Artificial intelligence detection device for computer system fault

Abstract

The invention relates to the technical field of artificial intelligence, and discloses a computer system fault artificial intelligence detection device comprising a data acquisition module configured to collect data from a plurality of data sources of a computer system and having data cleaning and sorting functions; the data encryption and uploading module is connected with the data acquisition module, encrypts the arranged data by adopting a homomorphic encryption algorithm, and uploads the encrypted data to a block chain; the block chain storage and index module is used for constructing a block structure for storing encrypted data on a block chain and establishing a data index mechanism based on a specific data structure; and an artificial intelligence diagnosis module. Through multi-source data acquisition, a comprehensive information basis is provided for fault diagnosis. In a data processing process, a convolutional neural network is adopted to extract spatial features, a recurrent neural network is adopted to process time sequence features and attention mechanism focusing key information, and a complex fault mode can be accurately identified.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence, and specifically to an artificial intelligence detection device for computer system failures. Background Art

[0002] With the rapid development of computer technology, computer systems are increasingly widely and deeply applied in various fields. From enterprise-level data centers to intelligent devices for personal daily use, computer systems have become an indispensable infrastructure for the operation of modern society. However, the complexity of computer systems also makes it inevitable for failures to occur. Once a failure occurs, it may bring huge economic losses to enterprises, such as business interruption, data loss, and decreased customer satisfaction. In some critical fields, it may even endanger life safety and social stability.

[0003] Traditional computer system failure detection methods mainly rely on manual experience and simple monitoring tools. Maintenance personnel judge whether the system has a failure by observing the basic operating parameters of the system, such as CPU usage rate, memory occupancy, etc., combined with an understanding of common failure modes. This method is inefficient and has limited accuracy when dealing with complex computer systems. On the one hand, manual monitoring is difficult to cover all aspects of the system in real time and comprehensively, and many potential failure hazards may be ignored. For example, some deep hardware failures may only show minor performance changes in the early stage and are not easily detected by humans; on the other hand, with the continuous update and replacement of computer system software and hardware technologies, the types and characteristics of failures are becoming increasingly complex and diverse. Traditional experience-based fault diagnosis methods are difficult to adapt to this rapid change, and it is often impossible to identify new types of failures in a timely and accurate manner. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides an artificial intelligence detection device for computer system failures, which solves the problem that traditional experience-based fault diagnosis methods often cannot identify new types of failures in a timely and accurate manner.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: An artificial intelligence detection device for computer system failures, comprising:

[0006] A data acquisition module, which is configured to collect data from multiple data sources of the computer system and has the functions of data cleaning and sorting;

[0007] A data encryption and upload module, which is connected to the data acquisition module, encrypts the sorted data using a homomorphic encryption algorithm, and uploads the encrypted data to the blockchain;

[0008] Blockchain storage and indexing module, which constructs a block structure for storing encrypted data on the blockchain and simultaneously establishes a data indexing mechanism based on a specific data structure;

[0009] Artificial intelligence diagnosis module, which can obtain encrypted data from the blockchain and decrypt it;

[0010] Audit module, which is responsible for recording event information related to data upload requests, download requests, and diagnostic operations.

[0011] Preferably, the data acquisition module includes: a data acquisition unit, which simulates the reading and writing of data blocks of a specific size to obtain the reading and writing speed when acquiring the memory reading and writing speed data; an operating system log acquisition unit, which extracts logs related to specific system core services when acquiring the operating system event logs in the system.

[0012] Preferably, the data encryption and upload module includes: a data encryption unit, which uses a homomorphic encryption algorithm with a specific encryption principle to encrypt data; an authorization interaction unit, which decides the data upload authorization according to the specific authorization decision logic in the smart contract.

[0013] Preferably, the blockchain storage and indexing module includes: a blockchain storage unit, which constructs blocks with a specific structure for storing encrypted data.

[0014] Preferably, the artificial intelligence diagnosis module includes: a convolutional neural network unit, which enables the convolutional layer to have specific parameter settings to achieve the function of extracting data spatial features.

[0015] Preferably, the audit module includes: an audit log recording unit, where the structure defined in the smart contract contains specific information for recording audit events, and stores audit log events through a specific array, and creates corresponding instances in the smart contract function of the relevant operation and adds them to the array.

