Assessment platform response method, device and equipment based on cloud service

Abstract

The invention relates to an assessment platform response method, device and equipment based on cloud services, and belongs to the technical field of cloud services, and the method comprises the steps: receiving an assessment request data packet of a user side, extracting a behavior feature identification code, screening an effective feature code based on user authority data, and generating a dynamic operation white list; allocating the request carrying the dynamic operation white list to an edge node, and dynamically adjusting the allocation weight according to the hardware resource occupancy rate of each node; establishing a cache hierarchy associated with user behaviors at the cloud, generating a standardized response template for the high-frequency assessment items, and storing the standardized response template in a nonvolatile memory area; performing integrity verification on response data before the response data is returned to the user side, and if the data is missing, extracting compensation data from a standby node according to the structural characteristics of the missing data to perform recombination; and monitoring the number of abnormal requests of the user side, and when a threshold value is reached, adding the feature code into the interception list and synchronizing the feature code to all edge nodes.

Description

Technical Field

[0001] The present invention relates to the field of cloud service technology, and in particular to a cloud service-based assessment platform response method, device and equipment. Background Art

[0002] When processing large-scale concurrent requests, current cloud-based assessment platforms often experience response delays due to uneven resource allocation at the edge nodes. Traditional permission verification mechanisms lack dynamic adjustment capabilities, making them easily bypassed by unauthorized requests. Existing caching strategies often rely on fixed templates and fail to dynamically optimize storage structures based on user behavior, resulting in inefficient processing of high-frequency requests. Furthermore, response data is prone to loss during transmission due to network fluctuations, and abnormal request monitoring typically relies on post-analysis, making it difficult to intercept malicious activity in a timely manner. This presents significant deficiencies in the real-time and security of the overall system. Summary of the Invention

[0003] The main purpose of this invention is to provide a cloud service-based assessment platform response method, device and equipment, To achieve the above-mentioned purpose, the present invention provides a cloud service-based assessment platform response method, comprising the following steps: Receive the assessment request data packet from the user end and extract the behavioral feature identification code, filter the valid feature codes based on the user permission data and generate a dynamic operation whitelist; Requests carrying a dynamic operation whitelist are assigned to edge nodes, and the allocation weight is dynamically adjusted based on the hardware resource occupancy of each node. When the GPU memory usage of the target node exceeds the threshold, its weight is reduced. A cache layer associated with user behavior is established in the cloud. Standardized response templates are generated for high-frequency assessment items and stored in non-volatile memory. When a cached feature identifier is detected, the template is preferentially called to generate a response data stream. Perform integrity check on the response data before returning to the user end. If data is missing, extract compensation data from the backup node based on the missing data structure characteristics and reconstruct it; Monitor the number of abnormal requests from the user side. When the threshold is reached, add the signature code to the interception list and synchronize it to all edge nodes.

[0004] Furthermore, the step of receiving the assessment request data packet from the user terminal and extracting the behavior feature identification code includes: After receiving the assessment request data packet from the user, a dynamic encryption key is generated based on the timestamp in the data packet header and the user's IP address, and the behavioral feature identification code is decrypted and verified; If the decrypted identification code deviates from the hash value of the pre-stored registration sequence by more than the preset tolerance range, the secondary authentication process on the user side is triggered, and a temporary interception log is generated at the edge node; The decryption verification process simultaneously detects the device fingerprint features carried by the identification code. If a virtual machine feature or an unauthorized hardware driver signature is detected, the request processing is directly terminated and the dynamic interception list is synchronously updated to the cloud service layer.

[0005] Furthermore, the steps of screening valid signature codes based on user permission data and generating a dynamic operation whitelist include: When filtering valid signature codes, the signature code is matched based on the role attributes in the user permission database and the real-time permission status. When it is detected that the remaining time of the permission validity period is less than the preset operation time, the expiration mark is automatically added to the whitelist; After the dynamic operation whitelist is generated, a decision tree model generated by training with historical user behavior data is used to perform secondary filtering on the feature codes in the whitelist, deleting feature codes with historical violation rates exceeding a threshold and generating a final whitelist sequence.

