field to a memory constant isolation area; and constructing a virtual addressing table according to the redirected read pointer to generate a virtual view. The application blocks the link of an unauthorized data block to the memory in the storage index space, and eliminates the risk of sensitive data residing in the memory.

Description

Technical Field

[0001] This invention belongs to the field of electronic digital data processing and access control technology, and particularly relates to a method and system for generating multidimensional data views and providing secure access control that supports dynamic policies. Background Technology

[0002] Currently, digital data processing and secure access control technologies form the underlying foundation supporting multi-service collaborative systems. The industry generally adopts role-based or attribute-based access control mechanisms, generating corresponding business data views by setting filtering rules or desensitization algorithms at the application layer. This conventional approach follows a post-extraction filtering process, where the storage engine loads the original data blocks from persistent storage media into the volatile memory of the computing node. The processing unit traverses the memory space and performs field-level replacement or masking according to logical instructions. However, in flexible workforce scenarios with high-concurrency retrieval and multi-dimensional permission changes, the above design inertia leads to a constraint on the balance between performance and security. Before completing logical filtering, the original sensitive data resides in plaintext in the memory stack. Within this physical window, the system faces the risk of leakage caused by memory dumps or side-channel attacks. At the same time, the computing power overhead of view generation is linearly related to the size of the original data hit by the query, and the repeated loading of massive amounts of data and scalar processing logic produce significant response lag.

[0003] To address these challenges, improvements such as increasing hardware redundancy or pre-generating static de-identified datasets still have limitations. Increasing hardware computing power cannot fundamentally eliminate the risk of plaintext exposure at the physical addressing level. Pre-generated static views, due to their lack of dynamic adaptation capabilities, cannot meet differentiated compliance viewing needs. Besides limitations in hardware-level solutions, existing software-based security control methods also have shortcomings. For example, Chinese invention patent application CN114826759A discloses a verifiable fine-grained access control inner product function encryption method, which improves access control granularity by introducing a hierarchical access structure and an inner product function encryption model. The underlying mechanism still heavily relies on extracting the entire ciphertext to the computing node and then performing decryption and authorization verification through complex matrix algebra operations. When faced with extreme objective conditions such as flexible employment, multiple roles, and massive concurrent queries, the computational overhead of relying on post-hoc cryptographic operations to control the flow inevitably expands linearly with the data throughput. Furthermore, it does not prevent the loading of non-compliant data blocks into memory from the physical addressing source. Core data must flow and queue in the computing memory. The underlying mechanism cannot overcome the physical limits of extremely low response latency and zero exposure window in concurrent scenarios. Analysis shows that the bottleneck of the existing technology lies in the disconnect between the access control logic and the underlying data flow path, resulting in an inability to achieve a coordinated balance between security and processing performance.

[0004] Therefore, the technical problem to be solved by this invention is how to construct a zero-copy view projection mechanism based on addressing intervention and index masking to block the memory loading of non-compliant data before the physical read operation occurs and to achieve physical redirection and desensitization of sensitive fields. Summary of the Invention

[0005] This invention provides a method for generating multidimensional data views and implementing secure access control with dynamic policies, comprising the following steps: Step S1: Obtain an access request containing business role information and real-time environment parameters. Map the business role information and real-time environment parameters to a multi-dimensional permission mapping matrix through a preset policy judgment logic. Obtain a field access weight sequence corresponding to the current access context through the inner product operation of the multi-dimensional permission mapping matrix and the preset business policy vector. Calculate and generate an access permission mask bitmap indicating field-level access permissions based on the comparison results of the values ​​of each element in the field access weight sequence with the preset permission threshold. Step S2: Push the access permission mask bitmap down to the storage engine index layer, and perform a bitwise logical AND operation between the access permission mask bitmap and the preset data tag index bitmap to determine the set of physical storage addresses that conform to the permission policy before the data block is loaded into memory, so as to generate a physical address mapping table. Step S3: During the encapsulation of the multidimensional data view descriptor, the physical address mapping table is parsed to identify the field attributes of the restricted fields. By modifying the address pointer mapping relationship, the read pointer in the multidimensional data view descriptor that originally pointed to the original physical storage address of the restricted field is forcibly redirected to a pre-allocated constant isolation area in the kernel space memory. The constant isolation area stores mask characters that represent the desensitization status. Step S4: Construct a virtual addressing table based on the physical address mapping table and the redirected read pointer. Use the virtual addressing table to map application layer requests to the corresponding physical storage space or constant isolation area. While maintaining the logical structure of the multidimensional data view and without extracting the full amount of original plaintext data, generate a multidimensional virtual data view so that sensitive fields in the multidimensional virtual data view are redirected from the read pointer to the constant isolation area.

[0006] Preferably, step S1 includes: collecting dynamic environmental factors including visitor's geographical location, device fingerprint, historical behavior characteristics, and network access environment; normalizing the dynamic environmental factors and mapping them together with business role information into the feature space of the multi-dimensional permission mapping matrix; and extracting the field access weight sequence that indicates the legality of the current access behavior by calculating the correlation between the feature space and the preset business policy vector.

