A computer network data management system based on AI information processing technology

Abstract

The application provides a computer network data management system based on an AI information processing technology, and relates to the technical field of computer network management, and has the characteristics that the system comprises a data acquisition and processing module, an AI intelligent analysis module, a computing power scheduling module, a security protection module, a heterogeneous compatibility module, an operation and maintenance control module and a compliance scheduling module which are connected in communication; the modules are cooperatively operated, and full-process closed-loop management of computer network data based on the AI information processing technology is realized; and the application has the advantages that through the cooperative operation of the modules, full-process closed-loop management of network data from acquisition, preprocessing, intelligent analysis, computing power scheduling to security protection, operation and maintenance control and compliance scheduling is realized, and the intelligentization, high efficiency, security and compliance level of network data management is improved.

Description

Technical Field

[0001] This invention relates to the field of computer network management technology, and in particular to a computer network data management system based on AI information processing technology. Background Technology

[0002] As computer networks rapidly evolve towards cloud-edge-device collaboration, multi-vendor heterogeneity, and high-concurrency flow control, network data exhibits characteristics of multi-source, heterogeneity, strong real-time requirements, and a surge in data volume. Traditional computer network data management systems are no longer able to meet the core needs of current network operation and maintenance, security protection, and compliance management.

[0003] Existing network data management systems have limited data collection scope, making them incompatible with multi-vendor proprietary protocols and heterogeneous network architectures, resulting in data silos. Analysis methods rely on rule-driven approaches, lacking adaptability, leading to high false positive and false negative rates in anomaly detection, and a lack of interpretability in the decision-making process. Computing power allocation is fixed, unable to be dynamically adjusted based on AI model training / inference needs and real-time network load. Security protection lacks effective means against adversarial attacks and model poisoning attacks targeting AI models, and privacy protection is insufficient. Operation and maintenance management is fragmented, lacking unified visualization and automation processes. Compliance management is inadequate, and operation records lack an immutable evidence storage mechanism.

[0004] Therefore, there is an urgent need for a computer network data management system that can achieve closed-loop management of the entire network data process, and has the functions of heterogeneous compatibility, intelligent analysis, dynamic computing power scheduling, security protection, convenient operation and maintenance and compliance control, in order to solve the shortcomings of existing technologies. Summary of the Invention

[0005] The purpose of this invention is to provide a computer network data management system based on AI information processing technology. Through the collaborative operation of multiple modules, it realizes closed-loop management of network data from collection, preprocessing, intelligent analysis, computing power scheduling to security protection, operation and maintenance management, and compliance scheduling, thereby improving the intelligence, efficiency, security, and compliance of network data management.

[0006] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows:

[0007] A computer network data management system based on AI information processing technology includes interconnected data acquisition and processing modules, AI intelligent analysis modules, computing power scheduling modules, security protection modules, heterogeneous compatibility modules, operation and maintenance management modules, and compliance scheduling modules. These modules operate collaboratively to form a closed-loop management system encompassing data input, intelligent analysis, computing power support, security assurance, operation and maintenance management, and compliance constraints. The specific module functions and their collaborative relationships are as follows:

[0008] The data acquisition and processing module, as the core of the system's data input, is used to collect multi-source heterogeneous network data and perform standardized preprocessing. This provides a unified format, high-quality, full-volume real-time dataset for subsequent AI intelligent analysis, forming the foundation for the entire system's operation. It communicates directly with the AI ​​intelligent analysis module, transmitting the preprocessed dataset to it in real time and receiving dataset optimization requests from the AI ​​intelligent analysis module. It collaborates with the heterogeneous compatibility module, leveraging its protocol adaptation capabilities to achieve data acquisition from multiple vendors' devices and using heterogeneous protocols. Furthermore, it works in conjunction with the compliance scheduling module to ensure that the collected data meets hierarchical and categorized management requirements, while simultaneously synchronizing the collection logs to the compliance scheduling module for auditing and evidence storage.

[0009] The data acquisition and processing module includes a data aggregation unit, a data cleaning unit, an automatic labeling unit, a data middleware unit, and a data drift detection unit.

[0010] The data aggregation unit collects network packets, logs, routing signaling and IoT streaming data through the edge aggregation gateway, and is compatible with multiple vendors' private protocols to achieve unified aggregation of multi-source data.

