Safe storage method and system for intelligent operation and maintenance data in credential field

Abstract

The invention discloses a secure storage method and system for intelligent operation and maintenance data in the credential field, and the method comprises the steps: deploying an end-to-end intelligent operation and maintenance all-in-one machine embedded with a GPU and an NPU, and monitoring a resource state in real time; encrypting the original operation and maintenance data by using an encryption engine, and analyzing memory access characteristics of the encrypted data in the neural network model through an NPU storage encryption engine performance evaluation platform; classifying the data into hot, warm and cold data according to memory access characteristics; a trusted execution environment based on trusted hardware is constructed to form a trust chain, and calculation tasks are dynamically scheduled according to resource states and memory access characteristics; and embedding life cycle labels containing sources and sensitivity into all encrypted data, and recording a full life cycle audit log. The system comprises a resource sensing and monitoring module, an encryption and memory access analysis module, an intelligent hierarchical storage module, an intelligent scheduling and trusted computing module and a data security management module. According to the method, safe storage, efficient processing and whole-course tracing of the operation and maintenance data in the credential environment are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of signal creation and intelligence calculation, and in particular to a signal creation field intelligent operation and maintenance data security storage method and system. BACKGROUND

[0002] With the comprehensive and in-depth promotion of signal creation industry, the intelligent operation and maintenance of its key infrastructure and core business systems are facing unprecedented data flood and severe security challenges. Intelligent operation and maintenance collects massive logs, indicators, tracking data, etc., and uses artificial intelligence models for real-time analysis and prediction to ensure system stability and optimize performance. However, the current intelligent operation and maintenance data storage and processing method in the signal creation environment has exposed many urgent problems in dealing with the dual demands of high-performance computing and data privacy security: First, modern intelligent operation and maintenance all-in-one machines generally use heterogeneous computing architecture composed of GPU and NPU. GPU is good at large-scale parallel computing, while NPU is good at neural network inference. However, existing systems often lack real-time sensing capability for GPU and NPU computing load and storage I / O pressure. This leads to rigid computing task scheduling and inability to dynamically and intelligently allocate resources according to real-time resource status. For example, in a high-concurrency analysis scenario, computationally intensive tasks may be inappropriately assigned to fully loaded NPU instead of idle GPU, causing computing power bottlenecks and task delays, and failing to fully utilize the collaborative potential of heterogeneous computing power.

[0003] Second, to meet the strict data security compliance requirements in the signal creation field, operation and maintenance data must be encrypted before storage. However, traditional encryption methods change the data access mode and locality characteristics after being introduced. Especially for NPU architecture optimized for plaintext data, the randomization characteristics of encrypted data can cause frequent cache invalidation and memory wall problems, resulting in a sharp decline in NPU computing efficiency.

[0004] Third, intelligent operation and maintenance data has significant access hotness differences. Real-time alarm data (hot data) with high-frequency access and historical archive data (cold data) with infrequent queries are stored together. Existing storage solutions use static or simple strategies for data classification, and cannot accurately respond to dynamic changes based on encrypted memory characteristics. The cache mechanism (such as traditional LRU) has low hit rate when facing multi-dimensional and complex access patterns, resulting in high-value hot data being unable to respond quickly, while the storage system is burdened with high storage costs and access delays for cold data.

[0005] Fourth, although the trusted execution environment (TEE) and homomorphic encryption technology make privacy computing possible, in the domestic environment of China's information technology, how to start from the hardware root trust, build a complete trust chain running through the hardware, operating system to the application layer, and seamlessly integrate the TEE with the blockchain, smart contract and other technologies to form a self-verifiable and trusted collaborative computing framework.

[0006] Fifth, the operation and maintenance data is widely sourced and has different sensitivities, and its storage, processing, sharing and destruction strategy should be dynamically adjusted. The current management method is extensive, lacks a fine-grained and data-bound life cycle tag system, which leads to the inability of security policy to accurately route and execute. At the same time, the whole chain operation of data from generation to destruction lacks tamper-proof audit logs, making it difficult to achieve fast and accurate traceability and responsibility when a security incident or compliance review occurs. SUMMARY

[0007] The purpose of the present application is to provide a smart operation and maintenance data security storage method and system in the field of China's information technology, to solve the problems of processing performance decline caused by data encryption and static storage strategy that cannot adapt to the dynamic access characteristics of data in the prior art.

