Cloud data archiving method and system

Abstract

The invention relates to the technical field of data processing, and provides a cloud data archiving method and system, and the method comprises the steps: obtaining a source data set of to-be-archived data, carrying out the grade division, obtaining a data grade label, dividing the source data set into a plurality of data blocks, carrying out the data encryption of different grades, obtaining a plurality of encrypted data blocks, and storing the encrypted data blocks in a database; analyzing and coding to obtain a multi-dimensional data storage structure, uploading the multi-dimensional data storage structure to a preset cloud storage node, and generating a storage index information set based on multiple data blocks in the uploading process; and when a data recovery request is received, acquiring the data block from the cloud storage node according to the storage index information set, and performing decryption and data recombination. Through the scheme, classification, hierarchical encryption and structured archiving of the source data set are realized, the security and management efficiency of the data in a cloud storage process are improved, and the problem that the encryption level cannot be dynamically adjusted by data sensitivity or content features when a large-scale heterogeneous data source is used is solved.

Description

Technical Field

[0001] The present application relates to the technical field of data processing, and in particular to a cloud data archiving method and system. Background Art

[0002] With the rapid development of technologies like big data, artificial intelligence, and cloud computing, businesses and institutions are generating massive amounts of structured and unstructured data during their operations. To reduce local storage costs and improve data access efficiency and disaster recovery capabilities, more and more organizations are migrating their data to cloud storage platforms. Cloud data archiving, a key component of this process, not only ensures long-term data preservation but also ensures data integrity, security, and recoverability.

[0003] Among related technical approaches, a common practice is to compress and convert raw data files to a standardized format before archiving, and then encrypt the archived data entirely using a static encryption algorithm (such as AES or RSA). The encrypted data is uploaded to a cloud storage server, and a unique file identifier or metadata index is generated and stored in a local or external database for future on-demand query or recovery. This achieves the goals of centralized data storage, conserving local storage resources, and improving disaster recovery capabilities.

[0004] Regarding the above technical solutions, although basic data security protection and retrieval capabilities can be achieved through unified encryption and index management, when faced with large-scale heterogeneous data sources, the existing fixed or coarse-grained encryption mechanisms are usually used, which makes it difficult to dynamically adjust the encryption level according to data sensitivity or content characteristics, resulting in excessive use of security computing resources for low-sensitivity data resources, and there is a risk of leakage of highly sensitive data when accessed or transmitted. Summary of the Invention

[0005] In order to improve the problem that when using large-scale heterogeneous data sources, the encryption level cannot be dynamically adjusted according to data sensitivity or content characteristics, resulting in excessive use of security computing resources for low-sensitivity data resources and the risk of leakage of highly sensitive data when accessed or transmitted, the present application provides a cloud data archiving method and system.

[0006] The present invention provides a cloud data archiving method, comprising: obtaining a source data set of data to be archived, hierarchically dividing the source data set using a preset data grading strategy to obtain data grade labels, and dividing the source data set into a plurality of data blocks using the data grade labels; performing different levels of data encryption on the corresponding data blocks based on the data grade labels to obtain a plurality of encrypted data blocks, analyzing and encoding all the encrypted data blocks to obtain a first data summary set and a second feature vector set; generating a clustering label set based on the first data summary set, and constructing a multidimensional data storage structure using the clustering label set and the second feature vector set; uploading the multidimensional data storage structure to a preset cloud storage node, generating a storage index information set based on the data blocks in the multidimensional data storage structure during the uploading process, and storing the storage index information set in a local index server; when a data recovery request is received, obtaining the data blocks in the multidimensional data storage structure from the cloud storage node according to the storage index information set of the data recovery request, and performing decryption and data reorganization to recover the source data set.

[0007] As a preferred solution, the steps of obtaining a source data set of data to be archived, hierarchically classifying the source data set using a preset data classification strategy to obtain data class labels, and dividing the source data set into a number of data blocks using the data class labels include: performing format recognition and structural analysis on the source data set to extract data subsets, performing content analysis and logical grouping on the data subsets to obtain data block groups and meta-tag information; performing rule matching and priority evaluation on the meta-tag information using a preset data classification strategy to obtain candidate levels, cross-checking the candidate levels with the data block groups to generate data class labels, and using the data class labels to divide the source data set into a number of data blocks.

[0008] As a preferred solution, the steps of performing format recognition and structural analysis on the source data set, extracting data subsets, performing content analysis and logical grouping on the data subsets, and obtaining data block groups and meta-tag information include: performing format analysis on the source data set, identifying file types and data structures, and using the file types and data structures to divide the data in the source data set into data subsets of text, images, and audio; performing content analysis on each of the data subsets to obtain metadata, using the metadata to construct meta-tag information, and logically grouping the data subsets through the meta-tag information to generate data block groups.

[0009] As a preferred solution, the steps of performing different levels of data encryption on the corresponding data blocks based on the data level labels to obtain several encrypted data blocks, analyzing and encoding all the encrypted data blocks to obtain a first data summary set and a second feature vector set include: using the data level labels to match the preset encryption algorithm type and key strength for the corresponding data blocks, encrypting the corresponding data blocks using the encryption algorithm type and the key strength to obtain several encrypted data blocks, and generating corresponding encryption summary information for each encrypted data block; wherein the encryption summary information includes a hash summary and key identification information; merging all the encrypted data blocks with their corresponding encryption summary information to generate intermediate encrypted data, performing semantic feature analysis and context vector encoding on the intermediate encrypted data to obtain a feature vector group and a structural summary; combining the hash summary and the key identification information with the structural summary to generate a first data summary set, and fusing the feature vector group with the meta tag information to generate a second feature vector set.

[0010] As a preferred solution, the steps of generating a clustering label set based on the first data summary set and constructing a multidimensional data storage structure using the clustering label set and the second feature vector set include: constructing an initial index comparison table based on the hash summary and key identifier of each data block in the first data summary set, performing similarity analysis on the feature vectors in the initial index comparison table to obtain similarity results for each data block; wherein the feature vector refers to a vector representing the semantics and structure of each data block extracted based on the content; assigning the data blocks to different clustering groups based on the similarity results to obtain a clustering label set, fusing the clustering label set with the second feature vector set to construct a multidimensional index vector group, performing dimensionality reduction processing and spatial mapping conversion on the multidimensional index vector group to obtain index partitions; and constructing a multidimensional data storage structure using the index partitions and clustering labels.