[0016] Preferably, the data acquisition module further includes: an application data acquisition unit, which uses its internal performance monitoring interface to obtain data when acquiring application operation data, and parses the access log to extract specific key information and combines it with the error log information to form a comprehensive data record.

[0017] Preferably, the artificial intelligence diagnosis module further includes: a recurrent neural network unit, where the long short-term memory network adopted processes the data time series features according to a specific calculation formula.

[0018] Preferably, the artificial intelligence diagnosis module further includes: a fault classification and prediction unit, which classifies and predicts faults based on the fully connected layer and the softmax layer.

[0019] Preferably, the audit module further includes: a risk monitoring unit that analyzes the audit log data and monitors potential risk patterns of data access and usage through a preset risk model and algorithm.

[0020] The present invention provides an artificial intelligence detection device for computer system failures. It has the following beneficial effects:

[0021] 1. Through multi-source data collection, covering hardware status data, software operation data, and network data, the present invention provides a comprehensive information basis for fault diagnosis. During the data processing process, a convolutional neural network is used to extract spatial features, a recurrent neural network is used to process temporal features, and an attention mechanism is used to focus on key information, which can accurately identify complex fault patterns, effectively avoiding diagnostic errors caused by information loss or misjudgment, greatly improving the accuracy of fault diagnosis, and reducing unnecessary maintenance costs and system downtime caused by misjudged faults.

[0022] 2. The data encryption and upload module of the present invention uses a homomorphic encryption algorithm to encrypt the data to ensure the security of the data during transmission and storage. Even if the data is transmitted in an untrusted environment or stored on the blockchain, the encryption measures can effectively prevent the data from being stolen or tampered with. At the same time, the smart contract strictly manages the authorization of data upload and access, comprehensively considering multiple factors such as the reputation score of the requester, business relevance, and system security risk level. Only data operations with authorized approvals are allowed, which not only protects the privacy rights and interests of data owners but also makes the entire computer system more secure and reliable during data interaction.

[0023] 3. Through the blockchain-based storage and indexing mechanism, the present invention makes the storage of data immutable and traceable. Each data block contains the hash value of the previous block, forming a complete data chain, and any modification to the data can be easily detected. The audit unit constructs a detailed audit log in the smart contract by recording event information such as data upload requests, download requests, and diagnostic operations, including relevant information of the requester, authorization decision results, timestamps, and fault categories. This not only provides a clear record of the data flow and usage for data owners or regulatory agencies to monitor and review but also enables quick tracing of relevant responsible persons in case of data leakage or illegal use events, enhancing the transparency and compliance of data management, contributing to the establishment of a reliable data management ecosystem, and promoting the effective management and utilization of data assets by enterprises and organizations. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is the architecture diagram of the artificial intelligence detection device in the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0026] Please refer to the attached Figure 1 , the embodiment of the present invention provides an artificial intelligence detection device for computer system failures, including:

[0027] 1. Data acquisition module

[0028] In this embodiment, the data acquisition module is configured to collect data from multiple data sources of the computer system. The data sources cover hardware sensors, operating system APIs, and application program monitoring interfaces. The collected data includes hardware status information, operating system event logs, and application program running data. And this module has the function of cleaning and sorting the collected data, which can remove outliers and redundant information, and integrate the data from different sources into a unified format for subsequent processing, specifically as follows:

[0029] In this embodiment, when the data acquisition unit collects data for the hardware part of the computer system, it establishes connections with a variety of hardware sensors. For the CPU temperature sensor, it communicates with it through a specially adapted hardware driver according to the SMBus communication protocol, and collects temperature data at an interval of every 3 seconds. When collecting data of the memory module, it interacts with the SPD chip of the memory. According to its specific communication specifications, it can not only obtain information such as the basic capacity and frequency of the memory, but also accurately obtain the read and write speed data of the memory through a series of read and write test operations. When testing the read and write speed, it will simulate read and write operations of data blocks of different sizes such as 4KB, 64KB, and 1MB, record the time consumed for each read and write operation, and thus calculate the memory read and write speed under different data block sizes. For hard disk data acquisition, it uses the SMART interface of the hard disk and obtains comprehensive hard disk health status information, including power-on time, start-up times, seek error rate, reallocated sector count, etc. according to the SMART protocol standard instruction set. At the same time, in order to evaluate the I / O performance of the hard disk, various types of file read and write operation tests, such as sequential read and write and random read and write operations, will be carried out, measure the read and write time consumed under different operations, and calculate the average read and write speed. These data can intuitively reflect the performance status of the hard disk and potential failure risks.

