Cloud data distributed storage method and system based on block chain

Abstract

The invention belongs to the technical field of cloud data storage, and discloses a cloud data distributed storage method and system based on a block chain. The method comprises the following steps: deploying a distributed storage network and a consensus block chain network in a cloud data center, and constructing a data analysis model, a retrieval tag generation model and a distributed storage strategy generation model; performing data analysis by using the data analysis model to obtain a real-time data analysis result; generating a retrieval tag by using a retrieval tag generation model; generating a distributed storage strategy by using a distributed storage strategy generation model; performing consensus on the real-time storage request by using a consensus block chain network; and performing distributed storage on the plurality of encrypted data fragments and the real-time retrieval tags thereof by using a distributed storage network. According to the method, the problems of centralization risk, insufficient data privacy protection, high retrieval difficulty, high storage cost and poor data storage adaptability in the prior art are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of cloud data storage, and particularly relates to a distributed storage method and system for cloud data based on blockchain. Background Art

[0002] With the development of technology and the improvement of data informatization level, cloud data storage has become the primary choice for enterprises or government departments due to its high security and high reliability. However, there are still some significant defects in the existing technology for cloud data storage, mainly including:

[0003] 1) Centralization risk: Traditional cloud data storage usually relies on a centralized server architecture, which leads to a relatively high risk of single-point failure. Once the central server fails, it may cause large-scale data loss or service interruption;

[0004] 2) Insufficient data privacy protection: In the centralized storage mode, the data owner has limited control over the data, and there is a risk of internal abuse, external leakage, or forced access by government agencies for third-party service providers, seriously threatening data privacy;

[0005] 3) High storage cost: In addition to hardware investment, existing cloud storage also requires high maintenance, upgrade, and operation costs. Especially for the storage of long-term and large amounts of data, the cost problem is more prominent;

[0006] 4) Difficult retrieval: The existing cloud storage steps are simple, resulting in the lack of customized retrieval tags for cloud data, making subsequent data retrieval difficult and affecting the user experience;

[0007] 5) Poor adaptability of data storage: Most of the existing cloud storage is based on preset storage strategies and cannot be automatically adjusted according to data characteristics, resulting in limitations in data storage and affecting normal data storage work. Summary of the Invention

[0008] In order to solve the problems of centralization risk, insufficient data privacy protection, difficult retrieval, high storage cost, and poor adaptability of data storage existing in the prior art, the purpose of the present invention is to provide a distributed storage method and system for cloud data based on blockchain.

[0009] The technical solution adopted by the present invention is as follows:

[0010] A distributed storage method for cloud data based on blockchain, comprising the following steps:

[0011] In a cloud data center, deploy a distributed storage network and a consensus blockchain network, and construct a data analysis model, a retrieval tag generation model, and a distributed storage strategy generation model;

[0012] Use a cloud data center to receive real-time data, and based on the real-time data, use a data analysis model to perform data analysis and obtain real-time data analysis results;

[0013] Based on the real-time data and the corresponding real-time data analysis results, use a retrieval tag generation model to generate retrieval tags and obtain real-time retrieval tags;

[0014] Based on the real-time data analysis results, use a distributed storage strategy generation model to generate distributed storage strategies and obtain real-time distributed storage strategies;

[0015] Based on the real-time distributed storage strategy, fragment and encrypt the real-time data to obtain several encrypted data fragments and a real-time storage request, and use a consensus blockchain network to perform consensus on the real-time storage request;

[0016] If the consensus is successful, based on the real-time distributed storage strategy, use a distributed storage network to perform distributed storage on several encrypted data fragments and their real-time retrieval tags.