[0011] The data cleaning unit uses ETL tools to perform data deduplication, noise filtering, and format normalization to remove invalid and abnormal data and ensure data quality.

[0012] The automatic labeling unit combines transfer learning technology to achieve automatic labeling of network data, reduce manual labeling costs, and provide labeled sample data for AI model training.

[0013] The data platform unit uses protocol conversion middleware to break down data silos across vendors, network segments, and cloud-edge-device, providing a global data view and supporting data access from various modules. The data platform unit uses a distributed message queue to achieve real-time synchronization of data across network segments and has a data quality verification subunit to verify the integrity, consistency, and timeliness of the normalized data. Data that fails the verification is automatically fed back to the data cleaning unit for reprocessing.

[0014] The data drift detection unit monitors changes in network traffic characteristics in real time. When a significant shift in data distribution is detected, it automatically updates the training dataset and synchronizes it to the AI ​​intelligent analysis module to ensure the adaptability of the AI ​​model.

[0015] The AI ​​intelligent analysis module, as the core decision-making unit of the system, incorporates an interpretable AI unit and an incremental learning unit. It performs anomaly detection and traffic analysis on the dataset based on improved AI algorithms, generating interpretable and traceable scheduling decisions, thus representing the core of the system's intelligence. It establishes communication connections with the data acquisition and processing module, computing power scheduling module, security protection module, compliance scheduling module, and operation and maintenance management module: receiving the full real-time dataset transmitted by the data acquisition and processing module; sending computing power demand instructions to the computing power scheduling module and receiving computing power allocation feedback from the computing power scheduling module; sending security verification requests for AI decisions to the security protection module and receiving risk warnings and verification results from the security protection module; sending scheduling decisions to the compliance scheduling module and synchronizing decision logs for auditing and evidence preservation; and providing feedback on the AI ​​model's operating status and anomaly detection results to the operation and maintenance management module and receiving operation and maintenance instructions from the operation and maintenance management module.

[0016] The explainable AI unit uses the LIME or SHAP explainable algorithm to output the specific basis for AI decision-making. Together with the rule engine unit, it forms a dual decision-making mechanism of AI initial screening and rule fallback, which improves the reliability and traceability of decision-making.

[0017] The incremental learning unit adopts an online streaming training method, which uses new network traffic data in real time to iteratively fine-tune the AI ​​model without retraining the entire model, thereby improving the model's real-time adaptability and training efficiency.

[0018] The AI ​​intelligent analysis module also includes an integrated learning unit, which combines adaptive dynamic thresholds and integrates the analysis results of multiple sub-models to reduce the false alarm rate and false negative rate of the AI ​​model.

[0019] The rule engine unit presets network management rules to verify the AI ​​decision results. If the AI ​​decision exceeds the rule range, an alarm is automatically triggered and fed back to the operation and maintenance management module.

[0020] The AI ​​intelligent analysis module also includes a multimodal fusion unit, which is used to align and fuse different modal features of network packets, logs, routing signaling and IoT streaming data to generate a unified feature representation vector. The improved AI algorithm performs anomaly detection and traffic analysis based on the feature representation vector to improve the comprehensiveness and accuracy of the analysis.

[0021] The computing power scheduling module, as the system's computing power support unit, dynamically allocates AI training and inference computing power based on the computing power requirements of the AI ​​intelligent analysis module and the real-time network load, ensuring the efficient operation of the AI ​​model while avoiding waste of computing resources. It works in collaboration with the AI ​​intelligent analysis module, the heterogeneous compatibility module, and the operation and maintenance management module: receiving computing power requirements from the AI ​​intelligent analysis module in real time, collecting real-time network load data, and dynamically adjusting the computing power allocation scheme; providing feedback on computing power allocation status and resource usage to the operation and maintenance management module, and receiving computing power configuration instructions from the operation and maintenance management module; and achieving reasonable allocation of cloud-edge-device computing power through the cloud-edge collaboration capabilities of the heterogeneous compatibility module.

[0022] The computing power scheduling module includes a model lightweighting unit, a cloud-edge collaboration unit, and an elastic computing power unit:

[0023] The lightweight model unit employs pruning, quantization, or knowledge distillation techniques to compress the AI ​​model size and computing power consumption, adapting to the computing power resource limitations of edge nodes.