[0008] To achieve the above-mentioned purpose, the technical scheme provided by the present application is: a smart operation and maintenance data security storage method in the field of China's information technology, comprising the following steps: S1, deploying an end-to-end smart operation and maintenance all-in-one machine, embedding GPU and NPU in the smart operation and maintenance all-in-one machine to form a smart computing core; monitoring the computing power load and storage I / O pressure of the GPU and NPU in real time, and sensing the resource state of the smart operation and maintenance all-in-one machine; S2, encrypting the original operation and maintenance data using an encryption engine, inputting the encrypted data into an NPU storage encryption engine performance evaluation platform based on STONNE and DRAMsim3 simulation framework, analyzing and obtaining the memory access characteristics of the encrypted data in the neural network model running process; S3, according to the memory access characteristics of the encrypted data, classifying it into hot data, warm data and cold data; wherein the hot data uses a uniform pool generation algorithm and an update algorithm to construct and maintain a cache pool, and the warm data and cold data use block storage technology for persistent storage and compression archiving; S4, building a trusted execution environment based on trusted hardware to form a trust chain; dynamically scheduling the computing task according to the resource state of the real-time smart operation and maintenance all-in-one machine and the memory access characteristics and classification of the encrypted data; S5, embedding a life cycle tag containing the source and sensitivity for all encrypted data; during the storage, processing and use of the encrypted data, routing it to the corresponding security module according to the tag, and recording the whole process audit log from data generation to destruction.

[0009] To optimize the above technical solution, the specific measures taken also include: In step S4, the trusted execution environment based on trusted hardware is constructed, specifically including: Using the trusted instruction embedded in the trusted CPU, a hardware root of trust from system startup to runtime is constructed; Through the Linux operating system, the integrity measurement function is started, forming a trust chain from hardware to the application layer; All integrity measurement values are stored in the trusted platform module, and the trusted state proof is provided externally through remote authentication technology.

[0010] Further, in step S4, the resource state of the real-time intelligent operation and maintenance all-in-one machine and the memory access characteristics and classification of encrypted data are used to dynamically schedule computing tasks, and the specific process is: When the computing power load is higher than the preset threshold, the computing-intensive task is preferentially scheduled to the GPU for execution; when the memory access characteristic shows that the locality of encrypted data access is high, the corresponding task is scheduled to the NPU configured with a large-capacity on-chip cache for execution; Further, for the joint computing task involving multiple parties sensitive data in the scheduling process, it is routed to the trusted execution environment, and homomorphic encryption and differential privacy cryptography technology are used for privacy calculation combined with smart contract and alliance chain.

[0011] In step S3, the uniform pool generation algorithm is specifically: The Euclidean distance between individuals in the population and reference points is calculated, and a distance matrix is stored; The distance matrix is sorted to obtain individual index and distance value; Through the de-duplication processing, the uniqueness of individuals in the uniform pool is ensured, and the individuals closest to each reference point are stored in the uniform pool.

[0012] Further, the uniform pool updating algorithm is specifically: The Euclidean distance between the newly generated individual and each reference point is calculated; The new distance value is compared with the distance value of the corresponding individual in the original uniform pool, and if the new distance is smaller, the original individual is replaced to realize the dynamic updating of the uniform pool.

[0013] Further, in step S3, the block storage technology is: Elasticsearch is used as a persistent log storage node, and is deployed in a three-replica cluster mode; A uniform pool is constructed in each log storage node as a cache mechanism, and data block lookup and prefetching are performed through digital fingerprints and distributed hash tables.

[0014] In step S2, the NPU storage encryption engine performance evaluation platform based on the STONNE and DRAMsim3 simulation framework refers to the NPU on-chip cache design of Eyeriss and ShiDianNao, and partitions the cache for management.