[0006] Furthermore, the steps of allocating the request carrying the dynamic operation whitelist to the edge node and dynamically adjusting the allocation weight according to the hardware resource occupancy of each node include: When dynamically adjusting the allocation weight, the real-time load factor is calculated based on the weighted value of the GPU memory occupancy rate and the CPU core utilization rate of the edge node. When the load factor exceeds the preset threshold, the weight allocation value of the node is proportionally reduced; At the same time, according to the priority field of the whitelist in the request data packet, high-priority requests are directed to the node group with a load factor lower than the threshold, and after the allocation is completed, the weight mapping table of each node is updated to the cloud service layer.

[0007] Furthermore, the steps of establishing a cache layer associated with user behavior in the cloud, generating standardized response templates for high-frequency assessment items, and storing them in a non-volatile memory area include: When establishing the cache hierarchy, the assessment items are divided into hot and cold data intervals according to the user's historical request frequency. For the assessment items in the hot data interval whose request frequency exceeds the set value, historical response data is extracted, and a standardized response template containing core fields is generated through cluster analysis; The templates are classified according to the request type and compressed using the LZ4 algorithm and stored in a designated partition of the non-volatile memory, and a bidirectional mapping index table of feature identification codes and storage offsets is established in the partition header.

[0008] Furthermore, when a cached feature identifier is detected, the step of preferentially calling the template to generate a response data stream includes: When calling the template to generate the response data stream, verify whether the difference between the timeliness mark of the standardized response template and the current request timestamp is within the preset window period; If it is within the window period, the template version is matched according to the user-side type code, and the static fields in the template are merged with the dynamic parameters of the current assessment request in a JSON nested manner; During the merging process, the dynamic parameters are hash-checked. If the check fails, the template regeneration process is triggered; the merged response data stream is encrypted using the AES-GCM algorithm and returned to the user end.

[0009] Furthermore, the steps of performing integrity check on the response data before returning to the user end and extracting compensation data from the backup node for reassembly based on the structural characteristics of the missing data if the data is missing include: Extract the header checksum from the response data that fails the integrity check, and locate the type identifier of the missing data segment based on the last four digits of the checksum; Nodes that match the type identifier and have a network delay of less than 30ms are selected from the backup node cluster. After obtaining the compensation data, a field difference algorithm is used to perform bit alignment and merge with the original data. The version mark of the original data is retained during the merging process and an incremental check field is added after the reorganization.

[0010] Furthermore, the steps of monitoring the number of abnormal requests from the user end and adding the signature code to the blocking list and synchronizing it to all edge nodes when the threshold is reached include: When monitoring abnormal requests from the user end, the number of triggers of the same feature code is counted within a preset time window and compared with the second threshold; When the threshold is exceeded, an interception mark containing an interception reason code and a timestamp is generated, and the feature code is bound to the interception mark and added to the dynamic interception list; During synchronization, a hash summary list is used for difference comparison, and only the newly added interception signature code and its mark are encrypted and transmitted to the edge node; After receiving the list, the edge node immediately replaces the current list in the memory and backs up the old list to the buffer, and returns the synchronization confirmation code to the cloud after the replacement is completed.

[0011] The present invention proposes a cloud service-based assessment platform response device, comprising: The request unit is used to receive the assessment request data packet from the user end and extract the behavior feature identification code, filter the valid feature code based on the user authority data and generate a dynamic operation whitelist; The weight unit is used to distribute requests carrying a dynamic operation whitelist to edge nodes, dynamically adjust the distribution weight according to the hardware resource occupancy of each node, and reduce its weight when the GPU memory usage of the target node exceeds the threshold; The cloud unit is used to establish a cache layer associated with user behavior in the cloud, generate standardized response templates for high-frequency assessment items and store them in the non-volatile memory area. When a cached feature identifier is detected, the template is preferentially called to generate a response data stream; A verification unit is used to perform integrity verification on the response data before returning it to the user end. If data is missing, compensation data is extracted from the backup node based on the structural characteristics of the missing data for reconstructing; The monitoring unit is used to monitor the number of abnormal requests from the user end. When the threshold is reached, the feature code is added to the interception list and synchronized to all edge nodes.

[0012] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned cloud service-based assessment platform response method are implemented.

[0013] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned cloud service-based assessment platform response method.