[0024] The cloud-edge collaborative unit adopts an architecture that combines edge local inference with cloud complex analysis. Edge nodes handle network anomalies and traffic scheduling with high real-time requirements, while cloud nodes handle complex data analysis tasks, realizing the hierarchical allocation of computing power.

[0025] The elastic computing unit is based on containerization technology and combined with a streaming computing framework to realize automatic expansion and contraction of computing resources. When the demand for AI computing power surges, it automatically adds computing nodes and automatically releases redundant computing power when the demand for computing power decreases, thereby improving the utilization rate of computing power.

[0026] The security protection module, serving as the system's security assurance unit, is used to defend against adversarial attacks and model poisoning attacks targeting the AI ​​model, protect network data privacy, and ensure the security and confidentiality of system operation. It establishes communication connections with all modules, providing security protection for the entire process, including data collection, AI analysis, computing power scheduling, and operation and maintenance: it encrypts data collected by the data collection and processing module; it performs adversarial training and security verification on the AI ​​intelligent analysis module's models to resist adversarial attacks and model poisoning attacks; it monitors the computing power allocation process of the computing power scheduling module to prevent unauthorized access to computing power; it verifies permissions and audits the operations of the operation and maintenance management module; and it encrypts and protects the evidence-based data of the compliance scheduling module to ensure data immutability.

[0027] The security protection module includes an adversarial training unit, a privacy protection unit, and a human-machine dual control unit:

[0028] The adversarial training unit pre-trains the AI ​​model using adversarial examples, allowing the model to learn the characteristics of the adversarial examples. Combined with the traffic baseline whitelist, it can effectively identify disguised malicious traffic and improve the model's resistance to attacks.

[0029] The privacy protection unit adopts a combination of federated learning and differential privacy technology. During the AI ​​model training process, only the AI ​​model parameters are transmitted, and the original network data is not transmitted. At the same time, privacy protection is achieved by adding Gaussian noise to the local model gradient to prevent data leakage.

[0030] The human-machine dual control unit is equipped with an AI decision-making manual approval whitelist, a policy circuit breaker mechanism, and a fault rollback mechanism. When the AI ​​issues an abnormal scheduling command, it automatically triggers the circuit breaker and rolls back to the safe configuration. At the same time, it is equipped with an emergency manual forced takeover entry point to ensure the security of the system in extreme situations.

[0031] The heterogeneous compatibility module, serving as the system's adaptation unit, is used to adapt to devices from multiple vendors, heterogeneous network protocols, and systems with different architectures. It addresses the issues of incompatibility between devices from multiple vendors, inconsistent protocols, and significant architectural differences in existing networks, ensuring the system can be flexibly deployed in various network environments. It works in conjunction with the data acquisition and processing module, the computing power scheduling module, and the operation and maintenance management module: providing multi-protocol adaptation capabilities to the data acquisition and processing module to ensure normal data acquisition from devices from multiple vendors; providing adaptation support for heterogeneous cloud-edge-device architectures to the computing power scheduling module, enabling computing power collaboration under different architectures; and providing a unified management interface for devices from multiple vendors to the operation and maintenance management module, enabling centralized operation and maintenance of heterogeneous devices.

[0032] The heterogeneous compatibility module includes an SDN management and control unit, a protocol adaptation unit, a network slicing unit, and a modular plug-in unit:

[0033] The SDN management and control unit adopts software-defined networking technology to uniformly manage switches, routers and legacy network equipment from multiple vendors, breaking down management barriers between vendors and achieving centralized management of equipment.

[0034] The protocol adaptation unit uses a protocol conversion gateway to convert between proprietary protocols from multiple vendors and standard protocols, ensuring data interoperability under different protocols.

[0035] The network slicing unit slices and isolates heterogeneous networks, isolates and transmits network data of different services, unifies data encapsulation format, and improves the security and stability of network transmission.

[0036] The modular plug-in unit adopts a plug-in architecture. When the network topology or services change, only the corresponding AI model plug-in needs to be replaced, without modifying the core system code, thus improving the system's flexibility and scalability.