[0030] In terms of collecting operating system event logs, when running on the Windows system, the data collection unit uses the EventLog API to obtain data. It can extract only the error, warning, and critical event logs related to the system core services in the past week, such as the key services in Windows services, such as DNS service, DHCP service, etc., and the driver, according to the preset filtering rules. During the collection process, the original log data will be format-converted from the Windows-specific binary log format to a unified parsable text format, and redundant information such as the fixed format description at the log header will be removed, only retaining key information such as the event description and event time for subsequent analysis and processing. In the Linux system, data is collected through the syslog mechanism, and important log information generated by the sshd system daemon process and kernel modules is collected according to the log level and source filtering rules set in the configuration file. The collected logs will be stored in the corresponding folders according to different service names or process IDs. For example, the logs of the sshd daemon process are stored in the " / var / log / sshd / " folder, so that when it is necessary to analyze the running status of a specific service later, the relevant log data can be quickly located and retrieved.

[0031] For the collection of application running data, the data collection unit makes full use of the monitoring interfaces or log output functions provided by the application itself. Taking the Web server application Apache as an example, in addition to collecting data through its access log and error log interfaces, it also uses its internal performance monitoring interface to obtain data such as the current number of connections, request processing time, and response status code distribution. When processing the access log, the log content will be parsed in real time, and key information including the client IP address, the requested URL path, and the request method POST will be extracted, and this information will be combined with the error codes and error description information in the error log to form a comprehensive record of the application running data. These records will be integrated and stored in hourly time periods. For example, all the application running data within an hour will be compressed and stored in a file in a specific format, and the file name contains time information, so that the change in the running status of the application at different time periods can be quickly queried and analyzed later, and potential fault hazards can be discovered in time. After the data collection unit completes data collection, it will transfer this data to the data preprocessing module for the next step of processing to ensure the integrity and availability of the data and lay a foundation for the subsequent fault detection process.

[0032] 2. Data Encryption and Upload Module

[0033] In this embodiment, the data encryption and upload module is connected to the data acquisition module. It is used to encrypt the sorted data using a specific homomorphic encryption algorithm and upload the data to the blockchain after encryption. This module interacts with the smart contract and decides whether to allow data upload according to the authorization rules based on multiple factors set in the smart contract. The multiple factors include the requester's reputation, the degree of business association with the data owner, and the system security risk level. Specifically as follows:

[0034] 2.1. Data Encryption:

[0035] The data encryption and upload unit encrypts the data using the Paillier homomorphic encryption algorithm. This algorithm is based on the composite residue class problem and has good additive homomorphic properties. During the encryption process, a public-private key pair is first generated. The public key is (g, n), where n = pq (p and q are two large prime numbers). For the plaintext data m to be encrypted, the encryption function is:

[0036] E(m) = g m ×r n mod n 2

[0037] where r is a random number. When encrypting the collected CPU temperature data t = 50°C (assuming the data has been preprocessed and digitized), according to the generated public key (g, n) and random number r, the encrypted ciphertext data E(t) is calculated according to the encryption function. This encryption method can protect data privacy while allowing specific operations on the ciphertext, providing security for subsequent data processing;

[0038] 2.2. Upload and Authorization Interaction

[0039] Before the encrypted data is uploaded to the blockchain, it needs to interact with the smart contract to determine whether it is authorized to be uploaded. The authorization rules set in the smart contract are based on multiple factors. Among them, the requester's reputation score R is an important factor. The initial value of the requester's reputation score is set to 0.5 and will be updated according to its behavior after each data access. If the requester complies with the rules during data access and it is found through the analysis after its data access that the accuracy of fault diagnosis has been improved by X% (X is greater than 20) by using the data it provides, for example, the original fault diagnosis accuracy of 70% has been improved to 90%, then the reputation score increases by 0.1. On the contrary, if the rules are violated, such as sharing the data with a third party without authorization or maliciously tampering with the data resulting in damaged data integrity, then the reputation score decreases by 0.2. The update operation of the reputation score is completed in the automatic execution process of the smart contract after each data access, and the updated reputation score is recorded in a specific data storage area of the blockchain. This area only allows authorized reputation management modules to perform read and modification operations;