[0007] Preferably, in step S2, a hard filtering barrier is established in the index space of the storage engine using an access permission mask bitmap. By blocking the loading link of unauthorized data blocks to computing memory before physical I / O is triggered, illegal field data cannot generate a physical resident image in the system memory address space, thereby eliminating the exposure window of sensitive plaintext information in the memory flow cycle from the physical source and enhancing data sovereignty security in a multi-tenant environment.

[0008] Preferably, in step S3, by identifying the field identifier of the restricted field in the multidimensional data view descriptor, the entry parameter in the physical address index table inside the multidimensional data view descriptor is modified using the address offset algorithm, and the physical base address that originally pointed to the original disk image area is offset to the logical starting address of the constant isolation area, so as to ensure that the value retrieval instruction initiated by the application layer is hard redirected to the preset static storage area at the micro-instruction execution level.

[0009] Preferably, the constant isolation zone is set as a read-only memory page in kernel mode, which stores a placeholder sequence composed of de-identified mask characters. The application layer is limited by the physical routing path of the read pointer and can only access the controlled physical page address through the virtual address table to read the de-identified mask characters in the constant isolation zone. This achieves data topology isolation and compliant viewing at the computing memory level without changing the upper-layer application logic.

[0010] Preferably, in step S4, the virtual addressing table is constructed as a mapping matrix between logical addresses and physical addresses. The mapping matrix includes: a regular field pointer pointing to a specific offset address of the original data storage area, and a restricted field pointer generated in step S3 pointing to a specific address space of the constant isolation area. The regular field pointer is used to pass through non-sensitive data, and the restricted field pointer is used to intercept and block the addressing path of sensitive original data.

[0011] Preferably, while keeping the same underlying original physical data in silent storage, the system supports dynamically generating multiple multi-dimensional virtual data views with different desensitization granularities and field visibility in the virtual mapping layer by switching or updating the access permission mask bitmap in real time; this process does not involve any data copying or moving operations in the physical storage layer, and achieves near real-time adaptive adjustment of data access permissions by changing the mapping topology of the logical addressing table.

[0012] Preferably, in step S2, the physical address is filtered using the following bitwise logical operations: ,in, For access control mask bitmap, This is the tag bitmap index for the corresponding data shard in the storage engine. A candidate bitmap used to identify the set of loadable physical addresses that conform to the current policy; when When a specific bit in the memory is 1, the corresponding physical storage unit is determined to be within the legal loading range of the current access request.

[0013] Preferably, the construction of the multidimensional virtual data view is entirely converged on the underlying index bit operations and address pointer remapping. The computational overhead of the system when handling massive concurrent query requests does not increase linearly with the increase of the number of physical data rows. The response latency of its view generation is limited by the index retrieval depth and the complexity of address table construction, thus providing sublinear growth in system throughput performance in high-dimensional differentiated access scenarios. The policy determination logic calculates differentiated field permissions based on the attributes of different business roles within the flexible employment ecosystem. Business roles include the enterprise as the employing entity, the service platform as the collaborating entity, and the administrative functional departments as the regulatory entity. By matching different preset business policy vectors to different business roles, differentiated views that meet compliance requirements can be generated for each participant within the same original data pool.

[0014] A multi-dimensional data view generation and secure access control system supporting dynamic policies includes: The permission calculation module is used to obtain access requests containing business role information and real-time environment parameters, map the business role information and real-time environment parameters to a multi-dimensional permission mapping matrix, and calculate and generate an access permission mask bitmap indicating field-level access permissions through the inner product operation of the multi-dimensional permission mapping matrix and the preset business policy vector. The storage index filtering module is used to push the access permission mask bitmap down to the storage engine index layer, so that the access permission mask bitmap and the data tag index bitmap preset in the underlying storage medium perform bit logic orthogonal operation, and determine the set of physical storage addresses that conform to the permission policy before the data block is loaded into the computing memory based on the bit operation result to generate a physical address mapping table. The addressing intervention module is used to parse the physical address mapping table and identify the field attributes of restricted fields during the encapsulation of multidimensional data view descriptors. By modifying the addressing pointer mapping relationship, it forces the read pointer in the multidimensional data view descriptor that points to the original physical storage address of the restricted field to a constant isolation area in kernel-mode memory. The constant isolation area stores desensitized mask characters. The virtual addressing image module is used to construct a virtual addressing table based on the physical addressing table and the redirected read pointer. The virtual addressing table is used to generate a multidimensional virtual data view while maintaining the logical structure of the multidimensional data view and without extracting all the original plaintext data.