[0037] The operation and maintenance management module, as the core unit of the system's operation and maintenance, provides a unified operation and maintenance interface and automated operation and maintenance processes for AI and the network, reducing operation and maintenance costs and improving efficiency. It establishes communication connections with all other modules to achieve full-process operation and maintenance management: receiving operational status data from each module and displaying it through a visual interface; sending operation and maintenance instructions to each module; and synchronizing operation and maintenance logs to the compliance scheduling module for auditing and evidence storage.

[0038] The operation and maintenance management module includes a visual operation and maintenance unit, an MLOps automated operation and maintenance unit, and a hierarchical permission unit:

[0039] The visualization operation and maintenance unit builds a graphical display interface for AI decision-making, network status, and anomaly alarms, supporting real-time viewing of network traffic, AI model analysis results, abnormal events, computing power usage, etc., to achieve visualized monitoring of operation and maintenance status.

[0040] The MLOps automated operation and maintenance unit constructs a fully automated operation and maintenance pipeline for model monitoring, version management, automatic fine-tuning, and fault alarms. When a degradation in AI model performance is detected, model fine-tuning is automatically triggered; when a module fault is detected, an alarm is automatically sent and preliminary fault investigation is performed.

[0041] The hierarchical permission unit assigns different operating permissions according to the responsibilities of the operation and maintenance personnel, and works in conjunction with an emergency takeover mechanism to ensure the security and standardization of operation and maintenance operations and prevent illegal operations.

[0042] The compliance scheduling module, as the system's compliance constraint unit, is used to perform hierarchical classification management and operational auditing of network data, ensuring that system operation complies with industry regulatory requirements and achieving traceability and delineation of responsibilities for data operations. It works collaboratively with all other modules: receiving collection logs from the data acquisition and processing module, decision logs from the AI ​​intelligent analysis module, computing power allocation logs from the computing power scheduling module, security event logs from the security protection module, and operation and maintenance logs from the operation and maintenance management module, and performing unified auditing and evidence storage; sending data hierarchical classification requirements to the data acquisition and processing module to ensure that collected data is managed according to its level; and sending compliance scheduling constraints to the AI ​​intelligent analysis module and the computing power scheduling module.

[0043] The compliance scheduling module includes a time-series prediction scheduling unit, a log auditing unit, and an integrity verification unit.

[0044] The time-series prediction scheduling unit uses a time-series traffic prediction algorithm to predict network traffic fluctuations and combines dynamic QoS scheduling technology to adjust bandwidth allocation priorities, ensuring the bandwidth requirements of core services while avoiding bandwidth resource waste, thus meeting the compliance requirements of network resource scheduling.

[0045] The log auditing unit automatically retains AI decision-making, scheduling, alarm logs and network data operation records to form a complete audit log, which supports log querying, filtering and exporting to meet regulatory auditing requirements.

[0046] The integrity verification unit will solidify AI decision records, operation records and log data on the blockchain. By utilizing the immutable and traceable characteristics of blockchain, it can achieve data immutability and the definition of rights and responsibilities. In the event of a security incident or compliance dispute, the responsible party can be traced through the blockchain-stored evidence data.

[0047] The beneficial effects of this invention are as follows:

[0048] 1. Through the interpretable AI unit, incremental learning unit and ensemble learning unit of the AI ​​intelligent analysis module, the anomaly detection and traffic analysis are made intelligent, reducing the false alarm rate and false negative rate, while providing decision interpretability and traceability, and solving the limitations of traditional rule-driven analysis.

[0049] 2. Through adversarial training, privacy protection, and human-machine dual control mechanisms in the security protection module, it effectively resists AI model attacks, protects network data privacy, and ensures system operation security.

[0050] 3. By collaborating with the data acquisition and processing module and the heterogeneous compatibility module, data silos are broken down, enabling unified acquisition and standardized preprocessing of data from multiple vendors' devices, heterogeneous protocols, and multiple nodes in the cloud, edge, and terminal, providing high-quality datasets for AI intelligent analysis.

[0051] 4. Through the dynamic allocation of computing power by the computing power scheduling module and the cloud-edge collaborative architecture, the computing power is adjusted in real time according to AI needs and network load, so as to achieve efficient utilization of computing power resources and avoid waste and insufficiency.

[0052] 5. By using the unified, visualized, and automated operation and maintenance process of the operation and maintenance management module, operation and maintenance costs can be reduced, while hierarchical permission management can ensure standardized operation and maintenance procedures.