[0011] As a preferred solution, the steps of uploading the multidimensional data storage structure to a preset cloud storage node, generating a storage index information set based on the data blocks in the multidimensional data storage structure during the uploading process, and storing the storage index information set in a local index server include: extracting the clustering labels and key identifiers of each data block in the multidimensional data storage structure, and using the clustering labels of the data blocks to calculate the redundant distribution level, generating an upload scheduling plan and a replica position relationship table for each data block according to the redundant distribution level; combining all the upload scheduling plans with their corresponding key identifiers to generate an encrypted upload metadata set; in the process of uploading each data block to the preset cloud storage node, generating a storage index record using the encrypted upload metadata set corresponding to each data block; merging the storage index record with the replica position relationship table corresponding to the data block to obtain a storage index information set, and writing the storage index information set to the local index server.

[0012] As a preferred solution, the step of obtaining the data blocks in the multidimensional data storage structure from the cloud storage node according to the storage index information set of the data recovery request, and performing decryption and data reorganization to recover the source data set includes: parsing the storage node and key identifier of the data block according to the storage index information set of the data recovery request, sending a data block request instruction to the cloud storage node based on the storage node, and obtaining the data blocks and feature vector group in the multidimensional data storage structure; using the key identifier to call the local key service to obtain the authorization key to obtain the key, decrypting the data block by using the key to obtain the decrypted data block; reconstructing the data level and sorting the data on the decrypted data block to generate a reorganized data linked list, performing semantic restoration and content reorganization on the reorganized data linked list and the feature vector group to recover the source data set.

[0013] The present application also provides a cloud data archiving method and system device, including: an acquisition module, used to obtain a source data set of data to be archived, classify the source data set using a preset data classification strategy to obtain a data class label, and use the data class label to divide the source data set into a number of data blocks; an encryption module, used to perform different levels of data encryption on the corresponding data blocks based on the data class label to obtain a number of encrypted data blocks, analyze and encode all the encrypted data blocks to obtain a first data summary set and a second feature vector set; a construction module, used to generate a clustering label set based on the first data summary set, and use the clustering label set and the second feature vector set to construct a multidimensional data storage structure; a storage module, used to upload the multidimensional data storage structure to a preset cloud storage node, generate a storage index information set based on the data blocks in the multidimensional data storage structure during the upload process, and store the storage index information set in a local index server; a recovery module, used to, when receiving a data recovery request, obtain the data blocks in the multidimensional data storage structure from the cloud storage node according to the storage index information set of the data recovery request, and perform decryption and data reorganization to recover the source data set.

[0014] Compared with the prior art, the present application has the following beneficial effects: strong security and high flexibility. After obtaining the source data set of the data to be archived, it is divided into several data blocks, and the data blocks are encrypted at different levels to generate several encrypted data blocks. All the encrypted data blocks are analyzed and encoded to obtain a first data summary set and a second feature vector set. Through the bidirectional feature construction method of the first data summary set and the second feature vector set, the data has stronger context relevance and content consistency during subsequent retrieval, classification and aggregation recovery. Then, a clustering label set is constructed through hash summary similarity analysis. During the upload process, with the help of upload scheduling plans, replica relationship tables and local index records, the reliable storage and traceability management of data in the cloud are effectively guaranteed, and refined classification, hierarchical encryption and structured archiving of the source data set are achieved, thereby improving the security and management efficiency of data in the cloud storage process, and improving the problem that in large-scale heterogeneous data sources, the encryption level cannot be dynamically adjusted according to data sensitivity or content characteristics, resulting in excessive use of security computing resources for low-sensitivity data resources, and the risk of leakage of highly sensitive data when accessed or transmitted. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] The structures, proportions, sizes, etc. depicted in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not intended to limit the conditions under which the present invention can be implemented and therefore have no substantive technical significance. Any structural modifications, changes in proportional relationships, or adjustments in size should still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and objectives that can be achieved by the present invention.

[0017] Figure 1 1 is a flow chart of a cloud data archiving method provided by an embodiment of the present invention; Figure 2 It is a schematic block diagram of the structure of the cloud data archiving system provided by an embodiment of the present invention.

[0018] Description of reference numerals: 10. Cloud data archiving system; 11. Acquisition module; 12. Encryption module; 13. Construction module; 14. Storage module; 15. Recovery module. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments derived by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0020] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor do they necessarily have to be performed in the order described. For example, some operations / steps may be decomposed, combined, or partially merged, so the actual order of execution may vary depending on the actual situation.

[0021] It should also be understood that the terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit the present application. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0022] It should be further understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0023] The technical solution of the present invention will be further described below with reference to the accompanying drawings and through specific implementation methods.

[0024] Example 1: like Figure 1 As shown, the present application provides a cloud data archiving method, including steps S100 to S500.

[0025] Step S100: Obtain a source data set of data to be archived, classify the source data set using a preset data classification strategy to obtain a data class label, and divide the source data set into several data blocks using the data class label.

[0026] In this step, the source data set to be archived is first obtained from the user's local or edge device. A format recognition engine is then used to parse the source data set, identifying its file type (e.g., .docx, .jpg, .mp3) and structural information (e.g., paragraphs, image metadata, and audio waveform structure). Based on the recognition results, the source data set is divided into several data subsets, such as text, image, and audio. A content parsing algorithm is then used to extract metadata such as semantics, tags, and timestamps from each data subset. Meta-tag information is then constructed using a rule-matching system and a priority calculation module. A pre-defined data classification strategy (e.g., based on the ISO / IEC 27001 classification model) is then used to match the meta-tag information and generate candidate level tags. Finally, through the cross-mapping relationship between the candidate levels and the data subsets, data level tags are generated and the source data set is logically segmented into multiple data blocks with security level identifiers.

[0027] For example, a source data set contains company meeting minutes, employee photos, and audio recordings. The system first identifies the document type as .docx, the photos as .jpg, and the audio as .wav. The document contains the keyword "company strategy," which matches the confidential level. The photos are classified as general data. The audio content, after speech transcription, is identified as an internal discussion and is classified as restricted. The resulting three data level labels are "high," "low," and "medium," which are then used to divide the data into three blocks.

[0028] Step S200: encrypt the corresponding data blocks at different levels based on the data level labels to obtain a number of encrypted data blocks, analyze and encode all the encrypted data blocks to obtain a first data summary set and a second feature vector set.