[0015] As another important technical solution, the application further provides an intelligent operation and maintenance data security storage system in the Xinyuan field, comprising: A resource perception and monitoring module is used for deploying an end-to-end intelligent operation and maintenance all-in-one machine, embedding a GPU and an NPU in the intelligent operation and maintenance all-in-one machine to form an intelligent computing core, monitoring the computing power load and storage I / O pressure of the GPU and the NPU in real time, and perceiving the resource state of the intelligent operation and maintenance all-in-one machine. An encryption and memory access analysis module is used for encrypting original operation and maintenance data by using an encryption engine, inputting the encrypted data into an NPU storage encryption engine performance evaluation platform based on the STONNE and DRAMsim3 simulation framework, analyzing and obtaining the memory access characteristics of the encrypted data in the running process of a neural network model. An intelligent scheduling and trusted computing module is used for constructing a trusted execution environment based on domestic hardware to form a trust chain, and dynamically scheduling a computing task according to the resource state of the real-time intelligent operation and maintenance all-in-one machine and the memory access characteristics. An intelligent hierarchical storage module is used for constructing a trusted execution environment based on trusted hardware to form a trust chain, and dynamically scheduling a computing task according to the resource state of the real-time intelligent operation and maintenance all-in-one machine, the memory access characteristics and classification of the encrypted data. A data security governance module is used for embedding a life cycle label containing a source and sensitivity into all encrypted data, routing the encrypted data to a corresponding security module according to the label during the storage, processing and use of the encrypted data, and recording an audit log of the whole process from data generation to destruction.

[0016] The application further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method for storing intelligent operation and maintenance data securely in the Xinyuan field when executing the computer program.

[0017] The application further provides a computer readable storage medium storing a computer program, wherein the computer program enables a computer to execute the method for storing intelligent operation and maintenance data securely in the Xinyuan field.

[0018] Compared with the prior art, the application has the following beneficial effects: The application constructs an intelligent operation and maintenance all-in-one machine with deep perception capability, deeply fuses the heterogeneous computing power of GPU and NPU, and realizes precise matching and dynamic balance between resources and demand by flexibly deploying computing tasks in real time according to the load of computing power and the storage pressure, so that the parallel computing flood of GPU and the special inference efficiency of NPU can be fully released, and sufficient and flexible computing power support is provided for complex data analysis and model operation in the upper layer.

[0019] The application introduces an advanced simulation evaluation platform, so that the system can accurately understand the micro access behavior of data in the NPU core runtime after the data is encrypted, and the key breakthrough makes the subsequent storage optimization no longer a blind decision, but a precise strategy based on a deep understanding of its inherent characteristics, thereby ensuring the confidentiality of the whole link of data and effectively ensuring the high-speed operation of the neural network processing engine.

[0020] The application distinguishes hot data, warm data and cold data according to their dynamic access characteristics. The hot data is active in the cache through a clever uniform pool algorithm to ensure immediate response, and the warm data and cold data are orderly guided to the hierarchical storage with high reliability and low cost. This storage strategy achieves the dual optimization of system response speed and storage economy as a whole.

[0021] The application constructs a complete trust chain running through the system layer and the application layer from the hardware root of trust, and places the privacy computing task under the strict protection of the trusted execution environment. By combining homomorphic encryption, differential privacy and other cryptographic technologies with the evidence storage capability of the blockchain, the system can safely complete the joint calculation of sensitive data of multiple parties while ensuring the availability and invisibility of the original data, and provides a trusted and compliant solution for operation and maintenance data analysis across security domains.

[0022] The application realizes intelligent routing and precise execution of security policies by embedding identity tags containing the source and sensitivity of each encrypted data. From the birth to the destruction of data, the flow, access and processing behavior of each link are clearly recorded to form an unalterable audit trail, which not only gives the system strong endogenous security attributes, but also realizes panoramic visualization and compliance control of the data life cycle, and provides a solid technical guarantee for coping with strict data security regulations. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 The flowchart of the application. DETAILED DESCRIPTION The above content of the present application will be further described in detail in the form of specific embodiments, but it should not be understood that the above-mentioned subject matter of the present application is limited to the following embodiments only, and any technology achieved based on the above-mentioned content of the present application falls within the scope of the present application.