[0017] Furthermore, in the cloud data center, deploy a distributed storage network and a consensus blockchain network, and construct a data analysis model, a retrieval tag generation model, and a distributed storage strategy generation model, including the following steps:

[0018] In the cloud data center, identify several data nodes participating in cloud data distributed storage and extract the historical behavior data of each data node;

[0019] Based on the historical behavior data of each data node, use an identity assignment method to divide several data nodes into several consensus nodes and several storage nodes;

[0020] Use blockchain technology to distribute and connect several consensus nodes to obtain a distributed storage network, and distribute and connect several storage nodes to obtain a consensus blockchain network;

[0021] Collect several historical data and preprocess the several historical data to obtain several preprocessed historical data;

[0022] Based on the several preprocessed historical data, use a deep learning algorithm to construct a data analysis model and generate several historical data analysis results;

[0023] Based on the several historical data and the corresponding historical data analysis results, use a natural language processing algorithm to construct a retrieval tag generation model and generate several historical retrieval tags;

[0024] Based on the several historical data analysis results, use a reinforcement learning algorithm to construct a distributed storage strategy generation model and generate several historical distributed storage strategy generation experiences.

[0025] Further, according to the historical behavior data of each data node, using an identity assignment method, several data nodes are divided into several consensus nodes and several storage nodes, including the following steps:

[0026] According to the preset node behavior rules, generate a consensus behavior evaluation matrix for the data nodes, and obtain the information entropy of the consensus behavior evaluation matrix;

[0027] According to the information entropy, obtain the reputation value of each data node, and set the node reputation value update strategy for the data nodes;

[0028] According to the reputation values of the data nodes, sort the several data nodes in descending order of power, and take the first M data nodes as consensus nodes, and the remaining data nodes as storage nodes.

[0029] Further, the data analysis model is constructed based on the LSTM-FPN-CNN-DBN algorithm, and the data analysis model includes a first text feature extraction module constructed based on the LSTM algorithm, an image feature extraction module constructed based on the FPN algorithm, an audio feature extraction module constructed based on the CNN algorithm, an attention weight module constructed based on the Attention mechanism, and a data analysis module constructed based on the DBN algorithm. The first text feature extraction module, the image feature extraction module, and the audio feature extraction module are all connected to the attention weight module, and the attention weight module is connected to the data analysis module;

[0030] The retrieval label generation model is constructed based on the BERT-LSTM-CRF algorithm, and the retrieval label generation model includes a multi-modal feature extraction module, a second text feature extraction module constructed based on the LSTM algorithm, and a retrieval label generation model constructed based on the CRF algorithm;

[0031] The distributed storage strategy generation model is constructed based on the MOPPO-cGAN algorithm, and the distributed storage strategy generation model includes a distributed storage strategy generation module constructed based on the MOPPO algorithm and an adversarial training module constructed based on the cGAN algorithm connected in sequence. The distributed storage strategy generation module is provided with a set of objective functions, an experience replay pool, an Actor network, a Critic network, and an agent. The adversarial training module is provided with a generator and a discriminator.

[0032] Further, use a cloud data center to receive real-time data, and according to the real-time data, use the data analysis model to perform data analysis to obtain real-time data analysis results, including the following steps:

[0033] Use the cloud data center to receive real-time data and preprocess the real-time data to obtain preprocessed real-time data;

[0034] Parse the preprocessed real-time data to obtain preprocessed real-time image data, preprocessed real-time text data, and preprocessed real-time audio data;

[0035] Use the first text feature extraction module of the data analysis model to extract the first real-time text data features of the preprocessed real-time text data;

[0036] Use the image feature extraction module of the data analysis model to extract the real-time image data features of the preprocessed real-time image data;

[0037] Use the audio feature extraction module of the data analysis model to extract the real-time audio data features of the preprocessed real-time audio data;

[0038] According to the preset attention weight values, use the attention weight module of the data analysis model to perform weighted fusion on the first real-time text data features, real-time image data features, and real-time audio data features to obtain real-time weighted fusion features;

[0039] According to the real-time weighted fusion features, use the data analysis module of the data analysis model to perform data analysis to obtain real-time data analysis results.