[0029] In this step, the system searches the pre-set encryption policy library based on the data-level label corresponding to each data block, matches the encryption algorithm type (e.g., AES, RSA, SM4, etc.) and key strength (e.g., 128-bit, 256-bit, 512-bit, etc.), calls the local encryption module to complete the corresponding data encryption operation, and outputs the encrypted data block. Specifically, each encrypted data block generates a corresponding encryption summary, including a hash digest (using SHA-256) and a key identifier (e.g., in UUID format). Next, all encrypted data blocks are combined with the corresponding encryption summary information to generate intermediate encrypted data. The system performs semantic feature extraction (using the BERT pre-trained model) and context vector encoding (e.g., a TF-IDF + Word2Vec combined model) on the intermediate encrypted data to form a feature vector group and structural summary information. Finally, the encryption summary (hash and key ID) and structural summary are combined to form the first data summary set, and the feature vector group is merged with the aforementioned meta-tag information to form the second feature vector set.

[0030] For example, high-level data blocks are encrypted using AES-256 to generate a hash summary value H1, with the key ID K1; mid-level data blocks are encrypted using SM4 to generate a summary H2, with the key ID K2; these encrypted data blocks are sent to the semantic analyzer to extract keyword vectors such as "finance" and "personnel changes" to form feature vector groups V1 and V2, which are then merged with the original meta tags such as "confidential" and "internal" to form a second feature vector set.

[0031] Step S300: Generate a cluster label set according to the first data summary set, and construct a multidimensional data storage structure using the cluster label set and the second feature vector set.

[0032] In this step, an initial index comparison table is constructed to associate the hash digest of each data block in the first data digest set with its key identifier. Specifically, the MinHash method and Jaccard similarity calculation are used to perform pairwise similarity measurement on all hash digests. The similarity calculation formula is: in, and The system then merges the cluster labels with the second set of eigenvectors to construct a multidimensional index vector set. Principal component analysis (PCA) is used to reduce the vector dimensions to a manageable range (e.g., 256 dimensions), and t-SNE is used for spatial mapping. Finally, the data index is partitioned according to spatial distribution, creating a complete multidimensional data storage structure.

[0033] For example, the summary similarity of two data blocks D1 and D2 is 0.82, and they are classified into the same cluster C1; their feature vectors are V1=[0.12,0.86] and V2=[0.10,0.88] respectively. After fusion and dimensionality reduction, they are constructed into a 128-dimensional multidimensional index structure M1, which is mapped into the storage partition corresponding to the cluster label C1.

[0034] Step S400: Upload the multidimensional data storage structure to a preset cloud storage node, generate a storage index information set based on the data blocks in the multidimensional data storage structure during the uploading process, and store the storage index information set in a local index server.

[0035] In this step, the clustering label and key identifier corresponding to each data block in the multidimensional data storage structure are extracted, and the redundant distribution level is calculated based on the clustering label (which can be set according to the data level, such as 3 copies for high level and 2 copies for medium level). Specifically, an upload schedule (including node number, upload time, and concurrency level) and a replica location relationship table (recording which nodes the data block is replicated) are generated based on the redundant distribution level. The schedule is bound to the key identifier to form an encrypted upload metadata set. During the upload process, a storage index record (including hash value, location, timestamp, key ID, etc.) is automatically generated for each data block upload event, and is merged with the replica location relationship table to generate a complete storage index information set. Finally, the storage index information set is written to the local index server for subsequent data recovery calls.

[0036] For example, data block D1 is marked as "high level" and needs to be stored in Node1, Node3, and Node5; a scheduling plan U1 and a replica table R1 are generated, and an index record I1={hash:H1,key:K1,nodes:[1,3,5],timestamp:T} is formed during the upload process and finally written to the local server.

[0037] Step S500: When a data recovery request is received, data blocks in the multidimensional data storage structure are obtained from the cloud storage node according to the storage index information set of the data recovery request, and decrypted and reorganized to restore the source data set.

[0038] In this step, the system first receives a recovery request and extracts the storage index information set, parsing the target data block's storage location and key identifier. Specifically, based on the storage node list, the system initiates a pull request to the cloud storage node, retrieving the data block and its accompanying feature vector group. The key identifier is then used to call the corresponding key from the local key service for decryption. After decryption, the system logically sorts and semantically reconstructs the data blocks based on the original data's hierarchical and structural labels. Combined with the feature vector group, the system performs contextual content restoration and sequence verification to reconstruct the complete source data set.

[0039] For example, the document block D1 requested for recovery is available in Node1, and the corresponding key ID is K1. After retrieving the key, the system decrypts the original text block, restores the original paragraph order through the sorting weight, and restores the title content through the "Strategic Meeting" semantic information contained in V1, completely reproducing the source data set.

[0040] In this embodiment, by obtaining the source data set of the data to be archived, the source data set is graded using a preset data classification strategy to generate data grade labels, and the source data set is divided into several data blocks based on the data grade labels; then, different levels of data encryption are performed on the corresponding data blocks based on the data grade labels to generate several encrypted data blocks, and all the encrypted data blocks are analyzed and encoded to obtain a first data summary set and a second feature vector set; then, a clustering label set is generated based on the first data summary set, and a multidimensional data storage structure is constructed by combining the clustering label set and the second feature vector set; then, the multidimensional data storage structure is uploaded to a preset cloud storage node, and during the uploading process, a storage index information set is generated based on the data blocks in the multidimensional data storage structure, and the storage index information set is stored in a local index server; finally, when a data recovery request is received, the data blocks in the multidimensional data storage structure are obtained from the cloud storage node based on the storage index information set of the data recovery request, and decryption and data reorganization are performed to restore the source data set.

[0041] Through the above technical solution, it is possible to achieve refined classification, hierarchical encryption and structured archiving of source data sets, thereby significantly improving the security and management efficiency of data in the cloud storage process; the generated first data summary set and second feature vector set not only improve the intelligence of data retrieval and clustering, but also enhance the content association and semantic restoration capabilities of data during the recovery process; the index generation mechanism in the construction and upload process of the multi-dimensional data storage structure effectively supports efficient positioning and copy management of data; in the data recovery stage, combined with the local index server and decryption process, it can ensure the integrity, controllable and reliable recovery of the source data set, significantly improving the availability and security of the cloud data archiving system, and improving the problem that in the case of large-scale heterogeneous data sources, the encryption level cannot be dynamically adjusted according to data sensitivity or content characteristics, resulting in excessive use of security computing resources for low-sensitivity data resources, and there is a risk of leakage of highly sensitive data when accessed or transmitted.