[0014] The cloud service-based assessment platform response method, device, and equipment provided by the present invention have the following beneficial effects: By calculating the GPU memory and CPU core occupancy of edge nodes in real time, the request allocation weight is dynamically adjusted, and high-load requests are distributed in a targeted manner based on the priority field, which significantly reduces the risk of node overload and improves resource utilization and response speed. Based on user behavior characteristics, hot and cold data intervals are divided and standardized response templates are generated. LZ4 compression and bidirectional index tables are combined to quickly match requests, reducing repeated calculation overhead. At the same time, template version matching and dynamic parameter nesting and merging are used to ensure the accuracy and real-time nature of response data. Dynamic encryption keys are used to verify device fingerprint features, combined with secondary authentication and interception log generation mechanisms to effectively identify forged requests. By real-time statistics of abnormal request feature codes and encrypted synchronization of dynamic interception lists to edge nodes, malicious behavior blocking at the millisecond level is achieved, thereby improving the overall defense capabilities of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a flow chart of a cloud service-based assessment platform response method according to an embodiment of the present invention; Figure 2 This is a structural block diagram of a cloud service-based assessment platform response device in one embodiment of the present invention; Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0016] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] Reference Figure 1 , which is a flow chart of a cloud service-based assessment platform response method proposed by the present invention, the method includes the following steps: S1, receives the assessment request data packet from the user end and extracts the behavioral feature identification code, filters the valid feature codes based on the user permission data and generates a dynamic operation whitelist; In step S1, after receiving the assessment request data packet from the user side, a dynamic encryption key is generated based on the timestamp in the data packet header and the user side IP address, and the behavioral feature identification code is decrypted and verified; if the decrypted identification code deviates from the hash value of the pre-stored registration sequence by more than the preset tolerance range, the user side secondary authentication process is triggered, and a temporary interception log is generated at the edge node; the decryption and verification process synchronously detects the device fingerprint feature carried by the identification code. If a virtual machine feature or an unauthorized hardware driver signature is detected, the request processing is directly terminated and the dynamic interception list is synchronously updated to the cloud service layer. When screening valid feature codes, feature codes are matched based on the role attributes and real-time permission status in the user permission database. When it is detected that the remaining time of the permission validity period is less than the preset operation time, the time limit identifier is automatically added to the whitelist; after the dynamic operation whitelist is generated, the feature codes in the whitelist are secondary filtered using a decision tree model generated by training the user's historical behavior data, and feature codes with historical violation rates exceeding the threshold are deleted to generate the final whitelist sequence.

[0019] During implementation, the assessment request packet sent by the user is first received. The timestamp and user IP address are extracted from the packet header, and a dynamic encryption key is generated based on these two factors (for example, hashing the concatenated timestamp and IP address using the SHA-256 algorithm). This key is used to decrypt the behavioral signature code within the packet and compare the decrypted code with the hash value of the pre-stored registration sequence. If the hash value deviation exceeds a preset tolerance (e.g., ±5%), a secondary authentication request containing a dynamic verification code is pushed to the user. A temporary interception log containing the request source IP address and the deviation value is generated at the edge node. During the decryption and verification process, the device fingerprint embedded in the behavioral signature code (e.g., GPU model hash, motherboard driver signature) is simultaneously parsed. If a virtual machine signature (e.g., VMware / VirtualBox driver signature) or an unauthorized hardware driver signature (e.g., a third-party driver certificate not on the whitelist) is detected, the request processing is immediately terminated. The signature code is then associated with the interception reason and encrypted and transmitted to the cloud service layer to update the global dynamic interception list. When screening valid feature codes, the computer device accesses the user permission database and matches the corresponding feature code based on the user's role attributes (such as administrator / ordinary user) and real-time permission status (such as activated / frozen). When the remaining time of the permission validity period is less than the preset operation time (for example, the remaining time is less than 30 minutes), the time limit identifier (such as the expiration timestamp field) is automatically added to the generated whitelist. Subsequently, the preliminarily generated whitelist is input into the pre-trained decision tree model (based on user historical behavior data training, features include historical request frequency, violation operation type, etc.), and the feature codes in the whitelist are filtered twice: the feature codes with historical violation rates exceeding the set threshold (such as violation rate > 15%) are deleted, and finally a whitelist sequence containing a unique serial number and version mark is generated for subsequent allocation process calls.