[0015] Compared to existing technologies, this invention supports a dynamic policy-based multidimensional data view generation and secure access control method, which has the following advantages: 1. In the generation of multidimensional data views, by compiling the refined permission decision results into dynamic permission bitmap vectors and pushing them down to the storage engine index layer to participate in bitwise logical AND operations with the pre-set data tag bitmap index, the system pre-locks the set of physical addresses that conform to the permission policy and forms a view mapping base table before the physical read operation occurs. Since this mechanism directly blocks the loading link of unauthorized data blocks to computing memory in the index space, it removes the risk of illegal residence of sensitive information at the physical source of the original data flowing from the storage medium to memory, avoids the memory plaintext exposure window caused by the first extraction and then processing method in the existing technology, and improves the data sovereignty security strength of the system in complex multi-tenant environments.

[0016] 2. For sensitive field spaces that need to be de-identified or hidden, the system utilizes a dynamic memory addressing intervention mechanism. When assembling multidimensional data view descriptors, the view data structure shell remains unchanged, but the read pointers pointing to the physical addresses of the underlying sensitive fields are forcibly redirected to a pre-allocated independent constant isolation area in kernel mode. This mechanism truncates the scalar substitution action in the traditional data processing flow, making the upper-layer application limited by the hard physical routing of the redirected pointers and only able to read the pre-set de-identification symbols in the isolation area. This achieves topological isolation at the computation memory level and solves the technical problem that traditional logical de-identification methods cannot defend against memory dump attacks. Attached Figure Description

[0017] Figure 1 This is a flowchart of the data view generation steps for the dynamic strategy and permission mask of this invention; Figure 2 This is the pointer redirection topology diagram of the virtual addressing mapping and constant isolation area of ​​the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0019] It should be noted that all directional and positional terms used in this invention, such as: up, down, left, right, front, back, vertical, horizontal, inner, outer, top, bottom, transverse, longitudinal, center, etc., are only used to explain the relative positional relationship and connection between components in a specific state (as shown in the accompanying drawings). They are only for the convenience of describing this invention and do not require that this invention be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention. In addition, the descriptions of "first," "second," etc., in this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated.

[0020] In the description of this invention, unless otherwise explicitly specified and limited, the terms installation, connection, and linking should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections; they can refer to direct connections or indirect connections through an intermediate medium; they can refer to the internal connection of two components. For those skilled in the art, the specific meaning of the above terms in this invention can be understood according to the specific circumstances.

[0021] In the description of this specification, references to the terms "an embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example, and the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0022] Example 1: In a flexible employment multi-role collaborative access scenario with high-concurrency retrieval and multi-dimensional permission changes, the enterprise (employer), the service platform (collaborating entity), and the administrative department need to cross-domain retrieve the same physical mapping record from the underlying core data pool. Traditional extraction and filtering methods require the storage engine to load the entire original data block containing sensitive information into the volatile memory of the computing node, and the computing unit to match filtering rules through traversal. This data flow path causes the original plaintext information to reside in the memory stack, creating a physical exposure window. At the same time, the repeated memory loading of massive amounts of data causes the computing power overhead to expand linearly with the query scale, resulting in a structural conflict between data defense requirements and high-throughput business response requirements. The system obtains access requests containing business role information and real-time environment parameters, maps the business role information and real-time environment parameters to a multi-dimensional permission mapping matrix, and obtains a field access weight sequence through the inner product operation of the multi-dimensional permission mapping matrix and the preset business strategy vector. Based on the comparison results of each element value in the field access weight sequence with the preset permission threshold, an access permission mask bitmap indicating field-level access permissions is calculated and generated.

[0023] For the multi-dimensional mapping of business role information and real-time environment parameters, the system extracts the business role identifier carried in the access request and uses a one-hot encoding algorithm to convert the business role identifier into a multi-dimensional orthogonal discrete feature vector. Simultaneously, it collects the terminal location coordinates and device trust scores from the real-time environment parameters. In the coordinate dimensionality reduction stage, the system extracts the latitude and longitude of the enterprise core data center gateway stored in the security configuration library as the baseline trust physical origin. It calculates the absolute spherical distance between the current terminal location coordinates and this baseline trust physical origin, measured in kilometers. A maximum trust mapping radius of 50 kilometers is set as the decision boundary for hardware execution. When the calculated absolute spherical distance is equal to or greater than 50 kilometers, the distance deviation factor is directly locked to an upper limit scalar of 1 at the logical decision layer. When the distance is less than 50 kilometers, the processor divides the absolute spherical distance value by a constant 50. The distance deviation factor is obtained by taking a fractional proportion between 0 and 1 from the distance in kilometers. This deterministic unidirectional distance deviation factor replaces the multidimensional coordinate system. Based on this result, a linear normalization function is called to quantize the terminal positioning coordinates and the device trust score to a continuous numerical range of zero to one. The system sequentially concatenates the discrete feature vector with the quantized continuous values ​​to generate a comprehensive feature array of the current access link and writes it into the address space of the column vector of the multidimensional permission mapping matrix. The hardware processor retrieves the multidimensional permission mapping matrix and the preset business policy vector and executes a register matrix multiplication instruction. For the numerical tracing and calibration of the preset business policy vector, the system extracts the legal release frequency of each field in the first 1000 historical audit logs during the initial power-on self-test phase, divides it by the total number of requests to obtain a specific fractional proportion between 0 and 1 as the fixed basic weight of the field and writes it into the vector register.