[0042] Example 2: In step S100, the source data set is format identified and structure parsed to extract a data subset, which is then content analyzed and logically grouped to obtain a data block group and meta tag information.

[0043] Through format recognition and structural analysis, the system identifies various types of data within the source data set. Specifically, the system first parses the source data set to identify file types (such as documents, images, and audio) and data structures (such as paragraphs, image matrices, and audio sequences). Based on the file type and data structure, the system divides the source data set into multiple data subsets, most common of which are text data, image data, and audio data subsets. For each data subset, the system analyzes its internal structural characteristics and logically groups it for further processing based on its content.

[0044] For example, if the source data set contains a text file, several images, and an audio clip, the system first parses the data. The text file is identified as a .txt file and segmented based on its structural information, such as paragraphs and titles. The image is identified as a .jpg file, and its pixel matrix features are extracted using an image processing algorithm. The audio file is identified as an .mp3 file, and its audio waveform features are extracted using spectral analysis. These data are then divided into three data subsets: a text subset, an image subset, and an audio subset.

[0045] The preset data classification strategy is used to perform rule matching and priority evaluation on the meta-tag information to obtain candidate levels, cross-check the candidate levels with the data block groups, generate data level labels, and use the data level labels to divide the source data set into several data blocks.

[0046] By performing content analysis on each data subset, the system extracts important metadata, such as keywords in text, color histogram features in images, and frequency domain features in audio. This metadata is used to generate meta-tag information. For example, for the text subset, labels such as "confidential," "public," and "internal" are identified; for the image subset, the security level is marked based on its content and resolution; and for the audio subset, the "restricted" label is generated based on the sensitivity of the audio content. Specifically, the preset data classification strategy will match the rules of this meta-tag information and evaluate its priority. For example, based on rules such as data sensitivity, access frequency, or department to which it belongs, the system will assign a candidate level to each subset.

[0047] For example, in a source data set containing internal company reports, the system extracted keywords such as "finance" and "strategic plan" using a text analysis algorithm, and based on these keywords, determined the security level of the text file to be "confidential." For a subset of images, the system labeled them "for internal use" based on information such as size, resolution, and whether they contained employee faces. For an audio subset, speech recognition technology identified "project discussion" content and marked it as "restricted." Ultimately, candidate levels were determined as "high," "medium," and "low," which were then cross-checked with the corresponding data block groups to ensure the security and accuracy of the data subsets.

[0048] The steps of performing format recognition and structural analysis on the source data set, extracting data subsets, performing content analysis and logical grouping on the data subsets, and obtaining data block groups and meta-tag information specifically include: The source data set is formatted and the file type and data structure are identified. The file type and data structure are used to divide the data in the source data set into text, image and audio data subsets.

[0049] Through the format parsing module, the system can perform preliminary parsing of the source data based on the file extension or internal file identifier of the source data set. Specifically, the system will scan the format identifier of each file, identify the file type (such as text file, image file, audio file, etc.), and further analyze the internal data structure of the file. For example, text files can be further structured according to paragraphs, titles, keywords in paragraphs, etc.; image files will use image processing algorithms to identify features such as pixel matrix, color distribution, and image size; audio files will use audio decoding technology to extract spectral features, waveform features and other data. Subsequently, the source data set will be divided into multiple data subsets based on these features, including text data subsets, image data subsets, and audio data subsets.

[0050] For example, in a source data set containing different types of data, the file doc1.txt is identified as a text file containing a company's annual report; image1.jpg is identified as an image file containing employee photos; and audio1.mp3 is identified as an audio file containing a meeting recording. Based on the file type and data structure, the system divides the source data set into three subsets: a text subset (containing doc1.txt), an image subset (containing image1.jpg), and an audio subset (containing audio1.mp3).

[0051] Perform content analysis on each data subset to obtain metadata, use the metadata to construct meta-tag information, and logically group the data subsets based on the meta-tag information to generate data block groups.

[0052] By analyzing the content of each data subset, the system extracts important metadata from each subset. Specifically, for the text subset, natural language processing algorithms are used to extract keywords, key sentences, and paragraph information; for the image subset, image feature extraction algorithms are used to extract color histograms, edge features, and texture information; and for the audio subset, methods such as spectrum analysis and audio time-domain analysis are used to extract the frequency distribution and tonal characteristics of the audio signal. All of this metadata is used to construct meta-tag information, which can include data sensitivity, usage scenarios, and content descriptions. For example, meta-tags for text files include "confidential" and "annual report"; meta-tags for image files include "employee photo" and "public"; and meta-tags for audio files include "meeting recording" and "internal discussion." Based on this meta-tag information, the system can logically group the data subsets, ultimately forming multiple data chunks.

[0053] By analyzing the content of each data subset, the system can extract keywords such as "annual report" and "company finances" from text files, visual information such as "employee faces" and "office environment" from image files, and audio content such as "project discussions" and "meeting minutes" from audio files. This generates relevant meta-tag information and then assigns these data subsets to different groups. For example, the text subset "annual report" is assigned to the "confidential" group, the image subset "employee photos" is assigned to the "public" group, and the audio subset "meeting recordings" is assigned to the "internal discussion" group. These groupings will serve as important basis for subsequent encryption, storage, and retrieval processes.

[0054] In step S200, the data level label is used to match the preset encryption algorithm type and key strength for the corresponding data block, and the corresponding data block is encrypted using the encryption algorithm type and key strength to obtain several encrypted data blocks, and corresponding encryption summary information is generated for each encrypted data block; wherein the encryption summary information includes hash summary and key identification information.

[0055] By applying data level labels, the system can select the appropriate encryption algorithm and key strength based on the data's security level. Specifically, the system matches each data block with a corresponding preset encryption algorithm type (such as AES-128, AES-256, RSA, etc.) and key strength (such as 128-bit, 256-bit, 512-bit). Based on the sensitivity level of each data block, the system will select a different encryption scheme: highly sensitive data blocks will use a stronger encryption algorithm (such as AES-256 or RSA), while less sensitive data blocks will use a lower key strength and encryption algorithm (such as AES-128). After encryption, the system will generate a cryptographic summary for each encrypted data block, including a hash digest (such as SHA-256) and a key identifier (such as a UUID).