[0053] 6. Through the hierarchical and classified management, log auditing and blockchain evidence storage of the compliance scheduling module, the traceability of data operations and the definition of rights and responsibilities are realized, which complies with industry regulatory standards. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the overall structure of a computer network data management system based on AI information processing technology according to the present invention.

[0055] Figure 2 This is a schematic diagram of the data acquisition and processing module framework of the present invention.

[0056] Figure 3 This is a schematic diagram of the AI ​​intelligent analysis module framework of the present invention. Detailed Implementation

[0057] To make the content of this invention easier to understand, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Identical components are indicated by the same reference numerals.

[0058] Example 1: Specific Deployment and Operation of a Computer Network Data Management System Based on AI Information Processing Technology

[0059] System hardware deployment: Two high-performance servers are deployed in the cloud for running core functions such as AI model training, complex data analysis, and computing power scheduling; 10 edge gateways are deployed at the edge for data collection, local inference, and real-time scheduling; and operation and maintenance terminals and security management terminals are deployed at the terminal.

[0060] System software deployment: The operating system uses CentOS 8.5, data acquisition uses Flume and Kafka, data cleaning uses Spark ETL, AI models use the TensorFlow framework, computing power scheduling uses the Kubernetes container platform, security protection uses the federated learning framework, and visual operation and maintenance uses Grafana.

[0061] Specific operation process of each module

[0062] Data Acquisition and Processing Module: The data aggregation unit collects multi-source data (network packets, device logs, routing signaling, IoT streaming data) from the campus network via the edge gateway, and converts it to a standard protocol format via the protocol adaptation unit of the heterogeneous compatibility module. The data cleaning unit performs deduplication, noise filtering, and JSON formatting. The automatic labeling unit uses a pre-trained model to assist in labeling data categories (normal traffic, DDoS attacks, port scans, etc.). The data middle platform unit achieves cross-network segment synchronization via Kafka, and the quality verification subunit checks data integrity, consistency, and timeliness; unqualified data is fed back to the cleaning unit for reprocessing. The data drift detection unit monitors traffic characteristics in real time, and notifies the AI ​​module to update the training set when the offset exceeds a threshold.

[0063] AI Intelligent Analysis Module: The multimodal fusion unit receives the full real-time dataset transmitted by the data acquisition and processing module, aligns and fuses the different modal features of network packets, logs, routing signaling, and IoT streaming data: extracts the port, IP, and packet length features of packets; the time, device ID, and operation type features of logs; the routing path and bandwidth usage features of routing signaling; and the monitoring values ​​and device status features of IoT data. These features are then fused into a unified feature representation vector through an attention mechanism and input into the improved AI algorithm.

[0064] The ensemble learning unit combines adaptive dynamic thresholds with the analysis results of three sub-models: CNN, LSTM, and XGBoost. An anomaly is identified when any two sub-models detect it; otherwise, the rule engine unit is triggered for verification. The rule engine unit pre-sets network management rules to verify the AI ​​decision-making results, ensuring accuracy.

[0065] The explainable AI unit uses the SHAP explainable algorithm to interpret the AI ​​decision results. For example, when a DDoS attack is detected, it outputs the IP address, port, attack type, and traffic characteristics on which the decision depends, thus enabling anomaly tracing. At the same time, the decision basis is synchronized to the visualization interface of the operation and maintenance management module for operation and maintenance personnel to view.

[0066] The incremental learning unit adopts an online streaming training method, receiving training datasets updated by the data acquisition and processing module in real time, and iteratively fine-tuning the AI ​​model. Each fine-tuning takes no more than 10 minutes, eliminating the need to retrain the entire model and ensuring that the model can adapt to changes in network traffic in real time. At the same time, the model training status and analysis results are synchronized to the computing power scheduling module and the operation and maintenance management module.

[0067] Computing power scheduling module: The model lightweighting unit uses knowledge distillation technology to distill complex AI models trained in the cloud into lightweight models, reducing computing power consumption and adapting to the computing power resource limitations of edge gateways.

[0068] The cloud-edge collaboration unit adopts an architecture that combines edge-local inference with cloud-based complex analysis: the lightweight model of the edge gateway handles tasks with high real-time requirements and immediately triggers local scheduling instructions when abnormal traffic is detected; the cloud server handles complex data analysis tasks and synchronizes the analysis results to the edge gateway and the operation and maintenance management module.