[0020] S2 distributes requests carrying a dynamic operation whitelist to edge nodes, dynamically adjusts the distribution weight based on the hardware resource occupancy of each node, and reduces its weight when the target node's GPU memory occupancy exceeds the threshold; In step S2, when dynamically adjusting the allocation weight, the real-time load coefficient is calculated based on the weighted value of the GPU memory occupancy and CPU core utilization of the edge node. When the load coefficient exceeds the preset threshold, the weight allocation value of the node is reduced proportionally; at the same time, according to the priority field of the whitelist in the request data packet, the high-priority request is directed to the node group with a load coefficient lower than the threshold, and after the allocation is completed, the weight mapping table of each node is updated to the cloud service layer.

[0021] During implementation, requests are allocated based on the priority field of the dynamic operation whitelist and the real-time load status of edge nodes. First, the cloud service layer obtains the GPU memory utilization of each edge node (for example, real-time collection via the NVML interface) and CPU core utilization (parsed via / proc / stat). The real-time load factor is calculated based on a preset ratio (e.g., 60% for GPU and 40% for CPU). If a node's load factor exceeds a threshold (e.g., 85%), its weight allocation is linearly reduced based on the excess ratio (e.g., a 3% reduction for every 1% increase).

[0022] At the same time, the priority field carried by the whitelist in the request data packet is parsed (for example, high priority is marked as "urgent assessment task"), and such requests are preferentially assigned to the node group with a load factor below the threshold and the highest resource redundancy (for example, a node cluster with a load factor of less than 70% is selected). After the allocation is completed, the computer device generates a weight mapping table containing the node ID, current weight value, and allocation timestamp, and uploads it to the cloud service layer through a TLS encrypted channel for dynamic call during subsequent request allocation. During this process, the allocation path of high-priority requests is independently recorded in the audit log and bound to the user feature code for traceability.

[0023] S3 establishes a cache layer associated with user behavior in the cloud, generates standardized response templates for high-frequency assessment items and stores them in non-volatile memory. When a cached feature identifier is detected, the template is preferentially called to generate the response data stream; In step S3, when establishing the cache level, the assessment items are divided into hot and cold data intervals according to the user's historical request frequency, and the historical response data is extracted for the assessment items whose request frequency in the hot data interval exceeds the set value. A standardized response template containing core fields is generated through cluster analysis; the template is classified according to the request type and compressed using the LZ4 algorithm and stored in the designated partition of the non-volatile memory, and a bidirectional mapping index table of feature identification code and storage offset is established in the partition header. When calling the template to generate the response data stream, it is verified whether the difference between the timeliness mark of the standardized response template and the current request timestamp is within the preset window period; if it is within the window period, the template version is matched according to the user-side type code, and the static fields in the template are merged with the dynamic parameters of the current assessment request in a JSON nested manner; the dynamic parameters are hashed during the merging process, and if the verification fails, the template regeneration process is triggered; the merged response data stream is encrypted using the AES-GCM algorithm and returned to the user end.

[0024] During implementation, the computer first divides assessment items into hot and cold data intervals based on historical user request frequencies (e.g., the number of requests in the past 30 days). The hot data interval is defined as assessment items with a request frequency exceeding 100 times per day, while the cold data interval is defined as those with a request frequency less than 10 times per day. For assessment items in the hot data interval, historical response data (e.g., successful response records from the past seven days) is extracted. The data structure is analyzed using the K-means clustering algorithm. After removing low-frequency fields, a standardized response template is generated containing only core fields (e.g., user ID, assessment score, and timestamp). Timeliness markers are also added (e.g., template generation time + 72-hour validity period).

[0025] Generated templates are categorized by request type (e.g., theoretical assessment, practical assessment), losslessly compressed using the LZ4 algorithm (with compression level set to 12), and stored in a separate partition of non-volatile memory (e.g., NVMe SSD). A bidirectional mapping index table is created in the partition header, using the feature identifier as the key and the offset as the value. A reverse index is also established to support fast reverse queries (e.g., locating the feature identifier by offset).