[0024] In the microinstruction loop that performs matrix multiplication and weight allocation, the processor sequentially extracts each horizontal row vector representing a single-dimensional user attribute factor from the multidimensional permission mapping matrix. It then performs a bitwise-aligned dot product operation with the corresponding field weights in the preset business strategy vector. An accumulation instruction is then executed to sum all product results in the same row, thus physically aggregating discrete environmental risk factors into an overall addressing access probability for a single storage field. This outputs a field access weight sequence composed of the inner product calculations of the corresponding underlying data fields. The system then displays the access permission mask bitmap. Push down to the storage engine index layer to make the access permission mask bitmap Compared with the data tag index bitmap pre-built in the underlying storage medium Perform bitwise AND operation according to the formula Filter physical addresses, among which, This represents the candidate bitmap. This represents a bitmap containing access permission masks. This represents a data label index bitmap, where the data format of each variable is a binary sequence; when the candidate bitmap When a specific bit in the memory is set to 1, the corresponding physical storage unit is determined to be within the authorized loading range of the current access request. Based on this, the system determines the set of physical storage addresses that conform to the permission policy before the data block is loaded into memory, thereby generating a physical address mapping table. This mechanism cuts off the loading link of unauthorized data blocks to computing memory within the index space, so that the physical address filtering process and the subsequent view encapsulation action form a cascade effect, and the front-end filtering barrier provides an addressing baseline without unauthorized redundancy for the back-end pointer-level redirection, thus resolving the hardware layer conflict between high-concurrency access verification and memory capacity exhaustion.

[0025] During the encapsulation of the multidimensional data view descriptor, the system parses the physical address mapping table to identify the field attributes of restricted fields, replacing the scalar text masking and replacement logic in the application-layer memory data desensitization mechanism, and adopts a low-level spatial addressing topology isolation scheme. The system uses an address offset algorithm to modify the entry parameters in the physical address index table inside the multidimensional data view descriptor, offsetting the physical base address pointing to the original disk image area to the logical starting address of a pre-allocated constant isolation area in kernel-mode memory. The read pointer in the multidimensional data view descriptor pointing to the original physical storage address of the restricted field is redirected to this constant isolation area. This constant isolation area is set as a read-only memory page in kernel-mode, and it stores desensitization mask characters representing the desensitization status. The application layer reads the desensitization mask characters in the constant isolation area following the physical routing path of the read pointer, transforming the multidimensional dynamic policy management requirements into a pointer address routing change task.

[0026] During the forced redirection process of the multidimensional data view descriptor read pointer, the system calls kernel-level allocation instructions during the underlying operating system initialization phase to request contiguous physical memory pages as constant isolation areas. It writes a specified byte length of de-identified mask characters to the constant isolation areas and modifies the corresponding page table entry control bits to lock the physical memory pages as read-only. The system triggers a shared memory mapping instruction to cross-map the physical base address of the constant isolation areas to the user-mode virtual address space of the current application-layer database processing process, generating a user-mode image entry address with read-only permissions. The addressing intervention module extracts the user-mode image entry address, calculates the hexadecimal offset vector between the address and the original physical storage address of the restricted field, and adds it to the recorded value of the restricted field read pointer addressing register, thus changing the final physical route endpoint of the read pointer directly from the original disk image area to the restricted field. The constant isolation area is constructed by the system based on the physical address mapping table and the redirected read pointers, which includes a virtual addressing table containing the mapping relationship between logical addresses and physical addresses. The virtual addressing table is used to map application layer requests to the corresponding physical storage space or constant isolation area. This generates a multidimensional virtual data view while maintaining the logical structure of the multidimensional data view and eliminating the need to extract all original plaintext data. Sensitive fields in this multidimensional virtual data view are directed to the constant isolation area by the read pointers, while non-sensitive fields are passed through non-sensitive data via regular field pointers pointing to specific offset addresses in the original data storage area. This reconstruction of the addressing topology makes the computational overhead of view generation independent of the increase in the number of physical data rows. The system achieves a structural combination of access control logic and underlying storage hardware processing path without requiring physical storage layer data copying and movement operations.

[0027] Example 2: In a high-concurrency data access test scenario, a data server with a 128-core processor and 2TB of physical memory was used as the test platform. The data server internally stores a 10TB structured relational personnel record set. The test request generation tool was configured to generate a sequence of concurrent access requests, with the concurrency of the sequence increasing in a gradient from 10,000 to 100,000. The sequence included a 20% cross-domain privilege escalation probe data stream, which served as background access noise. When the concurrent request increment step was set, the system obtained the maximum throughput rate and transmission response latency parameters of the test platform's network card. When the test platform's network bandwidth reached 80Gbps and the transmission response latency was less than 1ms, it was determined that the impact of the underlying protocol stack queuing latency on the application layer time measurement was less than 0.1ms. Based on this measurement result, the system set the test script to generate a test data stream with a concurrency increment step of 10,000, eliminating the measurement error caused by the underlying protocol stack queuing latency.