[0024] Some noun explanations in the present application are as follows: GPU refers to the abbreviation of Graphics Processing Unit in the field of Computer Graphics and general computing. GPU is a hardware device specially designed for processing graphics and image computing, which is responsible for accelerating graphics rendering and image processing tasks in computers. GPU and central processing unit (CPU) complement each other and together form the core components of a computer, but they have great differences in the way and characteristics of processing tasks.

[0025] Data assets refer to data resources that a specific subject legally owns or controls and can bring economic benefits, including the following characteristics: data assets are resources that record data in physical or electronic form, such as digital information, text information, image information, etc.; data assets are assets that are legally owned or controlled by a specific subject and can bring economic benefits in the future; data assets are factors, like land, capital, labor, and technology, which are important factors in production and business activities.

[0026] Information technology application innovation industry, that is, information technology application innovation industry, aims to realize the independent controllability of information technology, avoid external technology sanctions and risks, and promote the localization of the whole industry chain. The information technology application innovation industry is the foundation of data security and network security, and is also an important part of new infrastructure. It involves multiple fields, including IT infrastructure (such as CPU chips, servers, storage, switches, routers, etc.), basic software (such as databases, operating systems, middleware), and application software (such as OA systems, ERP systems, office software, government applications, etc.). In addition, the information technology application innovation industry also includes information security-related products and services, such as border security products and terminal security products.

[0027] In some embodiments, as shown in Figure 1 The present application provides a data security storage method for intelligent operation and maintenance in the field of information technology application innovation, which comprises the following steps: S1, deploying an end-to-end intelligent operation and maintenance all-in-one machine, embedding GPU and NPU in the intelligent operation and maintenance all-in-one machine to form a smart computing core; monitoring the computing power load and storage I / O pressure of GPU and NPU in real time, and sensing the resource state of the intelligent operation and maintenance all-in-one machine; In some embodiments, after embedding the end-to-end intelligent operation all-in-one machine with GPU and NPU, the GPU improves the task execution efficiency and throughput by virtue of parallel computing capability, the NPU focuses on neural network inference tasks, and the two work together to support the intelligent calculation and analysis application of large models on the operation of business data in the all-in-one machine, solving the problem of few intelligent calculation measurement methods in this scenario. It provides basic computing power support for subsequent steps, ensures that the subsequent privacy calculation module safely processes intelligent calculation data, and the NPU storage encryption engine evaluates the memory access characteristics of the encrypted neural network, ensuring that subsequent data security and performance evaluation are efficiently carried out.

[0028] S2, encrypt the original operation and maintenance data using the encryption engine, input the encrypted data into the NPU storage encryption engine performance evaluation platform based on the STONNE and DRAMsim3 simulation framework, analyze and obtain the memory access characteristics of the encrypted data during the operation of the neural network model; In some embodiments, the encryption engine uses national encryption algorithms (such as SM4) or international general encryption algorithms (such as AES) to ensure the confidentiality of data during transmission and storage. The NPU storage encryption engine performance evaluation platform based on the STONNE and DRAMsim3 simulation framework refers to the on-chip cache design of Eyeriss and ShiDianNao NPU, and partitions the cache for management.

[0029] Eyeriss divides the cache into 25 banks in total, and each layer exclusively occupies the bank resources during execution; ShiDianNao divides the cache into three parts according to input, filter, and psum to achieve efficient data access and management.

[0030] S3, according to the memory access characteristics of the encrypted data, it is classified into hot data, warm data and cold data; among them, the hot data uses the uniform pool generation algorithm and the update algorithm to construct and maintain the cache pool, and the warm data and cold data use block storage technology for persistent storage and compression archiving; In some embodiments, the memory access characteristics include access frequency, access latency, data locality, etc. The classification standard sets a threshold according to the access frequency. When the access frequency is higher than the set upper limit, it is hot data, lower than the set lower limit, it is cold data, and between the two, it is warm data.

[0031] The uniform pool generation algorithm is as follows: Calculate the Euclidean distance between individuals in the population and reference points, and store the distance matrix; Sort the distance matrix to obtain the individual index and distance value; Ensure the uniqueness of individuals in the uniform pool through deduplication processing, and store the individuals closest to each reference point in the uniform pool.