[0026] When a request contains a cached signature, the computer reads the corresponding template and verifies whether the difference between the timeliness marker and the current request timestamp is within a preset window (e.g., ±2 hours). If the verification passes, the computer matches the corresponding template version (e.g., a simplified field structure is used on mobile devices) based on the client type code (e.g., "Android_v10" for mobile devices or "Chrome_v98" for web devices). The static fields in the template (e.g., assessment rule descriptions) are then merged with the dynamic parameters of the current request (e.g., real-time test results) in a JSON nested format. The dynamic parameters are encapsulated under the "dynamic_data" key and hashed using the SHA-3 algorithm and embedded into the response header. If the hash verification fails (e.g., parameter tampering), the current template is discarded and a new template is generated based on the latest historical data. The merged response data stream is encrypted using the AES-GCM algorithm (with a key dynamically generated by the cloud and bound to the user's session ID), and a message authentication code (MAC) is appended before being returned to the client.

[0027] S4, perform integrity check on the response data before returning to the user end. If data is missing, extract compensation data from the backup node based on the missing data structure characteristics and reconstruct it; In step S4, the header check code is extracted from the response data that fails the integrity check, and the type identifier of the missing data segment is located according to the last four bits of the check code; Nodes that match the type identifier and have a network delay of less than 30ms are selected from the backup node cluster. After obtaining the compensation data, a field difference algorithm is used to perform bit alignment and merge with the original data. The version mark of the original data is retained during the merging process and an incremental check field is added after the reorganization.

[0028] During implementation, when the computer device performs an integrity check on the response data returned to the user, it first extracts the checksum from the response data header (e.g., generated using the CRC32 algorithm). The last four digits of the checksum (e.g., the last four hexadecimal characters) are mapped to a preset type identifier table (e.g., the last four digits "A3F1" correspond to "Missing Assessment Result Details Field"). The structural characteristics of the missing data segment (e.g., the missing field length, data type, and position offset) are determined based on the type identifier.

[0029] Subsequently, a node in the backup node cluster that stores matching type identifier data and has a network latency below 30ms (determined through real-time ping detection and node status reports) is selected. A compensation data request is then sent to the target node. After obtaining the compensation data, a field difference algorithm (e.g., based on binary bit mask comparison) is used to bit-align and merge the compensation data with the original data. The version tag of the original data (e.g., V1.2.3) is retained in the reassembled data header, and an incremental checksum field (e.g., a new SHA-256 hash value and a merge timestamp) is appended to the merged data.

[0030] If the merged data still fails the secondary integrity check, the standby node switching mechanism is triggered, and a node with a delay of less than 15ms is reselected to extract the compensatory data until the reorganization is successful or the maximum number of retries (such as 3 times) is reached, and then an error code is returned to the cloud service layer.

[0031] S5, monitors the number of abnormal requests from the user end. When the threshold is reached, the signature code is added to the blocking list and synchronized to all edge nodes.

[0032] In step S5, When monitoring abnormal requests from the user end, the number of triggers of the same feature code is counted within a preset time window and compared with the second threshold; When the threshold is exceeded, an interception mark containing an interception reason code and a timestamp is generated, and the feature code is bound to the interception mark and added to the dynamic interception list; During synchronization, a hash summary list is used for difference comparison, and only the newly added interception signature code and its mark are encrypted and transmitted to the edge node; After receiving the list, the edge node immediately replaces the current list in the memory and backs up the old list to the buffer, and returns the synchronization confirmation code to the cloud after the replacement is completed.

[0033] During implementation, the computer monitors user requests in real time, counting the number of times each signature code is triggered within a preset time window (e.g., 5 minutes) (e.g., using a sliding window algorithm), and comparing this count with a second threshold (e.g., >10 triggers within a single window). If the threshold is exceeded, the computer generates an interception tag embedded with an interception reason code (e.g., "Frequent abnormal request: Code_503") and a millisecond-accurate timestamp (e.g., 2023-08-25T14:30:45.123Z). The computer then binds the signature code to the tag and adds it to the dynamic interception list.

[0034] During synchronization, the cloud service layer generates a hash digest of the current block list (using the SHA-256 algorithm to hash the concatenated string of the signature and tag) and compares it with the digest list currently stored on the edge node. Only the newly added block signature and tag are encrypted with AES-256 and transmitted to the edge node (using QUIC protocol to reduce latency).