[0056] For example, if a data block is marked "Confidential," the system selects the AES-256 encryption algorithm and generates a key identifier, K1. For a data block marked "Public," the system selects the AES-128 encryption algorithm and generates a key identifier, K2. Each encrypted data block also generates a hash digest, H1 (for "Confidential") or H2 (for "Public"), which is used for subsequent data verification and recovery.

[0057] All encrypted data blocks are merged with their corresponding encryption summary information to generate intermediate encrypted data. Semantic feature analysis and context vector encoding are performed on the intermediate encrypted data to obtain a feature vector group and a structural summary.

[0058] By combining encrypted data blocks with their corresponding encryption digest information, the system forms an intermediate encrypted dataset. Specifically, all encrypted data blocks are encapsulated in the intermediate encrypted data along with their hash digest and key identifier. This intermediate encrypted data is then fed into a semantic analysis module, where the system uses a natural language processing (NLP) model (such as BERT or GPT) to extract semantic features from the intermediate encrypted data. The system then further characterizes the encrypted data using contextual vector encoding methods (such as Word2Vec or TF-IDF models), resulting in a feature vector group and a structural summary.

[0059] For example, suppose the system has two encrypted data blocks D1 and D2, each with hash digests H1 and H2, and key identifiers K1 and K2. The system merges these encrypted data (including D1 and D2), their corresponding hash digests H1 and H2, and key identifiers K1 and K2, to form intermediate encrypted data E1 = [D1, H1, K1] and E2 = [D2, H2, K2]. The system then performs semantic analysis on E1 and E2, generating feature vectors V1 = [0.32, 0.88] and V2 = [0.21, 0.93]. These vectors describe the context and semantic structure of the encrypted data.

[0060] The hash summary and key identification information are combined with the structure summary to generate a first data summary set, and the feature vector group is fused with the meta tag information to generate a second feature vector set.

[0061] By combining the hash digest, key identification information, and structured digest of the encrypted data, the system forms a first set of data digests. Specifically, this first set includes the hash digest, key identification information, and structured digest of the data block. This information can be used to verify the integrity of the data and ensure accurate decryption during recovery. Simultaneously, the system combines the feature vector group with the meta-tag information of the data block to generate a second set of feature vectors. This integration of meta-tag information (such as the data's security level, file type, and usage scenario) with the feature vectors allows the semantics of the data to be effectively linked to its encryption characteristics during storage and recovery. This second set of feature vectors will be used for subsequent data recovery, querying, and reassembly.

[0062] For example, suppose the hash digest of a data block is H1, its key identifier is K1, and the structure digest of the data block is S1. By combining H1, K1, and S1, the system generates the first data digest set S1_comb = [H1, K1, S1]. Furthermore, suppose the feature vector of the data block is V1 = [0.32, 0.88] and the metatag information is "confidential." The system fuses V1 and the metatag "confidential" to generate the second feature vector set V2_comb = [V1,"confidential"], which can be used for subsequent indexing and recovery.

[0063] In step S300, an initial index comparison table is constructed based on the hash summary and key identifier of each data block in the first data summary set, and a similarity analysis is performed on the feature vectors in the initial index comparison table to obtain similarity results for each data block; wherein the feature vector refers to a vector representing the semantics and structure of each data block extracted based on its content.

[0064] By combining feature vectors and key identifiers, the system is able to construct an initial index comparison table. Specifically, the system first generates a preliminary index comparison table based on the feature vectors and key identifiers of each data block in the first data summary set, where each record contains the feature vector, key identifier, and its location identifier information of the data block. Next, the system performs a similarity analysis on the feature vectors in the comparison table. By calculating a similarity metric (such as cosine similarity, Euclidean distance, or Manhattan distance) between feature vectors, the system can determine the similarity between different data blocks. For example, if the feature vectors of two data blocks have a high degree of similarity, it means that their contents have a large semantic or structural similarity, and therefore can be classified as the same or similar categories.

[0065] For example, for a data block and , whose eigenvectors are and , the system will calculate using the cosine similarity formula: in, Represents the feature vector and The dot product of and Respectively and Based on the calculated similarity value, the system can determine whether the two data blocks are similar enough. If the similarity exceeds a preset threshold (for example, 0.8), they can be considered to have similar content or structure and can be classified into the same cluster group.

[0066] Based on the similarity results, the data blocks are assigned to different clustering groups to obtain a cluster label set, the cluster label set is fused with the second feature vector set, a multidimensional index vector group is constructed, and the multidimensional index vector group is subjected to dimensionality reduction and spatial mapping transformation to obtain index partitions.

[0067] By analyzing the similarity of hash digests, the system can assign data blocks to different cluster groups. Specifically, based on the hash digest similarity results, the system uses a clustering algorithm (such as K-means, DBSCAN, or hierarchical clustering) to group similar data blocks together. For example, if two data blocks are highly similar, they are assigned to the same cluster group, while those with lower similarity are assigned to different groups. Each cluster group is assigned a cluster label, which represents the common characteristics or features of the data blocks in that group. The system then combines the cluster label set with the second set of feature vectors to construct a set of multidimensional index vectors. These index vectors contain both cluster information and feature vector information for the data blocks. The system then uses dimensionality reduction techniques (such as PCA or t-SNE) to reduce the dimensionality of these multidimensional index vectors and perform spatial mapping transformations for more efficient storage and retrieval. Ultimately, the system obtains one or more index partitions with an optimized storage structure for improved access performance.

[0068] For example, data blocks D1, D2, and D3 are clustered into a group with the cluster label C1. Assuming the feature vectors of these data blocks are V1, V2, and V3, respectively, the system fuses these feature vectors with the cluster label information to generate a new set of index vectors M1 = [V1, C1], M2 = [V2, C1], and M3 = [V3, C1]. Then, using PCA technology, this high-dimensional data is compressed into lower-dimensional vectors for more efficient storage and fast retrieval. Finally, the system maps these index vectors to a new space, resulting in a more optimized index partition.

[0069] Use index partitioning and clustering labels to build multidimensional data storage structures.