[0040] Further, according to the real-time data and the corresponding real-time data analysis results, use the retrieval tag generation model to perform retrieval tag generation to obtain real-time retrieval tags, including the following steps:

[0041] Use the second text feature extraction module of the retrieval tag generation model to extract the second real-time text data features of the real-time data analysis results;

[0042] Use the multi-modal feature extraction module of the retrieval tag generation model to extract the real-time key features of the real-time weighted fusion features corresponding to the preprocessed real-time data;

[0043] According to the second real-time text data features and the real-time key features, use the retrieval tag generation model of the retrieval tag generation model to perform retrieval tag generation to obtain real-time retrieval tags.

[0044] Further, according to the real-time data analysis results, use the distributed storage strategy generation model to perform distributed storage strategy generation to obtain real-time distributed storage strategies, including the following steps:

[0045] Parse the real-time data analysis results to obtain several real-time data analysis states, and according to the several real-time data analysis states, update the state space of the agent in the distributed storage strategy generation module of the distributed storage strategy generation model to obtain an updated state space;

[0046] Randomly extract a number of historical distributed storage policy generation experiences from the experience replay pool of the distributed storage policy generation module, and generate a number of possible distributed storage actions based on the number of historical distributed storage policy generation experiences;

[0047] According to a number of possible distributed storage actions, update the action space of the agent in the distributed storage policy generation module to obtain an updated action space;

[0048] Select a real-time objective function from the set of objective functions in the distributed storage policy generation module, and based on the real-time objective function, use the agent to control the Critic network to generate the real-time value of all possible distributed storage actions in the updated action space for each real-time data analysis state in the updated state space;

[0049] According to a number of real-time values, use the agent to control the Actor network to generate the probability distribution of all possible distributed storage actions corresponding to each real-time data analysis state;

[0050] Take the possible distributed storage action with the highest probability distribution in the updated action space as the executed distributed storage action for the corresponding real-time data analysis state;

[0051] Integrate the executed distributed storage actions of all real-time data analysis states in the updated state space to obtain a real-time distributed storage policy.

[0052] Further, according to the real-time distributed storage policy, slice and encrypt the real-time data to obtain a number of encrypted data slices and real-time storage requests, and use the consensus blockchain network to perform consensus on the real-time storage requests, including the following steps:

[0053] According to the real-time data slicing decision in the real-time distributed storage policy and in combination with the slice replica mechanism, slice the real-time data to obtain a number of real-time data slices including replica slices;

[0054] According to the real-time random encryption decision in the real-time distributed storage policy, use a pseudorandom number generator to generate a random key, and according to the random key, use a key derivation function to generate an encryption seed;

[0055] According to the encryption seed, use a dynamic encryption algorithm to dynamically encrypt a number of real-time data slices to obtain a number of encrypted data slices, generate a real-time storage request, and send the real-time storage request to the consensus blockchain network;

[0056] Take the consensus node that receives the real-time storage request in the consensus blockchain network as the primary node, and based on the primary node, use the consensus algorithm to perform consensus on the real-time storage request.

[0057] Further, if the consensus is successful, according to the real-time distributed storage policy, use the distributed storage network to perform distributed storage on several encrypted data shards and their real-time retrieval tags, including the following steps:

[0058] If the consensus is successful, according to the real-time random sending decision of the real-time distributed storage policy, send several encrypted data shards and their real-time retrieval tags to several storage nodes of the distributed storage network;

[0059] Use all storage nodes to locally store the received encrypted data shards and their real-time retrieval tags, and send real-time storage success information to other storage nodes;

[0060] If a storage node receives several real-time storage success messages exceeding the preset quantity threshold, end the distributed storage step; otherwise, continue with the distributed storage.

[0061] A blockchain-based cloud data distributed storage system for implementing the cloud data distributed storage method. The system is set up in the cloud data center and includes an initialization unit, a data analysis unit, a retrieval tag generation unit, a policy generation unit, a data consensus unit, and a distributed storage unit.