[0032] As preferred, the uniform pool updating algorithm specifically comprises: calculating the Euclidean distance between the newly generated individual and each reference point; comparing the new distance value with the distance value of the corresponding individual in the original uniform pool, and replacing the original individual if the new distance is smaller, to realize dynamic updating of the uniform pool.

[0033] In some embodiments, the hot data and the cold data are stored and compressed in archive using block storage technology, specifically comprising: Elasticsearch is used as a persistent log storage node, deployed in a three-replica cluster mode, including full-text search function, distributed real-time file storage, and indexing of each field to enable searching. It is also a real-time analysis distributed search engine that can be extended to hundreds of servers to handle PB-level structured or unstructured data; A uniform pool constructed by an algorithm is deployed in each log storage node, including two algorithms of uniform pool generation and uniform pool updating. The uniform pool is equivalent to a cache, making the search space of the log storage node larger, ensuring that larger data fragments are discovered and placed in the cache.

[0034] A uniform pool is constructed in each log storage node as a cache mechanism, and data block lookup and prefetching are performed through digital fingerprints and a distributed hash table.

[0035] In some embodiments, when the cache is not hit, the digital fingerprint of the target data block is first searched in the distributed hash table (DHT): first, calculate the Euclidean distance between each unmarked individual and each digital fingerprint in the DHT, and filter the individuals according to the distance —— store the individual closest to each digital fingerprint in the uniform pool through the uniform pool generation algorithm, if there are multiple unmarked individuals with the same distance to a digital fingerprint, select the individual closest to the ideal point to enter the pool. If the DHT search hits, add the locality list corresponding to the digital fingerprint stored in the DHT to the cache to complete the prefetching, and the subsequent processing is consistent with the cache hit; if the DHT still does not hit, there is a false positive error in the fast new data block filtering module, and when the cache is full, the uniform pool is updated through the LRU cache replacement strategy, and the locality list of the replaced digital fingerprint is synchronized to the DHT, while ensuring that the locality list only affects prefetching and does not interfere with the correctness of the index service.

[0036] S4, a trusted execution environment based on trusted hardware is constructed to form a trust chain; and a computing task is dynamically scheduled according to the resource state of the real-time intelligent operation and maintenance all-in-one machine and the access characteristics and classification of encrypted data. In some embodiments, the construction of the trusted execution environment based on trusted hardware specifically comprises: The trusted chain based on the megachip CPU is constructed at system startup and runtime by using the megachip ZX-TCT trusted computing technology, and the trust relationship is transmitted from the megachip CPU embedded instruction to the OS layer; meanwhile, the IMA integrity measurement function is enabled for the Linux operating system, and the trust chain is further extended to the application layer, that is, the privacy computing framework layer; and all the integrity measurement values are stored in the domestic TPM2.0 hardware device to ensure the trust and non-tamperability; finally, a trusted execution environment is more controllably constructed through remote authentication technology, and the trusted computing of the upper application is realized.

[0037] In step S4, the computing task is dynamically scheduled according to the resource state of the real-time intelligent operation and maintenance all-in-one machine and the memory access characteristics and classification of the encrypted data, and the specific process is as follows: When the computing power load is higher than the preset threshold, the computing-intensive task is preferentially scheduled to the GPU for execution; when the memory access characteristic shows that the locality of encrypted data access is high, the corresponding task is scheduled to the NPU configured with a large-capacity on-chip cache for execution; As preferred, for the joint computing task involving multi-party sensitive data in the scheduling process, the joint computing task is routed to the trusted execution environment, and the homomorphic encryption and differential privacy cryptography technology are combined with the smart contract and the alliance chain to perform privacy computing.

[0038] S5, embedding a life cycle label containing a source and a sensitivity into all encrypted data; in the storage, processing and use process of the encrypted data, the data is routed to the corresponding security module according to the label, and the whole process audit log from data generation to destruction is recorded.

[0039] In some embodiments, the life cycle label includes fields such as data source, sensitivity level, creation time, expiration time, access permission, etc. According to the label content, the system automatically routes the data to the corresponding encryption module, access control module or desensitization processing module. At the same time, the system records the whole link audit log from data generation, transmission, processing to destruction, ensuring the traceability and compliance of the data life cycle.