[0028] A control group employing a full-load memory filtering mechanism and an experimental group employing a multi-dimensional permission mapping and pointer redirection mechanism were set up. The experimental group parsed the role information and environment parameters in concurrent access requests and calculated the access permission mask bitmap. The system will use the access permission mask bitmap. With data label index bitmap Perform bitwise AND operation according to the formula Obtain candidate bitmap Key intermediate state data were collected under different concurrency request gradients. Under the condition of 30,000 concurrent requests, the control group loaded 27.4GB of raw sensitive data into volatile memory, with a plaintext data exposure time window of 82.3ms. The experimental group's bit logic operation time was 0.7ms, the amount of raw sensitive data loaded into memory was 0GB, and the plaintext exposure time window was 0ms. Under the condition of 50,000 concurrent requests, the volume of sensitive data loaded in the control group increased to 45.6GB, and the plaintext exposure time window reached 125.6ms. The experimental group's bit logic operation time remained at 0.8ms, and the amount of memory loaded and the plaintext exposure time window remained constant at 0.

[0029] For the data view encapsulation process containing restricted fields, the control group performed scalar text masking and replacement calculations; the experimental group modified the physical address index of the multidimensional data view descriptor based on the physical address mapping table, redirecting the read pointer pointing to the original physical storage address of the restricted field to a constant isolation area storing the masked characters. The processing time was recorded as the number of rows hit by the requested query increased from 10,000 to 100,000 rows. When the number of hit rows was 10,000, the masking calculation time for the control group was 8.5ms, while the pointer redirection time for the experimental group was 1.1ms. When the number of hit rows increased to 50,000, the masking calculation time for the control group increased non-linearly to 42.3ms, while the pointer redirection time for the experimental group was 1.2ms. When the number of hit rows reached 80,000 and the concurrent request volume reached 8... Under extreme operating conditions of 0000, the control group experienced memory exhaustion and triggered abnormal process termination, resulting in loss of service responsiveness. The experimental group, relying on the single-instance de-identification mask characters of the constant isolation zone and the underlying index truncation mechanism, maintained an overall query latency between 5.1ms and 6.2ms, and the pointer redirection time was stable at 1.3ms. The multi-dimensional permission mapping matrix combined with the read pointer redirection mechanism, under high-concurrency conditions including background access noise, truncated the physical loading link of sensitive data to computing memory through bitwise logical AND operations, compressing the physical window of plaintext exposure in memory to 0ms. The addressing topology isolation of the constant isolation zone made the computing power and memory overhead of the view generation stage independent of the number of original data rows hit by the query. The system maintained consistent data processing latency and fine-grained access control state in a high-throughput retrieval environment.

[0030] Example 3: In high-concurrency cross-domain data access scenarios, the security risk level of the network environment fluctuates continuously. Fixed, constant access permission judgment benchmarks cannot adapt to sudden risk events, leading to unauthorized access and leakage under high-risk conditions or overload interception under low-risk conditions. The system extracts network node credit values ​​from network node monitoring logs, collects the security module authentication status of terminal devices, and quantifies and outputs device trust scores. Using device trust scores and network node credit values ​​as real-time environmental parameters, the system calculates a weighted sum of device trust scores and network node credit values ​​to generate an environmental risk scalar for the current access link. The system reads preset benchmark permission thresholds and risk penalty coefficients, and applies the formula... Calculate and generate dynamic preset permission thresholds; where, This indicates the preset permission threshold. Indicates the baseline permission threshold. Indicates environmental risk scalars, The system represents the risk penalty coefficient, and all variables are dimensionless real numbers. The system reads the inner product operation result of the multidimensional permission mapping matrix and the preset business strategy vector to generate a field access weight sequence. The system compares the value of each element in the field access weight sequence with the preset permission threshold. When the weight value of a specific element is greater than or equal to the preset permission threshold, the system sets the bit of the corresponding field in the access permission mask bitmap to the binary value 1 of the authorization identifier. When the weight value of a specific element is less than the preset permission threshold, the system sets the bit of the corresponding field in the access permission mask bitmap to the binary value 0 of the interception identifier. This quantitative comparison mechanism dynamically adjusts the bit assignment criteria according to the environmental risk scalar, so that the access permission mask bitmap has a security interception boundary that adapts to risk fluctuations.