[0069] The elastic computing unit monitors AI computing power demand and network load in real time: when the demand for AI model training surges, it automatically expands by 3 computing power nodes to increase computing power supply; when the demand for computing power decreases, it automatically releases 2 redundant computing power nodes to ensure efficient utilization of computing power resources.

[0070] The computing power scheduling module receives computing power demand instructions from the AI ​​intelligent analysis module in real time, and dynamically allocates computing power resources based on the real-time network load to ensure the efficient operation of the AI ​​model.

[0071] Security Protection Module: The adversarial training unit uses adversarial example pre-trained models and combines them with a traffic whitelist to identify malicious traffic. Privacy Protection Unit: Federated learning is employed; each edge node only uploads model parameters, not raw data, and adds differential privacy noise. Human-Machine Dual Control Unit: Scheduling commands issued by AI undergo whitelist approval; abnormal commands trigger circuit breakers and roll back to the previous security configuration, supporting manual emergency takeover.

[0072] Heterogeneous Compatibility Module: The SDN management unit manages devices from multiple vendors uniformly via the OpenFlow protocol. The protocol adaptation gateway converts proprietary protocols such as Huawei VRP and Cisco IOS to SNMP or TCP / IP. The network slicing unit divides the network into office, IoT, and core business slices, isolating transmissions. The modular plug-in unit allows adding new devices by replacing only the corresponding plug-ins.

[0073] Operations and Maintenance Management Module: The visualized operations and maintenance unit displays information such as network traffic, AI decisions, computing power status, and security alarms. The MLOps pipeline automatically monitors model performance, triggering automatic fine-tuning when the false alarm rate exceeds the limit, and supports version rollback. The hierarchical permission unit divides operations and maintenance personnel into administrators, operations and maintenance personnel, and auditors, and supports emergency privilege escalation.

[0074] Compliance scheduling module: The order prediction scheduling unit uses LSTM to predict traffic fluctuations over the next 24 hours and dynamically adjusts QoS bandwidth priorities. The log auditing unit retains logs from each module for at least 6 months and supports conditional queries. The integrity verification unit uses hash chains to store key decision records to prevent tampering and enable accountability.

[0075] System collaborative operation process

[0076] In this embodiment, the modules work together to form a complete closed-loop management of the entire network data process. The specific process is as follows: Data acquisition and preprocessing: The data acquisition and processing module acquires multi-source heterogeneous network data through the protocol adaptation capability of the heterogeneous compatibility module. After cleaning, labeling, and quality verification, it forms a full real-time dataset in a unified format and stores it in the data platform.

[0077] AI Intelligent Analysis: The AI ​​Intelligent Analysis module acquires datasets from the data platform, generates feature representation vectors through the multimodal fusion unit, uses improved AI algorithms for anomaly detection and traffic analysis, and combines interpretable AI units and rule engine units to generate interpretable scheduling decisions, which are then synchronized to the computing power scheduling module, security protection module, and compliance scheduling module.

[0078] Computing power scheduling: The computing power scheduling module dynamically allocates computing power resources based on the computing power requirements of the AI ​​intelligent analysis module and the real-time network load through model lightweighting, cloud-edge collaboration, and elastic computing power allocation, to support the efficient operation of AI models and synchronize the computing power allocation status to the operation and maintenance management module.

[0079] Security Protection: The security protection module provides security protection for the entire process of data collection, AI analysis, and computing power scheduling, resists AI model attacks, protects data privacy, performs security verification on AI decisions, triggers circuit breakers and rollbacks in case of anomalies, and synchronizes security events to the operation and maintenance management module and the compliance scheduling module.

[0080] Operation and maintenance management: The operation and maintenance management module receives the operating status data of each module, displays it through a visual interface, executes automated operation and maintenance processes, ensures standardized operation and maintenance through hierarchical permission management, and synchronizes operation and maintenance logs to the compliance scheduling module;

[0081] Compliance scheduling: The compliance scheduling module manages data in a hierarchical and classified manner, audits the operation logs of each module, and uses blockchain notarization to ensure data immutability and the definition of rights and responsibilities. At the same time, it adjusts bandwidth allocation based on time-series traffic prediction to ensure that the system operation meets compliance requirements.