[0070] By performing dimensionality reduction and spatial mapping on multidimensional index vector groups, the system constructs an efficient multidimensional data storage structure. Specifically, index partitions and cluster labels help the system determine the storage location of data blocks, ensuring efficient data distribution and rapid location during storage. Based on the cluster labels, index partitions, and storage characteristics of the data blocks, the system generates a multidimensional data storage structure, such as a multi-level directory structure, a tree structure, or a graph structure. This structure optimizes data storage efficiency and reduces search time during data access.

[0071] For example, suppose data blocks D1 and D2 are assigned to cluster C1, and their feature vectors are mapped to a new space S1 after dimensionality reduction. The system then stores these data blocks in the same physical or logical location based on the C1 label and the reduced index vector. In a cloud storage system, this manifests as storing data blocks associated with the C1 label in the same directory. This makes the storage and retrieval of data blocks more efficient and organized, significantly improving system performance and scalability, especially when dealing with large-scale data.

[0072] In step S400, the clustering labels and key identifiers of each data block in the multidimensional data storage structure are extracted, and the redundant distribution level is calculated using the clustering labels of the data blocks. The upload scheduling plan and replica location relationship table of each data block are generated according to the redundant distribution level.

[0073] By extracting the clustering label and key identification information of each data block, the system can analyze the redundancy needs and storage requirements of the data block. Specifically, the system first extracts the clustering label and corresponding key identification of each data block from the multidimensional data storage structure. The clustering label can reflect the sensitivity or access frequency of the data block. Based on the clustering label, the system calculates the redundant distribution level according to the preset redundancy strategy. For example, low-sensitivity data blocks only need to be stored on one storage node, while high-sensitivity data blocks need to be redundantly distributed on multiple nodes. The higher the redundancy level, the more copies of the data block in the cloud storage node. After the redundant distribution level calculation is completed, the system will generate an upload scheduling plan and a replica location relationship table for each data block, indicating which cloud storage nodes and replica locations the data block should be uploaded to.

[0074] For example, suppose a data block D1 belongs to cluster label C1 (high sensitivity). The redundancy distribution level of this data block is set to 3, indicating that this data block needs to be stored on three different cloud storage nodes. On the other hand, another data block D2 belongs to cluster label C2 (low sensitivity) and its redundancy distribution level is 1, indicating that it only needs to be stored on one cloud storage node. Based on the redundancy level, copies of D1 will be distributed to different nodes, while D2 will be uploaded to only one node.

[0075] Combine all upload schedules with their corresponding key identifiers to generate an encrypted upload metadata set.

[0076] By combining each data block's upload schedule with its corresponding key identifier, the system generates an encrypted upload metadata set. Specifically, the upload schedule includes information such as the time, location, and concurrency level of the data block's upload, while the key identifier ensures data protection and encryption management during the upload process. Combining this information, the system generates an encrypted upload metadata set for each data block, which includes the upload schedule, key identifier, encryption algorithm, and other necessary metadata.

[0077] For example, the upload schedule for data block D1 is to be redundantly distributed across nodes Node1, Node2, and Node3. Its corresponding key identifier is K1, and the encryption algorithm is AES-256. By combining this information, the generated encrypted upload metadata set is UploadInfo1 = [D1, Node1, Node2, Node3, K1, AES-256]. This information set is used during the upload process to ensure data security and accuracy.

[0078] During the process of uploading each data block to the preset cloud storage node, the storage index record is generated using the encrypted upload meta information set corresponding to each data block.

[0079] As each data block is uploaded to a pre-defined cloud storage node, the system generates a storage index record using its encrypted upload metadata set. Specifically, as data is uploaded to each node, the system generates a storage index record based on the upload schedule and the encrypted upload metadata set. The record includes the data block's hash value, the upload node's identifier, the upload timestamp, the key identifier, the encryption algorithm type, and replica information. This storage index record is crucial for data integrity verification and subsequent recovery.

[0080] For example, after data block D1 is uploaded to nodes Node1, Node2, and Node3, the generated storage index record Index1 = [D1, H1, Node1, Node2, Node3, Timestamp, K1, AES-256] records the data block's hash value H1, its corresponding replica node, upload timestamp, key identifier K1, and encryption algorithm AES-256. This information facilitates subsequent data retrieval and recovery.

[0081] The storage index record is merged with the replica location relationship table corresponding to the data block to obtain a storage index information set, and the storage index information set is written into the local index server.

[0082] By merging the storage index records with the replica location relationship table, the system can obtain a complete storage index information set. Specifically, the replica location relationship table records the replica locations and redundancy information of data blocks on different storage nodes. After merging this with the storage index records, the system obtains a complete storage index information set. This information set includes the storage locations of all data blocks, redundant replica information, encryption schemes, and other relevant metadata. Finally, the system writes this storage index information set to the local index server for rapid subsequent data retrieval, management, and recovery.

[0083] For example, the storage index record for data block D1 is Index1=[D1,H1,Node1,Node2,Node3,Timestamp,K1,AES-256], and the replica location relationship table record is ReplicaInfo1=[D1,Node1,Node2,Node3]. The resulting storage index information set is IndexInfo1=[D1,H1,Node1,Node2,Node3,Timestamp,K1,AES-256,ReplicaInfo1]. This information set is ultimately written to the local index server IndexServer for subsequent management and retrieval of data blocks.

[0084] In step S500, the storage node and key identifier of the data block are parsed according to the storage index information set of the data recovery request, and a data block request instruction is sent to the cloud storage node based on the storage node to obtain the data block and feature vector group in the multidimensional data storage structure.

[0085] Using the storage index information set in the data recovery request, the system can resolve the storage location and key identifier of the target data block. Specifically, the system obtains the storage index information set from the local storage index server and extracts the storage node (such as Node1, Node2, etc.) and its corresponding key identifier (such as K1, K2, etc.) for each data block. Next, based on the address information of these storage nodes, the system sends a data block request instruction to the cloud storage node, requesting the data block and corresponding feature vector group from the target storage node. The data block request instruction contains the data block's hash value, storage node information, and necessary authentication data.

[0086] For example, if a data recovery request is for data block D1, the system will parse the storage index information set to find that D1 is stored on nodes Node1 and Node2, and obtain the key identifier K1. The system then sends a request to the cloud storage node to obtain data block D1 and the corresponding feature vector group from Node1 and Node2.

[0087] The key identifier is used to call the local key service to obtain the authorization key, obtain the key, decrypt the data block using the key, and obtain the decrypted data block.