[0040] In another embodiment, the present application provides an intelligent operation and maintenance data security storage system in the field of Xingcheng, which comprises: A resource perception and monitoring module is used for deploying an end-to-end intelligent operation and maintenance all-in-one machine, embedding a GPU and an NPU in the intelligent operation and maintenance all-in-one machine to form a wisdom calculation core, monitoring the computing power load and storage I / O pressure of the GPU and the NPU in real time, and perceiving the resource state of the intelligent operation and maintenance all-in-one machine; The encryption and memory access analysis module is used for encrypting the original operation and maintenance data by using an encryption engine, inputting the encrypted data into an NPU memory encryption engine performance evaluation platform based on STONNE and DRAMsim3 simulation framework, and analyzing and obtaining the memory access characteristics of the encrypted data in the running process of the neural network model. The intelligent scheduling and trusted computing module is used for constructing a trusted execution environment based on domestic hardware, forming a trust chain, and dynamically scheduling computing tasks according to the resource state of the real-time intelligent operation and maintenance all-in-one machine and the memory access characteristics. The intelligent hierarchical storage module is used for constructing a trusted execution environment based on trusted hardware, forming a trust chain, and dynamically scheduling computing tasks according to the resource state of the real-time intelligent operation and maintenance all-in-one machine and the memory access characteristics and classification of the encrypted data. The data security governance module is used for embedding a life cycle label containing a source and a sensitivity into all encrypted data, routing the encrypted data to a corresponding security module according to the label during storage, processing and use of the encrypted data, and recording an audit log of the whole process from data generation to destruction.

[0041] In another embodiment, the application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for secure storage of intelligent operation and maintenance data in the field of signal creation.

[0042] In another embodiment, the application provides a computer readable storage medium storing a computer program, wherein the computer program enables a computer to execute the method for secure storage of intelligent operation and maintenance data in the field of signal creation.

[0043] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium, which can contain or store programs for use by or in conjunction with an instruction execution system, device or apparatus. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or apparatus, or any suitable combination of the above. More specific examples of computer storage media can include one or more wires, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0044] The above merely describes preferred embodiments of the present application, and is not intended to limit the present application in any form. Any modification, equivalent replacement, and improvement of the above embodiments made by any person skilled in the art, without departing from the technical solution of the present application, according to the technical essence of the present application, are still within the protection scope of the present application.

Claims

1. A method for secure storage of intelligent operation and maintenance data in the field of information technology innovation, characterized in that, Includes the following steps: S1. Deploy an end-to-end intelligent operation and maintenance integrated machine, embedding GPU and NPU in the intelligent operation and maintenance integrated machine to form an intelligent computing core; monitor the computing load and storage I / O pressure of GPU and NPU in real time, and perceive the resource status of the intelligent operation and maintenance integrated machine. S2. Use the encryption engine to encrypt the original operation and maintenance data, and input the encrypted data into the NPU storage encryption engine performance evaluation platform based on the STONNE and DRAMsim3 simulation framework to analyze and obtain the memory access characteristics of the encrypted data during the operation of the neural network model. S3. Based on the access characteristics of encrypted data, it is classified into hot data, warm data and cold data; hot data uses a uniform pool generation algorithm and update algorithm to build and maintain a cache pool, while warm data and cold data use block storage technology for persistent storage and compressed archiving. S4. Construct a trusted execution environment based on trusted hardware to form a chain of trust; Based on the resource status of the real-time intelligent operation and maintenance integrated machine and the access characteristics and classification of encrypted data, computing tasks are dynamically scheduled. S5. Embed a lifecycle tag containing source and sensitivity for all encrypted data; during the storage, processing and use of encrypted data, route it to the corresponding security module according to the tag, and record the entire process audit log from data generation to destruction.

2. The method for secure storage of intelligent operation and maintenance data in the field of information technology innovation according to claim 1, characterized in that: In step S4, the construction of a trusted execution environment based on trusted hardware specifically includes: By leveraging trusted instructions embedded within a trusted CPU, a hardware root of trust can be built from system startup to runtime. By enabling integrity measurement functionality through the Linux operating system, a trust chain is formed from the hardware to the application layer. All integrity metrics are stored in the trusted platform module, and trusted status proofs are provided externally through remote authentication technology.