[0062] The beneficial effects of the present invention are:

[0063] The present invention discloses a blockchain-based cloud data distributed storage method and system. By deploying a distributed storage network and a consensus blockchain network, the risk of single-point failure is reduced, the overall security is enhanced, and centralized attacks are resisted; using random encryption and a consensus mechanism, it ensures that the data owner has full control over the data, effectively preventing data leakage and unauthorized access; the distributed storage network can be flexibly expanded without a large amount of hardware investment, adapting to the rapid growth and changing needs of the data volume, and by using the distributed storage policy and the distributed network, the hardware and maintenance costs are reduced, achieving a more economical data storage solution; combining the consensus mechanism and the distributed storage policy ensures the consistency and high reliability of data between different nodes, reducing the risk of data loss and damage; through the data analysis model and the retrieval tag generation model, automated data analysis and retrieval tag generation are realized, customized retrieval tags are set for each data, providing a basis for subsequent data retrieval and improving the user experience; through the distributed storage policy generation model, the distributed storage policy is adjusted according to the data characteristics, improving the practicability and efficiency of the distributed storage and ensuring the normal progress of the data storage work.

[0064] Other beneficial effects of the present invention will be further described in the specific implementation manner. Description of the Drawings

[0065] Figure 1It is a flowchart of the method for distributed storage of cloud data based on blockchain in the present invention.

[0066] Figure 2 It is a structural block diagram of the system for distributed storage of cloud data based on blockchain in the present invention. Detailed implementation manners

[0067] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments.

[0068] Embodiment 1:

[0069] As Figure 1 shown, this embodiment provides a method for distributed storage of cloud data based on blockchain, including the following steps:

[0070] S1: In the cloud data center, deploy a distributed storage network and a consensus blockchain network, and construct a data analysis model, a retrieval tag generation model, and a distributed storage policy generation model, including the following steps:

[0071] S1-1: In the cloud data center, confirm a number of data nodes participating in the distributed storage of cloud data, and extract the historical behavior data of each data node;

[0072] S1-2: According to the historical behavior data of each data node, use an identity assignment method to divide the number of data nodes into a number of consensus nodes and a number of storage nodes, including the following steps:

[0073] S1-2-1: According to the preset node behavior rules, generate a consensus behavior evaluation matrix of the data nodes, and obtain the information entropy of the consensus behavior evaluation matrix, including the following steps:

[0074] The preset node behavior rules include defined honest behavior rules, including correct transaction review behavior, accurate identity verification behavior, and compliance with laws, and defined malicious behavior rules, including incorrect transaction review behavior, incorrect identity verification behavior, and violation of laws;

[0075] The consensus behavior evaluation matrix includes an honest behavior evaluation matrix and a malicious behavior evaluation matrix. Normalize each column in the honest behavior evaluation matrix and the malicious behavior evaluation matrix to obtain a normalized honest behavior evaluation matrix and a normalized malicious behavior evaluation matrix; specifically, use the critical value method for normalization, find the maximum and minimum values of each column, subtract the minimum value from each element, and divide by the difference between the maximum and minimum values to obtain a normalized matrix:

[0076] Obtain the information entropy of each column in the normalized honest behavior evaluation matrix and the normalized malicious behavior evaluation matrix. The formula is:

[0077]

[0078] Wherein, E i' is the information entropy of the i'-th honest / malicious behavior; i' is the indicator of honest / malicious behavior; j' is the column indicator of the evaluation matrix; n is the total number of honest / malicious behaviors; p i'j' is the normalized value of the i'-th honest / malicious behavior in the j'-th column;

[0079] S1-2-2: Obtain the reputation value of each data node according to the information entropy, and set the node reputation value update strategy of the data node, including the following steps:

[0080] S1-2-2-1: Obtain the weight of honest behavior and the weight of malicious behavior according to the proportion of the information entropy in the total information entropy. The smaller the information entropy, the greater the weight. The formula is:

[0081]

[0082] Wherein, w i' is the weight of the i'-th honest / malicious behavior;

[0083] S1-2-2-2: Classify the honest behaviors according to the weight of honest behavior, establish a corresponding reward behavior table, and classify the malicious behaviors according to the weight of malicious behavior, and establish a corresponding punishment behavior table;

[0084] S1-2-2-3: Obtain the reward value of the data node in the current consensus stage according to the reward behavior table, and obtain the punishment value of the data node in the current consensus stage according to the punishment behavior table;