[0040] Another factor is the business relevance B between the requester and the data owner. If the requester has a long-term hardware maintenance cooperation agreement with the data owner and the number of hardware failures successfully resolved in the past year accounts for more than Y% of the total number of hardware failures (Y is greater than 60), for example, more than 70 hardware failures out of a total of 100 have been successfully resolved, then the business relevance value is relatively high, such as 0.8. If it is a new partner or the number of business transactions with the data owner in the past three years is less than Z times (Z is less than 5) and the business type has a low relevance to the current data, such as only occasional marginal business cooperation, then the value is relatively low, such as. The evaluation and calculation process of the business relevance is performed by a special evaluation function in the smart contract. The input parameters of the evaluation function include cooperation agreement records, fault resolution records, and business transaction details. These data are stored in a specific data structure on the blockchain for smart contracts to read and process;

[0041] The system security risk level S is also a key factor in authorization decisions. It is determined through an assessment method based on a risk matrix, and network threats are divided into three levels: high, medium, and low, and software vulnerabilities are divided into three levels: serious, general, and minor. If the system detects a serious software vulnerability and suffers a high-level network threat, such as a large-scale port scanning attack from a specific malicious source address and a buffer overflow vulnerability that can be exploited remotely, the security risk level is close to 1; if the system security is good, there are no obvious vulnerabilities and the network threat is low, that is, the network traffic is normal, no abnormal connection requests are detected, and the software vulnerability scanning tool does not find medium- and high-risk vulnerabilities, then the value is close to 0. The assessment of the security risk level is triggered and executed periodically by the smart contract, for example, once every 12 hours. The network threat detection data based on the assessment process comes from the real-time monitoring data of the network intrusion detection system, and the software vulnerability information comes from the scanning results of professional vulnerability scanning tools. These data are transmitted to the smart contract through a specific data interface for comprehensive evaluation. Authorization decision function in the smart contract:

[0042]

[0043] Where I = {R, B, S} is the input vector, W h is the hidden layer weight matrix, b h is the hidden layer bias vector, is the ReLU activation function, W σ is the output layer weight vector, b σ is the output layer bias, σ is the sigmoid activation function, when A ≥ T (T is the authorization threshold, such as 0.5), the authorization is passed and the data can be uploaded to the blockchain; otherwise, the upload is rejected and the requester information and the rejection reason are recorded in the audit log.

[0044] 3. Blockchain Storage and Indexing Module

[0045] In this embodiment, the blockchain storage and indexing module constructs a block structure for storing encrypted data on the blockchain. The block contains encrypted data, its hash value, timestamp, data source identifier, and the hash value of the previous block. At the same time, a data indexing mechanism based on a specific data structure is established to quickly locate and retrieve the data stored in the blockchain, as follows:

[0046] 3.1 Blockchain Storage

[0047] The blockchain storage and indexing unit constructs a block structure for storing encrypted data on the Ethereum blockchain. Each block contains encrypted data E(X), its hash value H(E(X)) (calculated using the SHA-256 hash algorithm, that is, converting the encrypted data into a byte stream and then performing SHA-256 calculation to obtain a hash value of a fixed length), timestamp TS, data source identifier SI, and the hash value of the previous block PrevH. The formed block structure is:

[0048] Block = {E(X), H(E(X)), TS, SI, PrevH}

[0049] When storing the encryption result of the hardware state data of a computer system once, the encrypted hardware data is used as E(X). After calculating its hash value, the timestamp for collecting this data, the identifier indicating that the data source is a hardware sensor, and the hash value of the previous block are added to construct a complete block and store it on the blockchain. Such a storage structure ensures the immutability and traceability of the data. Once the data is stored on the blockchain, any modification to the data will cause a change in the hash value, which will be detected by the nodes in the blockchain network;