[0088] Using the key identifier of the data block, the system can call the local key service to obtain the authorization key. Specifically, the system sends the key identifier (e.g., K1) to the local key service, which returns the corresponding key (e.g., AES key K1), which is used to decrypt the data block. The decryption process uses the same algorithm used for encryption (e.g., AES-256, RSA, etc.), ensuring that the contents of the data block can be restored to its original plaintext form. After decryption is complete, the system obtains the decrypted data block, and the originally encrypted data can now be processed further or reassembled into the source data set.

[0089] For example, for data block D1, the system obtains the key identifier K1 from the index information, obtains the key K1 by calling the local key service, and then uses the key to decrypt the encrypted data blocks stored on Node1 and Node2 to obtain the decrypted data block D1_decrypted.

[0090] The decrypted data blocks are reconstructed and sorted to generate a reorganized data linked list. The reorganized data linked list and the feature vector group are semantically restored and content reorganized to recover the source data set.

[0091] Using the data blocks obtained through decryption, the system reconstructs and sorts the data hierarchically based on the data's hierarchical information and order requirements. Specifically, the system first reconstructs the decrypted data blocks by level based on the data level labels (such as "confidential" and "internal"), and sorts all related data blocks in a preset order. For example, the system can use the file's timestamp, modification history, or other logical sorting criteria to ensure that the order of the recovered data blocks is consistent with the original source data set. Then, by analyzing the feature vectors of each data block and combining it with contextual information, semantic restoration and content reorganization are performed to ultimately restore the source data set. Semantic restoration is achieved through contextual vector decoding technology (such as BERT, Word2Vec, etc.), which restores the content in the data block to its original structure and semantics, ensuring the accuracy and integrity of the recovered data.

[0092] For example, suppose data block D1 contains confidential financial data and project reports. After decryption, the system reconstructs this data block with other confidential data blocks based on the "confidential" label, sorting them by timestamp to ensure the financial data and reports are restored in their original order. The system then performs semantic restoration on each data block based on the feature vector group and contextual information, restoring the data content to its original format, ultimately recovering the source data set.

[0093] In this embodiment, by identifying the format, parsing the structure, and analyzing the content of the source data set, the system accurately divides the data into distinct subsets and logically groups them based on meta-tag information, thereby achieving categorized data management. Using a pre-set data classification strategy and encryption algorithm, the system assigns an appropriate encryption scheme to each data block, ensuring data security. By generating encryption summary information (such as hash digests and key identifiers), the system provides reliable support for subsequent verification and recovery. Through similarity analysis and clustering algorithms based on hash digests, the system groups similar data blocks into clusters. This is optimized using multi-dimensional index vectors and dimensionality reduction techniques, thereby improving data storage and retrieval efficiency. During data upload, the system generates an upload schedule and replica location relationship table based on the redundancy distribution level, ensuring redundant distribution and reliable storage across multiple cloud storage nodes, further ensuring high data availability. During recovery, the system uses the storage index information set to resolve the storage node and key identifier. The system then extracts the data blocks from the cloud storage nodes and, through a combination of key decryption, data reconstruction, and semantic restoration techniques, accurately recovers the original source data set. Overall, this embodiment improves the security, management efficiency and recovery accuracy of data storage through technical means such as encrypted storage, cluster optimization, redundancy control and intelligent recovery, ensuring the efficiency and reliability of cloud data archiving solutions in large-scale data environments.

[0094] Example 3: like Figure 2 As shown, the present application also provides a cloud data archiving system 10 , including an acquisition module 11 , an encryption module 12 , a construction module 13 , a storage module 14 and a recovery module 15 .

[0095] The acquisition module 11 is mainly used to obtain the source data set of the data to be archived, classify the source data set using a preset data classification strategy, obtain data class labels, and divide the source data set into several data blocks using the data class labels.

[0096] The encryption module 12 is mainly used to perform different levels of data encryption on corresponding data blocks based on data level labels to obtain several encrypted data blocks, analyze and encode all encrypted data blocks to obtain a first data summary set and a second feature vector set.

[0097] The construction module 13 is mainly used to generate a clustering label set according to the first data summary set, and construct a multidimensional data storage structure using the clustering label set and the second feature vector set.

[0098] The storage module 14 is mainly used to upload the multidimensional data storage structure to a preset cloud storage node, generate a storage index information set based on the data blocks in the multidimensional data storage structure during the uploading process, and store the storage index information set in a local index server.

[0099] The recovery module 15 is mainly used to obtain data blocks in the multidimensional data storage structure from the cloud storage node according to the storage index information set of the data recovery request when a data recovery request is received, and to perform decryption and data reorganization to restore the source data set.

[0100] In this embodiment, by designing a modular cloud data archiving system 10, the data storage and recovery process can be completed efficiently and securely. The acquisition module 11 accurately classifies and segments the source data set using a preset data classification strategy for subsequent processing and management. The encryption module 12 differentially encrypts the data blocks according to the different security levels of the data and generates corresponding encryption summary information and feature vector sets, thereby ensuring the security and validity of the data. The construction module 13 uses the first data summary set to generate a clustering label set, and combines it with the second feature vector set to construct a multidimensional data storage structure, further improving data storage efficiency and retrieval performance. The storage module 14 uploads the multidimensional data storage structure to the cloud storage node and generates a storage index information set to ensure redundant storage and high availability of the data. At the same time, this information is stored in the local index server, improving the flexibility and accuracy of data management. The recovery module 15 can quickly parse the storage index information set and obtain data blocks from the cloud storage node based on the data recovery request. The source data set is restored through decryption and data reassembly, ensuring the accuracy and integrity of data recovery. Through the collaborative work of these modules, this embodiment realizes an efficient, secure and reliable cloud data archiving and recovery system, ensuring redundant distribution, encryption protection and rapid recovery capabilities of data among different storage nodes, and has broad application prospects.

[0101] It should be noted that, those skilled in the art will clearly understand that, for the sake of convenience and brevity of description, the specific working processes of the above-described system and each module can refer to the corresponding processes in the aforementioned embodiment 1 and will not be repeated here.

[0102] The structures, proportions, sizes, etc. depicted in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not intended to limit the conditions under which the present invention can be implemented and therefore have no substantive technical significance. Any structural modifications, changes in proportional relationships, or adjustments in size should still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and objectives that can be achieved by the present invention.