[0082] Closed-loop optimization: When the data drift detection unit detects a data distribution shift, it updates the training dataset, the AI ​​intelligent analysis module fine-tunes the model, the computing power scheduling module adjusts the computing power allocation, and the operation and maintenance management module optimizes the operation and maintenance strategy to continuously improve system performance.

[0083] Example 2: Extended Applications of the System

[0084] Data Acquisition and Processing Module: A new storage data acquisition unit has been added to collect server-stored data and tenant business data. The data platform unit has been optimized to support distributed storage and fast querying of massive amounts of data.

[0085] AI Intelligent Analysis Module: A new tenant behavior analysis unit has been added, which analyzes tenant network usage behavior based on AI algorithms, identifies abnormal tenant behavior, and optimizes the integrated learning unit to improve the efficiency of large-scale data analysis.

[0086] Computing power scheduling module: Added tenant computing power isolation unit to realize computing power isolation between different tenants, avoid computing power contention between tenants, optimize elastic computing power unit, and support computing power scheduling of tens of thousands of containers;

[0087] Security Protection Module: A new tenant data encryption unit has been added, which adopts end-to-end encryption technology to protect tenant data privacy. The adversarial training unit has been optimized to resist targeted attacks on tenant data.

[0088] Compliance scheduling module: A new tenant compliance audit unit has been added, which can formulate personalized audit strategies based on the compliance requirements of different tenants, optimize the integrity verification unit, and support the isolated storage of multi-tenant data.

[0089] After adaptation and adjustments, the system can meet the core requirements of cloud computing data centers, enabling secure management, real-time analysis, and compliance control of tenant data, thereby improving the operational efficiency and security of data centers.

[0090] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A computer network data management system based on AI information processing technology, characterized in that: It includes interconnected data acquisition and processing modules, AI intelligent analysis modules, computing power scheduling modules, security protection modules, heterogeneous compatibility modules, operation and maintenance management modules, and compliance scheduling modules; The data acquisition and processing module is used to acquire multi-source heterogeneous network data and perform preprocessing to construct a full real-time dataset in a unified format. The AI ​​intelligent analysis module has built-in interpretable AI units and incremental learning units, which are used to perform anomaly detection and traffic analysis on the dataset based on improved AI algorithms, and generate scheduling decisions with interpretability and traceability. The computing power scheduling module is used to dynamically allocate AI training and inference computing power according to the computing power requirements of the AI ​​intelligent analysis module and the real-time network load. The security protection module is used to defend against adversarial attacks and model poisoning attacks against AI models, and to protect network data privacy. The heterogeneous compatibility module is used to adapt to devices from multiple vendors, heterogeneous network protocols, and systems with different architectures. The operation and maintenance management module is used to provide a unified operation and maintenance interface and automated operation and maintenance process for AI and network. The compliance scheduling module is used to perform hierarchical classification management and operational auditing of network data; The data acquisition and processing module, AI intelligent analysis module, computing power scheduling module, security protection module, heterogeneous compatibility module, operation and maintenance management module, and compliance scheduling module work together to achieve closed-loop management of computer network data throughout the entire process based on AI information processing technology.

2. The computer network data management system based on AI information processing technology according to claim 1, characterized in that: The data acquisition and processing module includes a data aggregation unit, a data cleaning unit, an automatic labeling unit, a data middleware unit, and a data drift detection unit. The data aggregation unit collects network packets, logs, routing signaling, and IoT streaming data through the edge aggregation gateway, and is compatible with multiple vendors' proprietary protocols; The data cleaning unit uses ETL tools to perform data deduplication, noise filtering, and format normalization. The automatic annotation unit combines transfer learning technology to achieve automatic annotation of network data; The data platform unit uses protocol conversion middleware to break down data silos across vendors, network segments, and cloud-edge-device, providing a global data view. The data drift detection unit monitors changes in network traffic characteristics in real time and automatically updates the training dataset.

3. The computer network data management system based on AI information processing technology according to claim 1, characterized in that: The explainable AI unit uses the LIME or SHAP explainable algorithm to output the specific basis for AI decision-making, and works with the rule engine unit to form a dual decision-making mechanism of AI initial screening and rule fallback. The incremental learning unit adopts an online streaming training method, which uses new network traffic data in real time to iteratively fine-tune the AI ​​model; The AI ​​intelligent analysis module also includes an integrated learning unit, which combines adaptive dynamic thresholds to reduce the false alarm rate and false negative rate of the AI ​​model. The rule engine unit presets network management rules to verify the AI ​​decision results.