[0035] After receiving the encrypted data, the edge node decrypts and verifies the hash digest consistency. It then replaces the current block list in memory with the updated version and backs up the old list to a fixed-size circular buffer (e.g., retaining the three most recent backups). After the replacement is complete, the edge node generates a synchronization confirmation code (e.g., "Node_15:Ver_2.1.8") containing the node ID and the new list version number. This code is then returned to the cloud service layer via a persistent HTTPS connection. If the cloud doesn't receive the confirmation code within a timeout (e.g., 2 seconds), a retransmission mechanism is triggered until synchronization is successful.

[0036] Reference Attachment Figure 2 This is a block diagram of the device structure of a cloud service-based assessment platform response device proposed by the present invention, including: The request unit is used to receive the assessment request data packet from the user end and extract the behavior feature identification code, filter the valid feature code based on the user authority data and generate a dynamic operation whitelist; The weight unit is used to distribute requests carrying a dynamic operation whitelist to edge nodes, dynamically adjust the distribution weight according to the hardware resource occupancy of each node, and reduce its weight when the GPU memory usage of the target node exceeds the threshold; The cloud unit is used to establish a cache layer associated with user behavior in the cloud, generate standardized response templates for high-frequency assessment items and store them in the non-volatile memory area. When a cached feature identifier is detected, the template is preferentially called to generate a response data stream; A verification unit is used to perform integrity verification on the response data before returning it to the user end. If data is missing, compensation data is extracted from the backup node based on the structural characteristics of the missing data for reconstructing; The monitoring unit is used to monitor the number of abnormal requests from the user end. When the threshold is reached, the feature code is added to the interception list and synchronized to all edge nodes.

[0037] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, memory, display screen, input device, network interface and database connected via a system bus. The processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the above method is implemented.

[0038] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0039] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the above-described method when executed by a processor. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0040] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware using a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Any reference to memory, storage, database, or other media provided herein and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM.

[0041] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, apparatus, article, or method comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, apparatus, article, or method. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, apparatus, article, or method comprising the element.

[0042] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A cloud service-based assessment platform response method, characterized in that: The following steps are involved: Receive the assessment request data packet from the user end and extract the behavioral feature identification code, filter the valid feature codes based on the user permission data and generate a dynamic operation whitelist; Requests carrying a dynamic operation whitelist are assigned to edge nodes, and the allocation weight is dynamically adjusted based on the hardware resource occupancy of each node. When the GPU memory usage of the target node exceeds the threshold, its weight is reduced. A cache layer associated with user behavior is established in the cloud. Standardized response templates are generated for high-frequency assessment items and stored in non-volatile memory. When a cached feature identifier is detected, the template is preferentially called to generate a response data stream. Perform integrity check on the response data before returning to the user end. If data is missing, extract compensation data from the backup node based on the missing data structure characteristics and reconstruct it; Monitor the number of abnormal requests from the user side. When the threshold is reached, add the signature code to the interception list and synchronize it to all edge nodes.

2. The cloud service-based assessment platform response method according to claim 1, characterized in that: The steps of receiving the assessment request data packet from the user end and extracting the behavior feature identification code include: After receiving the assessment request data packet from the user, a dynamic encryption key is generated based on the timestamp in the data packet header and the user's IP address, and the behavioral feature identification code is decrypted and verified; If the decrypted identification code deviates from the hash value of the pre-stored registration sequence by more than the preset tolerance range, the secondary authentication process on the user side is triggered, and a temporary interception log is generated at the edge node; The decryption verification process simultaneously detects the device fingerprint features carried by the identification code. If a virtual machine feature or an unauthorized hardware driver signature is detected, the request processing is directly terminated and the dynamic interception list is synchronously updated to the cloud service layer.

3. The cloud service-based assessment platform response method according to claim 2, characterized in that: The steps for filtering valid signatures based on user permission data and generating a dynamic operation whitelist include: When filtering valid signature codes, the signature code is matched based on the role attributes in the user permission database and the real-time permission status. When it is detected that the remaining time of the permission validity period is less than the preset operation time, the expiration mark is automatically added to the whitelist; After the dynamic operation whitelist is generated, a decision tree model generated by training with historical user behavior data is used to perform secondary filtering on the feature codes in the whitelist, deleting feature codes with historical violation rates exceeding a threshold and generating a final whitelist sequence.