[0031] For restricted fields with bit values ​​assigned to the intercepted state, the system initiates an addressing offset algorithm to assemble a multidimensional data view descriptor, extracts the original physical base address of the restricted field in the underlying data block, queries the system kernel-mode memory page table to obtain the logical start address of the pre-allocated constant isolation zone, calculates the hexadecimal address difference vector between the logical start address of the constant isolation zone and the original physical base address of the restricted field, extracts the read pointer pointing to the original physical storage address of the restricted field inside the multidimensional data view descriptor, and superimposes the hexadecimal address difference vector onto the address register record value of the read pointer to generate a redirected write instruction to overwrite the read instruction. The original address parameters of the pointer are modified so that the read pointer directly points to the address space of the constant isolation area. The overwritten read pointer is used to construct a virtual addressing table containing the mapping relationship between logical addresses and physical addresses. When the application layer initiates a data read request, the memory management unit parses the overwritten read pointer in the virtual addressing table, physically routes the read link to the constant isolation area, and outputs the desensitized mask characters pre-stored in the constant isolation area. The combination of dynamic threshold adjustment mechanism and address offset algorithm truncates the computation memory loading path of the original plaintext data. The system maintains the data sovereignty isolation state and fine-grained access control benchmark in the retrieval environment with fluctuating security risks.

[0032] Example 4: In the pre-deployment calibration scenario of a flexible multi-role collaborative system, the system establishes an initial baseline of the baseline permission threshold before accessing the business data stream. The test node imports a de-identified test set containing 10,000 overreach detection samples and 50,000 regular business samples. The system counts the total number of samples triggering overreach interception and the theoretical total number of samples expected to be intercepted in the de-identified test set, calculates the ratio of the total number of samples triggering overreach interception to the theoretical total number of samples, and obtains the initial interception rate. The system extracts preset security level indicators. In this mapping execution step, the system's underlying layer has a physical hardware lookup table consisting of three discrete key-value pairs. When the input security level indicator corresponds to Level 1 regular business protection, it outputs a value of 1.0 to the executor as the matching system sensitivity coefficient value; when it corresponds to Level 2 highly sensitive business data protection, it looks up the table and outputs a value of 1.5; when it corresponds to Level 3 highest system control permission protection, it looks up the table and outputs a value of 2.0. This physical process of step-by-step table lookup mapping maps the security level indicator to the corresponding system sensitivity coefficient value, according to the formula... Calculate the baseline permission threshold; where, λ represents the baseline permission threshold, and λ represents the dimensionless system sensitivity coefficient. This represents the dimensionless initial interception rate, which the system uses to write the baseline permission threshold into the underlying physical storage medium.

[0033] When faced with the initial data filling scenario of a multidimensional permission mapping matrix, the system extracts the compliant access record feature vectors of specific business roles, calculates the coordinates of the cluster center points of the compliant access record feature vectors in the multidimensional vector space, and writes them into the initial array of the multidimensional permission mapping matrix as the mapping benchmark value. When faced with the specific business scenario of a new business role being accessed, the system collects the first batch of 500 access behavior feature vectors of the new role, calculates the Euclidean distance between the first batch of access behavior feature vectors and the coordinates of each cluster center point, and corrects the local weight elements in the multidimensional permission mapping matrix based on the obtained shortest Euclidean distance. The initial setting operation of data filling and on-site calibration is to build the underlying physical association benchmark for the field access weight sequence generation link.

[0034] Example 5: In the offline calibration scenario of the data scheduling system, before accessing service traffic, the system establishes a quantitative benchmark of preset service policy vector and risk penalty coefficient, obtains a de-identified sample set containing regular read records and unauthorized access records, extracts the security module response latency of terminal devices and network node packet loss rate, reads a preset 1ms constant from non-volatile memory as a benchmark comparison latency, divides the extracted security module response latency by the 1ms constant, so that the physical unit of time is canceled out by the underlying divider and forcibly converted into a dimensionless time ratio scalar; at the same time, a constant offset of 0.01 is forcibly added to the obtained network node packet loss rate value to absolutely prevent the system crash singularity with a denominator of 0 in a completely smooth network state; executes a register multiplication instruction to multiply the dimensionless time ratio scalar by the packet loss rate after adding the offset, and defines the reciprocal of the product of security module response latency and packet loss rate as the sample environment risk feature value, extracting the sample environment risk feature corresponding to the unauthorized access record. The system identifies the boundary extreme values ​​of the eigenvalues ​​and sets these boundary extreme values ​​as the risk penalty coefficient. It then calculates the access frequency of each business role to the underlying data fields, converts the access frequency into a normalized frequency scalar, and uses this normalized frequency scalar to assemble a preset business strategy vector. When the system faces a multi-tenant collaborative testing scenario, the test probe inputs a verification request stream containing simulated environment risk characteristic values. The system calls the preset business strategy vector and the multi-dimensional permission mapping matrix to generate an inner product operation result, outputting a field access weight sequence. Based on the simulated environment risk characteristic values ​​and the risk penalty coefficient, it calculates the preset permission threshold and generates an access permission mask bitmap by combining the field access weight sequence. The system calculates the absolute difference between the actual number of blocked physical addresses and the theoretical number of blocked addresses. When the absolute difference is greater than zero, the risk penalty coefficient is decreased by a set step size, and the verification request stream is resent until the absolute difference equals zero. The risk penalty coefficient is then written to the underlying physical storage medium, completing the parameter calibration process before generating the multi-dimensional data view.