3. The method for secure storage of intelligent operation and maintenance data in the field of information technology innovation according to claim 1, characterized in that: In step S4, the dynamic scheduling of computing tasks based on the resource status of the real-time intelligent operation and maintenance integrated machine and the access characteristics and classification of encrypted data is carried out as follows: When the computing load exceeds a preset threshold, computationally intensive tasks will be prioritized for execution on the GPU; when the memory access characteristics indicate high locality of encrypted data access, the corresponding task will be scheduled for execution on an NPU equipped with a large-capacity on-chip cache. For joint computation tasks involving sensitive data from multiple parties during the scheduling process, they are routed to the trusted execution environment, and homomorphic encryption and differential privacy cryptography are used in combination with smart contracts and consortium blockchains for privacy computation.

4. A method for secure storage of intelligent operation and maintenance data in the field of information technology innovation according to claim 1, characterized in that: In step S3, the uniform pool generation algorithm is specifically as follows: Calculate the Euclidean distance between individuals in the population and the reference point, and store the distance matrix; Sort the distance matrix and obtain the individual index and distance value; The uniqueness of individuals in the uniform pool is ensured by deduplication, and the individuals closest to each reference point are stored in the uniform pool.

5. A method for secure storage of intelligent operation and maintenance data in the field of information technology innovation according to claim 4, characterized in that: The uniform pooling update algorithm is specifically as follows: Calculate the Euclidean distance between the newly generated individual and each reference point; The new distance value is compared with the distance value of the corresponding individual in the original uniform pool. If the new distance is smaller, the original individual is replaced, thus realizing the dynamic update of the uniform pool.

6. A method for secure storage of intelligent operation and maintenance data in the field of information technology innovation according to claim 1, characterized in that: In step S3, the block storage technology is: Elasticsearch is used as the persistent log storage node and deployed in a three-replica cluster. A uniform pool is built in each log storage node as a caching mechanism, and data block lookup and prefetching are performed using digital fingerprints and distributed hash tables.

7. A method for secure storage of intelligent operation and maintenance data in the field of information technology innovation according to claim 1, characterized in that: In step S2, the NPU storage encryption engine performance evaluation platform based on the STONNE and DRAMsim3 simulation framework refers to the NPU on-chip cache design of Eyeriss and ShiDianNao and performs cache partitioning management.

8. A secure data storage system for intelligent operation and maintenance in the field of information technology innovation, characterized in that, include: The resource awareness and monitoring module is used to deploy an end-to-end intelligent operation and maintenance all-in-one machine, in which a GPU and an NPU are embedded to form an intelligent computing core. Real-time monitoring of GPU and NPU computing load and storage I / O pressure, and perception of the resource status of the intelligent operation and maintenance all-in-one machine; The encryption and memory access analysis module is used to encrypt the original operation and maintenance data using the encryption engine, input the encrypted data into the NPU storage encryption engine performance evaluation platform based on the STONNE and DRAMsim3 simulation framework, and analyze and obtain the memory access characteristics of the encrypted data during the operation of the neural network model. The intelligent hierarchical storage module is used to classify encrypted data into hot data, warm data, and cold data according to the access characteristics of the data. Hot data uses a uniform pool generation algorithm and an update algorithm to build and maintain a cache pool, while warm data and cold data use block storage technology for persistent storage and compressed archiving. The intelligent scheduling and trusted computing module is used to build a trusted execution environment based on domestically produced hardware, forming a chain of trust. Based on the resource status and memory access characteristics of the real-time intelligent operation and maintenance integrated machine, computing tasks are dynamically scheduled. The data security governance module is used to embed lifecycle tags containing source and sensitivity information into all encrypted data; during the storage, processing and use of encrypted data, it routes the data to the corresponding security module according to the tags and records the entire process audit log from data generation to destruction.

9. An electronic device, characterized in that, include: The present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements a method for secure storage of intelligent operation and maintenance data in the field of information technology innovation as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program causes the computer to execute a method for secure storage of intelligent operation and maintenance data in the field of information technology innovation as described in any one of claims 1 to 7.

Images