[0050] 3.2 Data Index Construction

[0051] Meanwhile, the unit establishes a data indexing mechanism based on Merkle Patricia Trie. For each data feature in the data matrix X, it uses the feature as the key and stores the storage location (block number and offset) of the corresponding encrypted data in the blockchain as the value in the MPT. For example, for the CPU temperature data column, "CPU_Temperature" is used as the key, and the storage location information in the blockchain is used as the value. In Ethereum, the mapping data structure in the smart contract is used to implement some functions of the MPT. Define a mapping dataIndex, whose key is the data feature name, and the value is a structure type containing block number and offset information. When querying a specific type of data, such as querying the historical records of CPU temperature data, by passing "CPU_Temperature" as a parameter through the query function provided by the smart contract, the storage location of the data in the blockchain can be quickly located, greatly improving the data retrieval efficiency. This data indexing mechanism enables the rapid and accurate acquisition of the required data in large-scale blockchain data storage, providing efficient data support for fault diagnosis.

[0052] 4. AI Diagnosis Module

[0053] In this embodiment, the AI diagnosis module can obtain and decrypt the encrypted data from the blockchain. This module includes a convolutional neural network module, a recurrent neural network module, an attention mechanism module, and a fault classification prediction module. The convolutional neural network module is used to extract data spatial features, the recurrent neural network module processes data temporal features, the attention mechanism module focuses on key information, and the fault classification prediction module classifies and predicts faults based on the fully connected layer and the softmax layer, as follows:

[0054] 4.1. Data Acquisition and Decryption

[0055] When performing fault diagnosis, the AI diagnosis unit first obtains and decrypts the encrypted data from the blockchain. The private key (λ, μ) of the Paillier encryption algorithm is used for decryption, where:

[0056] λ = lcm(p - 1, q - 1)

[0057] (lcm represents the least common multiple), and the decryption function:

[0058]

[0059] where c is the encrypted data, For example, when obtaining the ciphertext c of the encrypted CPU temperature data, the original CPU temperature data is calculated according to the private key information using the decryption function;

[0060] 4.2. Feature Extraction of the CNN Layer

[0061] The decrypted data is divided into multiple data segments according to the time series, and then input into the convolutional neural network module. The convolutional layer of the CNN module adopts specific parameter settings. The size of the convolutional kernel is 3*3, the stride is 1, the padding is 1, and the number of input channels is determined according to the number of types of collected data features. After the output of the batch normalization layer is completed, it then passes through the ReLU activation function layer to map the output to non-negative. Through these operations, the CNN layer can effectively extract the spatial features of the data. For example, for hardware data, it can identify the spatial correlation patterns between different hardware parameters, and for software log data, it can extract the feature relationships of log events in the spatial dimension, thus providing important feature information for subsequent fault diagnosis.

[0062] 4.3. Temporal Feature Processing of the RNN Layer

[0063] The sequence of feature maps extracted by the CNN layer is then input into the recurrent neural network (RNN) module. Here, the long short-term memory network is adopted. The RNN layer can process the temporal features of the data through calculation. For example, it can model the sequence relationship of events occurring in the system log data and capture the change trend of the hardware state data in the time series, thus discovering potential fault patterns. For example, if the CPU temperature continues to rise for a period of time, the RNN layer can learn this temporal feature and use it as an important basis for fault diagnosis.

[0064] 5. Audit Module

[0065] In this embodiment, the audit module works in coordination with the data encryption and upload module and the artificial intelligence diagnosis module, and is responsible for recording event information related to data upload requests, download requests, and diagnostic operations, including information related to the requester, authorization decision results, timestamps, and fault categories, etc. Specific structures and arrays are defined in the smart contract to store audit logs, and the audit logs are stored in accordance with a specific encrypted storage method, and only the authorized audit management module can decrypt and view them. The authorization process is executed according to the comprehensive rules of the smart contract, as follows:

[0066] 5.1. Audit Log Recording

[0067] The audit unit works in collaboration with the data encryption and upload unit and the artificial intelligence diagnosis unit, and is responsible for recording event information related to data upload requests, download requests, and diagnostic operations. In the data upload request event, it records the reputation score Ru, business relevance Bu, system security risk level Su, authorization decision result Au, and request timestamp TSu of the requester. For example, when a hardware vendor requests to upload hardware detection data of a computer system, the audit unit will record the vendor's reputation score (such as derived from its past data usage compliance and contribution to fault diagnosis), business relevance to the data owner (such as based on their cooperation history and frequency of business transactions), the current system's security risk level (such as considering whether there are potential security vulnerabilities in the system targeting the hardware), the authorization decision result (whether to allow the upload), and the request timestamp. For the data download request event, it similarly records the relevant information Rd, Bd, Sd, Ad, and TSd of the requester. In the diagnostic operation event, it records the relevant information of the diagnostic requester, the fault category F obtained from the diagnosis, and the diagnostic operation timestamp TS. The audit logs are stored by defining specific structures and arrays in the smart contract.

[0068] 5.2. Audit Log Storage and Management

[0069] The storage of audit log data on the blockchain adopts a specific encrypted storage method to ensure the integrity and confidentiality of the audit log data. Only the authorized audit management module can decrypt and view the content of the audit logs, and the authorization process is also based on the authorization rules in the smart contract, which comprehensively considers factors such as the identity information of the auditor, the permission level, and the current system's audit requirements. The beneficial effect of this audit mechanism is that it can provide a complete audit trail of data usage, facilitating data owners or regulatory agencies to monitor and review the flow and usage of data. In the event of a data breach or illegal use, the relevant responsible parties can be quickly traced, enhancing data security and compliance. At the same time, through the analysis of audit logs, potential risk patterns in data access and usage can be discovered. For example, if a certain requester frequently requests to download specific types of data, there may be a risk of data abuse, and thus timely measures can be taken to prevent it, further ensuring the safe and stable operation of the computer system.

[0070] Working Principle: The present invention provides an artificial intelligence detection device for computer system failures, specifically as follows:

[0071] As the data source of the entire device, the data acquisition module widely collects data from multiple data sources such as the hardware sensors of the computer system, the operating system API, and the application monitoring interface. When collecting hardware data, through the adapted driver and communication protocol, such as the SMBus protocol for the CPU temperature sensor, the specific specification of the memory SPD chip, and the hard disk SMART interface standard, the hardware status information is accurately obtained, including temperature, read and write speed, health status, etc., and performance tests are carried out to evaluate the potential failure risk. For the operating system event logs, in the Windows system, key logs are extracted according to the EventLog API and combined with filtering rules, and the format is converted to remove redundancy; in the Linux system, important logs are collected and classified and stored according to the set conditions with the help of the syslog mechanism for subsequent analysis. In terms of collecting application running data, data such as the number of connections and processing time are obtained through the application's own interface, key information is parsed by combining access and error logs, and stored in an integrated manner according to time to monitor the changes in the application running status. The collected data is cleaned and sorted into a unified format and then passed to the data encryption and upload module.

[0072] The data encryption and upload module encrypts the data using the Paillier homomorphic encryption algorithm and uses its additive homomorphic property to ensure data privacy and the feasibility of subsequent operations. Before uploading the encrypted data to the blockchain, it interacts with the smart contract for authorization determination. The smart contract comprehensively considers multiple factors such as the requester's credit score, business relevance, and system security risk level. The credit score is dynamically updated and recorded in a specific area of the blockchain according to the impact of data use on the accuracy of fault diagnosis and whether there are any illegal operations; the business relevance is calculated by a special function by analyzing the cooperation history and fault resolution situation; the system security risk level is based on the evaluation results of network threats and software vulnerabilities, and the smart contract regularly triggers the evaluation process and synthesizes multi-party data to obtain. Only when the result of the authorization decision function meets the threshold condition, the data is allowed to be uploaded to the blockchain storage and indexing module.

[0073] The block structure constructed by the blockchain storage and indexing module contains information such as encrypted data, hash value, and timestamp. The SHA-256 algorithm is used to calculate the hash value to ensure the immutability and traceability of the data. Based on the Merkle Patricia Trie established data indexing mechanism, with the data characteristics as the key and the storage location as the value, the efficient data retrieval function is realized with the help of the mapping structure in the smart contract, enabling the artificial intelligence diagnosis module to quickly obtain the required data.