4. A computer network data management system based on AI information processing technology according to claim 1, characterized in that: The computing power scheduling module includes a model lightweighting unit, a cloud-edge collaboration unit, and an elastic computing power unit. The lightweight model unit employs pruning, quantization, or knowledge distillation techniques to compress the AI ​​model size and computing power consumption. The cloud-edge collaborative unit adopts an architecture that combines edge local inference with cloud-based complex analysis. Edge nodes handle network anomalies and traffic scheduling with high real-time requirements, while cloud nodes handle complex data analysis tasks. The elastic computing unit is based on containerization technology and combined with a streaming computing framework to achieve automatic expansion and contraction of computing resources.

5. A computer network data management system based on AI information processing technology according to claim 1, characterized in that: The security protection module includes an adversarial training unit, a privacy protection unit, and a human-machine dual control unit; The adversarial training unit pre-trains the AI ​​model using adversarial examples and combines it with a traffic baseline whitelist to identify disguised malicious traffic. The privacy protection unit adopts a combination of federated learning and differential privacy technology, only transmitting AI model parameters, and achieving privacy protection by adding Gaussian noise to the local model gradient; The human-machine dual control unit is equipped with an AI decision-making manual approval whitelist, a policy circuit breaker mechanism, and a fault rollback mechanism. When the AI ​​issues an abnormal scheduling command, it automatically triggers the circuit breaker and rolls back to the safe configuration with one click. It also has an emergency manual forced takeover entry point.

6. A computer network data management system based on AI information processing technology according to claim 1, characterized in that: The heterogeneous compatibility module includes an SDN management and control unit, a protocol adaptation unit, a network slicing unit, and a modular plug-in unit; The SDN management and control unit adopts software-defined networking technology to uniformly manage multi-vendor switches, routers and legacy network equipment. The protocol adaptation unit enables the conversion between proprietary protocols from multiple vendors and standard protocols through a protocol conversion gateway; The network slicing unit performs slicing and isolation on heterogeneous networks and unifies the data encapsulation format; The modular plug-in unit adopts a plug-in architecture, so when the network topology or services change, only the corresponding AI model plug-in needs to be replaced.

7. A computer network data management system based on AI information processing technology according to claim 1, characterized in that: The operation and maintenance management module includes a visual operation and maintenance unit, an MLOps automated operation and maintenance unit, and a hierarchical permission unit. The visualization operation and maintenance unit builds a graphical display interface for AI decision-making, network status, and anomaly alarms; The MLOps automated operation and maintenance unit constructs a fully automated operation and maintenance pipeline that includes model monitoring, version management, automatic fine-tuning, and fault alarms. The hierarchical permission unit assigns different operating permissions according to the responsibilities of the operation and maintenance personnel, and is coordinated with an emergency takeover mechanism.

8. A computer network data management system based on AI information processing technology according to claim 1, characterized in that: The compliance scheduling module includes a time-series prediction scheduling unit, a log auditing unit, and an integrity verification unit. The time-series prediction scheduling unit uses a time-series traffic prediction algorithm to predict network traffic fluctuations and combines dynamic QoS scheduling technology to adjust bandwidth allocation priorities. The log auditing unit automatically retains AI decision-making, scheduling, alarm logs, and network data operation records; The integrity verification unit will upload and solidify AI decision records, operation records, and log data to the blockchain to ensure that the data is tamper-proof and that responsibilities are clearly defined.

9. A computer network data management system based on AI information processing technology according to claim 2, characterized in that: The data platform unit uses a distributed message queue to achieve real-time synchronization of data across network segments, and is equipped with a data quality verification subunit to verify the integrity, consistency and timeliness of the normalized data. Data that fails the verification is automatically fed back to the data cleaning unit for reprocessing.

10. A computer network data management system based on AI information processing technology according to claim 1, characterized in that: The AI ​​intelligent analysis module also includes a multimodal fusion unit, which is used to align and fuse different modal features of network packets, logs, routing signaling and IoT streaming data to generate a unified feature representation vector. The improved AI algorithm performs anomaly detection and traffic analysis based on the feature representation vector.

Images