[0035] In a multi-tenant collaborative data scheduling environment, the underlying core data pool faces implicit negative event recording injection detection. The attack source constructs structured query requests that conform to syntax specifications and frequently cross field physical addresses to infer the storage topology of sensitive fields. The system collects the timestamp sequence of continuous access requests and the hexadecimal span sequence of the requested physical addresses. To ensure the continuity of underlying digital signal processing and prevent feature loss, the system allocates a fixed-capacity circular memory buffer in the kernel space and explicitly sets the absolute duration of a single data sampling window to 500ms, continuously capturing underlying access request commands at a fixed sampling frequency of 100Hz. To avoid boundary feature truncation, system instructions control a 50% hard overlap rate between two adjacent sampling windows. For each sampling time window containing 50 valid sampled data points, the first-order difference vector of the timestamp sequence is extracted. In addition to the discrete scalar of the hexadecimal span sequence, the system predetermines that the behavior trajectory matrix has a 10x10 two-dimensional memory arrangement structure. First-order difference vectors of length 90 floating-point numbers are sequentially filled into the first 9 rows of the matrix in row-major order. Discrete scalars containing 10 sampling periods are extracted and merged into a 1-dimensional array of length 10, which is then written into the 10th row of the matrix, thus completing the closed-loop loading of the fixed-dimensional matrix. Simultaneously, the first-order difference vectors and discrete scalars are concatenated to assemble the behavior trajectory matrix of the current request flow. The system reads the negative event fingerprint matrix pre-stored in the kernel-mode read-only memory page, calculates the Frobenius norm of the behavior trajectory matrix and the negative event fingerprint matrix, and generates a threat quantification feature value representing the strength of the injected threat. When the threat quantification feature value is greater than the preset injection warning boundary value, the system forcibly rewrites the environmental risk scalar to its highest-order value, according to the formula... Drive preset permission threshold If the upper limit of the underlying authorization is exceeded, the system will then use the full bit mark in the access permission mask bitmap to identify the intercepted binary value 0, and force the full read pointer in the virtual address table to be redirected to the constant isolation area. This mechanism is linked to the underlying pointer address routing change path based on the quantization comparison results of the behavior trajectory matrix and the negative event fingerprint matrix. While maintaining the physical network connection status, it continuously outputs de-identified mask characters to the injection probe source, eliminating the structural conflict between application layer syntax compliance probe and underlying physical topology defense.

[0036] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.

Claims

1. A method for generating multidimensional data views and implementing secure access control with dynamic policies, characterized in that, Includes the following steps: Step S1: Obtain an access request containing business role information and real-time environment parameters. Map the business role information and real-time environment parameters to a multi-dimensional permission mapping matrix through a preset policy judgment logic. Obtain a field access weight sequence corresponding to the current access context through the inner product operation of the multi-dimensional permission mapping matrix and the preset business policy vector. Calculate and generate an access permission mask bitmap indicating field-level access permissions based on the comparison results of the values ​​of each element in the field access weight sequence with the preset permission threshold. Step S2: Push the access permission mask bitmap down to the storage engine index layer, and perform a bitwise logical AND operation between the access permission mask bitmap and the preset data tag index bitmap to determine the set of physical storage addresses that conform to the permission policy before the data block is loaded into memory, so as to generate a physical address mapping table. Step S3: During the encapsulation of the multidimensional data view descriptor, the physical address mapping table is parsed and the field attributes of the restricted fields are identified. By modifying the address pointer mapping relationship, the read pointer in the multidimensional data view descriptor that originally pointed to the original physical storage address of the restricted field is forcibly redirected to the pre-allocated constant isolation area in the kernel space memory. The constant isolation area stores mask characters that represent the desensitization status. Step S4: Construct a virtual addressing table based on the physical address mapping table and the redirected read pointer. Use the virtual addressing table to map application layer requests to the corresponding physical storage space or constant isolation area. While maintaining the logical structure of the multidimensional data view and without extracting the full amount of original plaintext data, generate a multidimensional virtual data view so that sensitive fields in the multidimensional virtual data view are redirected from the read pointer to the constant isolation area.

2. The method for generating multidimensional data views and implementing secure access control supporting dynamic policies according to claim 1, characterized in that, Step S1 includes: collecting dynamic environmental factors including visitor's geographical location, device fingerprint, historical behavior characteristics, and network access environment; normalizing the dynamic environmental factors and mapping them together with business role information into the feature space of the multi-dimensional permission mapping matrix; and extracting the field access weight sequence that indicates the legality of the current access behavior by calculating the correlation between the feature space and the preset business policy vector.

3. The method for generating multidimensional data views and implementing secure access control supporting dynamic policies according to claim 1, characterized in that, In step S2, a hard filtering barrier is established in the index space of the storage engine using an access permission mask bitmap. By blocking the loading link of unauthorized data blocks to computing memory before physical I / O is triggered, illegal field data cannot generate a physical resident image in the system memory address space. This eliminates the exposure window of sensitive plaintext information in the memory flow cycle from the physical source, thereby enhancing data sovereignty security in a multi-tenant environment.