[0074] After the artificial intelligence diagnosis module obtains the encrypted data from the blockchain, it first uses the Paillier private key to decrypt it. The decrypted data fragments are sequentially input into the convolutional neural network module, and the spatial features are extracted through the convolution layer, batch normalization layer, and ReLU activation function layer with specific parameter settings to identify the intrinsic correlation pattern of the data. Then the feature map sequence enters the recurrent neural network module, and the long short-term memory network is used to process the time series features, model the trend of data changes over time, and mine potential fault modes. After that, the attention mechanism module dynamically focuses on key information according to task requirements, highlighting the data parts related to fault diagnosis. Finally, the fault classification prediction module uses the fully connected layer and the softmax layer to calculate the probability of each fault category based on the previously extracted and processed feature information, and determine the most likely fault category.

[0075] The audit module works with other modules to record event information related to data upload, download and diagnostic operations, and defines structures and arrays in smart contracts to store audit logs. The logs are stored in the blockchain in a specific encryption method, and only the audit management module authenticated by the smart contract authorization rules can decrypt and view them. In this way, full audit monitoring of data use is achieved, ensuring that data flows are clear and traceable, used in compliance with regulations, timely discovering and preventing potential risks, and maintaining the safe and stable operation of the system.

[0076] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence detection device for computer system failures, characterized in that, Including: A data collection module, which is configured to collect data from multiple data sources of a computer system and has the functions of cleaning and sorting data; A data encryption and upload module, which is connected to the data collection module, encrypts the sorted data using a homomorphic encryption algorithm, and uploads the data to the blockchain after encryption; A blockchain storage and indexing module, which constructs a block structure for storing encrypted data on the blockchain and simultaneously establishes a data indexing mechanism based on a specific data structure; An artificial intelligence diagnosis module, which can obtain encrypted data from the blockchain and decrypt it; An audit module, which is responsible for recording event information related to data upload requests, download requests, and diagnostic operations.

2. The artificial intelligence detection device for computer system failures according to claim 1, characterized in that, The data collection module includes: a data collection unit, which simulates data blocks of a specific size for reading and writing operations to obtain the reading and writing speed when collecting memory reading and writing speed data; an operating system log collection unit, which extracts specific system core service-related logs when collecting operating system event logs in the system.

3. An artificial intelligence detection device for computer system failures according to claim 1, characterized in that, The data encryption and upload module includes: a data encryption unit, and the homomorphic encryption algorithm used has a specific encryption principle to implement data encryption; an authorization interaction unit, which decides data upload authorization according to the specific authorization decision logic in the smart contract.

4. An artificial intelligence detection device for computer system failures according to claim 1, characterized in that, The blockchain storage and indexing module includes: a blockchain storage unit, and the constructed block has a specific structure for storing encrypted data.

5. An artificial intelligence detection device for computer system failures according to claim 1, characterized in that, The artificial intelligence diagnosis module includes: a convolutional neural network unit, and the convolutional layer has specific parameter settings to implement the function of extracting data spatial features.

6. An artificial intelligence detection device for computer system failures according to claim 1, characterized in that, The audit module includes: an audit log recording unit, and the structure defined in the smart contract contains specific information for recording audit events, and stores audit log events through a specific array, and creates a corresponding instance in the smart contract function of the relevant operation and adds it to the array.

7. An artificial intelligence detection device for computer system failures according to claim 1, characterized in that, The data collection module further includes: an application data collection unit, which uses its internal performance monitoring interface to obtain data when collecting application operation data, and parses the access log to extract specific key information and combines it with error log information to form a comprehensive data record.

8. An artificial intelligence detection device for computer system failures according to claim 1, characterized in that, The artificial intelligence diagnosis module further includes: a recurrent neural network unit, and the long short-term memory network used processes data time series features according to specific calculation formulas.

9. An artificial intelligence detection device for computer system failures according to claim 1, characterized in that, The artificial intelligence diagnosis module further includes: a fault classification and prediction unit, which classifies and predicts faults based on a fully connected layer and a softmax layer.

10. An artificial intelligence detection device for computer system failures according to claim 1, characterized in that, The audit module further includes: a risk monitoring unit, which analyzes audit log data and monitors potential risk patterns of data access and use through a preset risk model and algorithm.

Images