4. The cloud service-based assessment platform response method according to claim 1, characterized in that: The steps of allocating requests carrying a dynamic operation whitelist to edge nodes and dynamically adjusting the allocation weights based on the hardware resource occupancy of each node include: When dynamically adjusting the allocation weight, the real-time load factor is calculated based on the weighted value of the GPU memory occupancy rate and the CPU core utilization rate of the edge node. When the load factor exceeds the preset threshold, the weight allocation value of the node is proportionally reduced; At the same time, according to the priority field of the whitelist in the request data packet, high-priority requests are directed to the node group with a load factor lower than the threshold, and after the allocation is completed, the weight mapping table of each node is updated to the cloud service layer.

5. The cloud service-based assessment platform response method according to claim 1, characterized in that: The steps for establishing a cache layer associated with user behavior in the cloud, generating standardized response templates for frequently used assessment items, and storing them in a non-volatile memory area include: When establishing the cache hierarchy, the assessment items are divided into hot and cold data intervals according to the user's historical request frequency. For the assessment items in the hot data interval whose request frequency exceeds the set value, historical response data is extracted, and a standardized response template containing core fields is generated through cluster analysis; The templates are classified according to the request type and compressed using the LZ4 algorithm and stored in a designated partition of the non-volatile memory, and a bidirectional mapping index table of feature identification codes and storage offsets is established in the partition header.

6. The cloud service-based assessment platform response method according to claim 5, characterized in that: When a cached feature identifier is detected, the steps of preferentially calling the template to generate the response data stream include: When calling the template to generate the response data stream, verify whether the difference between the timeliness mark of the standardized response template and the current request timestamp is within the preset window period; If it is within the window period, the template version is matched according to the user-side type code, and the static fields in the template are merged with the dynamic parameters of the current assessment request in a JSON nested manner; During the merging process, the dynamic parameters are hash-checked. If the check fails, the template regeneration process is triggered; the merged response data stream is encrypted using the AES-GCM algorithm and returned to the user end.

7. The cloud service-based assessment platform response method according to claim 1, characterized in that: The steps of performing integrity check on the response data before returning to the user end and extracting compensation data from the backup node for reassembly based on the structural characteristics of the missing data if the data is missing include: Extract the header checksum from the response data that fails the integrity check, and locate the type identifier of the missing data segment based on the last four digits of the checksum; Nodes that match the type identifier and have a network delay of less than 30ms are selected from the backup node cluster. After obtaining the compensation data, a field difference algorithm is used to perform bit alignment and merge with the original data. The version mark of the original data is retained during the merging process and an incremental check field is added after the reorganization.

8. The cloud service-based assessment platform response method according to claim 1, characterized in that: The steps of monitoring the number of abnormal requests from the user end, adding the signature code to the blocking list when the threshold is reached, and synchronizing it to all edge nodes include: When monitoring abnormal requests from the user end, the number of triggers of the same feature code is counted within a preset time window and compared with the second threshold; When the threshold is exceeded, an interception mark containing an interception reason code and a timestamp is generated, and the feature code is bound to the interception mark and added to the dynamic interception list; During synchronization, a hash summary list is used for difference comparison, and only the newly added interception signature code and its mark are encrypted and transmitted to the edge node; After receiving the list, the edge node immediately replaces the current list in the memory and backs up the old list to the buffer, and returns the synchronization confirmation code to the cloud after the replacement is completed.

9. A cloud service-based assessment platform response device, characterized in that: include: The request unit is used to receive the assessment request data packet from the user end and extract the behavior feature identification code, filter the valid feature code based on the user authority data and generate a dynamic operation whitelist; The weight unit is used to distribute requests carrying a dynamic operation whitelist to edge nodes, dynamically adjust the distribution weight according to the hardware resource occupancy of each node, and reduce its weight when the GPU memory usage of the target node exceeds the threshold; The cloud unit is used to establish a cache layer associated with user behavior in the cloud, generate standardized response templates for high-frequency assessment items and store them in the non-volatile memory area. When a cached feature identifier is detected, the template is preferentially called to generate a response data stream; A verification unit is used to perform integrity verification on the response data before returning it to the user end. If data is missing, compensation data is extracted from the backup node based on the structural characteristics of the missing data for reconstructing; The monitoring unit is used to monitor the number of abnormal requests from the user end. When the threshold is reached, the feature code is added to the interception list and synchronized to all edge nodes.

10. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the cloud service-based assessment platform response method described in any one of claims 1 to 8 are implemented.

Images