[0103] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A cloud data archiving method, characterized in that: include: Obtaining a source data set of data to be archived, classifying the source data set using a preset data classification strategy to obtain data class labels, and dividing the source data set into a plurality of data blocks using the data class labels; Performing different levels of data encryption on the corresponding data blocks based on the data level labels to obtain a plurality of encrypted data blocks, and analyzing and encoding all the encrypted data blocks to obtain a first data summary set and a second feature vector set; generating a clustering label set according to the first data summary set, and constructing a multidimensional data storage structure using the clustering label set and the second feature vector set; Uploading the multidimensional data storage structure to a preset cloud storage node, generating a storage index information set based on the data blocks in the multidimensional data storage structure during the uploading process, and storing the storage index information set in a local index server; When a data recovery request is received, the data blocks in the multidimensional data storage structure are obtained from the cloud storage node according to the storage index information set of the data recovery request, and decrypted and reorganized to recover the source data set.

2. The cloud data archiving method according to claim 1, characterized in that: The steps of obtaining a source data set of data to be archived, classifying the source data set using a preset data classification strategy to obtain data class labels, and dividing the source data set into a plurality of data blocks using the data class labels include: Performing format recognition and structural analysis on the source data set to extract a data subset, performing content analysis and logical grouping on the data subset to obtain data block groups and meta tag information; The meta-tag information is subjected to rule matching and priority evaluation using a preset data classification strategy to obtain a candidate grade, the candidate grade is cross-checked with the data block group to generate a data grade label, and the source data set is divided into several data blocks using the data grade label.

3. The cloud data archiving method according to claim 2, characterized in that: The steps of performing format recognition and structural analysis on the source data set, extracting a data subset, performing content analysis and logical grouping on the data subset, and obtaining a data block group and meta tag information include: Performing format analysis on the source data set to identify file types and data structures, and using the file types and data structures to divide the data in the source data set into data subsets of text, image, and audio; Content analysis is performed on each of the data subsets to obtain metadata, meta tag information is constructed using the metadata, and the data subsets are logically grouped using the meta tag information to generate data block groups.

4. The cloud data archiving method according to claim 2, characterized in that: The steps of performing different levels of data encryption on the corresponding data blocks based on the data level labels to obtain a plurality of encrypted data blocks, and analyzing and encoding all the encrypted data blocks to obtain a first data summary set and a second feature vector set include: Using the data level label to match a preset encryption algorithm type and key strength for the corresponding data block, encrypting the corresponding data block using the encryption algorithm type and the key strength to obtain a plurality of encrypted data blocks, and generating corresponding encryption summary information for each of the encrypted data blocks; wherein the encryption summary information includes a hash digest and key identification information; Merging all the encrypted data blocks and their corresponding encryption summary information to generate intermediate encrypted data, performing semantic feature analysis and context vector encoding on the intermediate encrypted data to obtain a feature vector group and a structure summary; The hash summary and the key identification information are combined with the structure summary to generate a first data summary set, and the feature vector group is fused with the meta tag information to generate a second feature vector set.

5. The cloud data archiving method according to claim 1, characterized in that: The step of generating a cluster label set according to the first data summary set, and constructing a multidimensional data storage structure using the cluster label set and the second feature vector set includes: Constructing an initial index comparison table based on the hash digest and key identifier of each data block in the first data summary set, and performing similarity analysis on the feature vectors in the initial index comparison table to obtain similarity results for each data block; wherein the feature vector refers to a vector representing the semantics and structure of each data block extracted based on its content; Allocating the data blocks to different clustering groups based on the similarity results to obtain a clustering label set, fusing the clustering label set with the second feature vector set to construct a multidimensional index vector set, and performing dimensionality reduction and spatial mapping conversion on the multidimensional index vector set to obtain index partitions; A multidimensional data storage structure is constructed using the index partitions and cluster labels.

6. The cloud data archiving method according to claim 1, characterized in that: The step of uploading the multidimensional data storage structure to a preset cloud storage node, generating a storage index information set based on data blocks in the multidimensional data storage structure during the uploading process, and storing the storage index information set in a local index server includes: Extracting the clustering label and key identifier of each data block in the multidimensional data storage structure, calculating the redundancy distribution level using the clustering label of the data block, and generating an upload scheduling plan and a replica location relationship table for each data block according to the redundancy distribution level; Combining all the upload scheduling plans with their corresponding key identifiers to generate an encrypted upload meta-information set; In the process of uploading each data block to the preset cloud storage node, a storage index record is generated using the encrypted upload meta information set corresponding to each data block; The storage index record is merged with the replica location relationship table corresponding to the data block to obtain a storage index information set, and the storage index information set is written into a local index server.

7. The cloud data archiving method according to claim 1, characterized in that: The step of obtaining data blocks in the multidimensional data storage structure from the cloud storage node according to the storage index information set of the data recovery request, and performing decryption and data reorganization to recover the source data set includes: Parsing the storage node and key identifier of the data block according to the storage index information set of the data recovery request, sending a data block request instruction to the cloud storage node based on the storage node, and obtaining the data block and feature vector group in the multidimensional data storage structure; Using the key identifier, calling a local key service to obtain an authorization key, obtaining a key, and decrypting the data block using the key to obtain a decrypted data block; The decrypted data blocks are subjected to data level reconstruction and data sorting to generate a reorganized data linked list, and the reorganized data linked list and the feature vector group are subjected to semantic restoration and content reorganization to recover the source data set.

8. A cloud data archiving method and system device, characterized in that: include: An acquisition module is used to acquire a source data set of data to be archived, classify the source data set using a preset data classification strategy to obtain data class labels, and divide the source data set into a number of data blocks using the data class labels; an encryption module, configured to perform different levels of data encryption on the corresponding data blocks based on the data level labels to obtain a plurality of encrypted data blocks, and to analyze and encode all of the encrypted data blocks to obtain a first data summary set and a second feature vector set; A construction module, configured to generate a clustering label set according to the first data summary set, and construct a multidimensional data storage structure using the clustering label set and the second feature vector set; A storage module is configured to upload the multidimensional data storage structure to a preset cloud storage node, generate a storage index information set based on the data blocks in the multidimensional data storage structure during the uploading process, and store the storage index information set in a local index server; The recovery module is used to obtain the data blocks in the multidimensional data storage structure from the cloud storage node according to the storage index information set of the data recovery request when a data recovery request is received, and to perform decryption and data reorganization to restore the source data set.

Images