4. The method for generating multidimensional data views and implementing secure access control supporting dynamic policies according to claim 1, characterized in that, In step S3, by identifying the field identifier of the restricted field in the multidimensional data view descriptor, the entry parameter in the physical address index table inside the multidimensional data view descriptor is modified using the address offset algorithm. The physical base address that originally pointed to the original disk image area is offset to the logical starting address of the constant isolation area, ensuring that the value retrieval instruction initiated by the application layer is hard redirected to the preset static storage area at the micro-instruction execution level.

5. The method for generating multidimensional data views and implementing secure access control supporting dynamic policies according to claim 1, characterized in that, The constant isolation zone is set as a read-only memory page in kernel mode, which stores a placeholder sequence composed of de-identified mask characters. The application layer is limited by the physical routing path of the read pointer and can only access the controlled physical page address through the virtual address table to read the de-identified mask characters in the constant isolation zone. This allows for data topology isolation and compliant viewing at the computing memory level without changing the upper-layer application logic.

6. The method for generating multidimensional data views and implementing secure access control supporting dynamic policies according to claim 1, characterized in that, In step S4, the virtual addressing table is constructed as a mapping matrix between logical addresses and physical addresses. The mapping matrix includes: a regular field pointer pointing to a specific offset address of the original data storage area, and a restricted field pointer generated in step S3 pointing to a specific address space of the constant isolation area. The regular field pointer is used to pass through non-sensitive data, and the restricted field pointer is used to intercept and block the addressing path of sensitive original data.

7. The method for generating multidimensional data views and implementing secure access control supporting dynamic policies according to claim 1, characterized in that, While keeping the same underlying raw physical data silently stored, the system supports dynamically generating multiple multi-dimensional virtual data views with different desensitization granularities and field visibility in the virtual mapping layer by switching or updating the access permission mask bitmap in real time. This process does not involve any data copying or moving operations in the physical storage layer, and achieves near real-time adaptive adjustment of data access permissions by changing the mapping topology of the logical addressing table.

8. The method for generating multidimensional data views and implementing secure access control supporting dynamic policies according to claim 1, characterized in that, In step S2, the physical address is filtered using the following bitwise logical operations: ,in, For access control mask bitmap, This is the tag bitmap index for the corresponding data shard in the storage engine. A candidate bitmap used to identify the set of loadable physical addresses that conform to the current policy; when When a specific bit in the memory is 1, the corresponding physical storage unit is determined to be within the legal loading range of the current access request.

9. The method for generating multidimensional data views and implementing secure access control supporting dynamic policies according to claim 1, characterized in that, The construction of multidimensional virtual data views is entirely converged on the underlying index bit operations and address pointer remapping. The computational overhead of the system when handling massive concurrent query requests does not increase linearly with the increase of the number of physical data rows. The response latency of view generation is limited by the index retrieval depth and the complexity of address table construction, thus providing sublinear growth in system throughput performance in high-dimensional differentiated access scenarios. The policy determination logic calculates differentiated field permissions based on the attributes of different business roles within the flexible employment ecosystem. Business roles include enterprises as the main employing entity, service platforms as collaborative entities, and administrative functional departments as regulatory entities. By matching different preset business policy vectors to different business roles, differentiated views that meet compliance requirements can be generated for each participant within the same original data pool.

10. A multidimensional data view generation and secure access control system supporting dynamic policies, used to implement the multidimensional data view generation and secure access control method supporting dynamic policies as described in claim 1, characterized in that, include: The permission calculation module is used to obtain access requests containing business role information and real-time environment parameters, map the business role information and real-time environment parameters to a multi-dimensional permission mapping matrix, and calculate and generate an access permission mask bitmap indicating field-level access permissions through the inner product operation of the multi-dimensional permission mapping matrix and the preset business policy vector. The storage index filtering module is used to push the access permission mask bitmap down to the storage engine index layer, so that the access permission mask bitmap and the data tag index bitmap preset in the underlying storage medium perform bit logic orthogonal operation, and determine the set of physical storage addresses that conform to the permission policy before the data block is loaded into the computing memory based on the bit operation result to generate a physical address mapping table. The addressing intervention module is used to parse the physical address mapping table and identify the field attributes of restricted fields during the encapsulation of multidimensional data view descriptors. By modifying the addressing pointer mapping relationship, it forces the read pointer in the multidimensional data view descriptor that points to the original physical storage address of the restricted field to a constant isolation area in kernel-mode memory. The constant isolation area stores desensitized mask characters. The virtual addressing mapping module is used to construct a virtual addressing table based on the physical address mapping table and the redirected read pointer. The virtual addressing table is used to generate a multidimensional virtual data view while maintaining the logical structure of the multidimensional data view and without extracting the full amount of original plaintext data.

Images