[0085] S1-2-2-4: Obtain the corresponding reputation value according to the reward value and punishment value of the data node in the current consensus stage. The formula is:

[0086]

[0087] Wherein, is the reputation value of node i" at iteration number t; is the reward value of the honest behavior level L; is the count of the honest behavior level L; is the punishment value of the malicious behavior level L'; is the count of the malicious behavior level L'; i" is the data node indicator; t is the iteration number indicator; L is the honest behavior level; L' is the malicious behavior level;

[0088] S1-2-3: Sort several data nodes in descending order according to the reputation value of the data node, and use the top M data nodes as consensus nodes, and the remaining data nodes as storage nodes;

[0089] S1-3: Use blockchain technology to distributively connect a number of consensus nodes to obtain a distributed storage network, and distributively connect a number of storage nodes to obtain a consensus blockchain network;

[0090] S1-4: Collect a number of historical data and preprocess the number of historical data to obtain a number of preprocessed historical data;

[0091] The preprocessing includes data cleaning, format conversion, and normalization processing to improve the quality of the data and provide data support for subsequent model construction;

[0092] S1-5: According to a number of preprocessed historical data, use deep learning algorithms to construct a data analysis model and generate a number of historical data analysis results;

[0093] The data analysis model is constructed based on the Long Short-Term Memory (LSTM)-Feature Pyramid Networks (FPN)-Convolutional Neural Networks (CNN)-Deep Belief Network (DBN) algorithm, and the data analysis model includes a first text feature extraction module constructed based on the LSTM algorithm, an image feature extraction module constructed based on the FPN algorithm, an audio feature extraction module constructed based on the CNN algorithm, an attention weight module constructed based on the Attention mechanism, and a data analysis module constructed based on the DBN algorithm. The first text feature extraction module, the image feature extraction module, and the audio feature extraction module are all connected to the attention weight module, and the attention weight module is connected to the data analysis module;

[0094] The image feature extraction module extracts hierarchical features in the image through a series of convolutional layers, activation functions, and pooling layers. It fuses the feature maps of different levels extracted by the CNN using skip connections, constructs a feature pyramid through upsampling and lateral connections, and transmits the high-level semantic information to the low-level, enhancing the semantic expression ability of the low-level features. This can effectively combine the low-level detailed features and high-level semantic features, improving the accuracy of image feature extraction and the expression ability of features. The first text feature extraction module is specifically used to process text data. By leveraging the characteristics of the long short-term memory network (LSTM), it captures the temporal information and long-term dependencies in the text, effectively solving the vanishing gradient problem in the traditional recurrent neural network (RNN), and thus better extracting the deep features in the text. The audio feature extraction module focuses on the processing of audio data. It extracts the frequency features and temporal features in the audio using a convolutional neural network (CNN), and can efficiently capture the key information in the audio signal through local connections and weight sharing mechanisms. The attention weight module serves as a bridge connecting the text, image, and audio feature extraction modules with the data analysis module. Through the Attention mechanism, it assigns different weights to different types of features, enabling the model to pay more attention to the features that have a greater impact on the analysis results, thereby improving the accuracy of the analysis. The data analysis module is the core of the entire model, responsible for comprehensively analyzing the features from different feature extraction modules. As a deep neural network, the deep belief network (DBN) can learn the complex representations in the data, thus achieving more accurate data analysis;

[0095] S1-6: According to a number of historical data and the corresponding historical data analysis results, use natural language processing algorithms to construct a retrieval label generation model and generate a number of historical retrieval labels;

[0096] The retrieval label generation model is constructed based on the Bidirectional Encoder Representations from Transformers (BERT)-LSTM-Conditional Random Field (CRF) algorithm, and the retrieval label generation model includes a multi-modal feature extraction module, a second text feature extraction module constructed based on the LSTM algorithm, and a retrieval label generation model constructed based on the CRF algorithm;

[0097] Through the pre-training of weighted fusion features of a large amount of multimodal data, the multimodal feature extraction module can accurately identify the key features in the weighted fusion features for subsequent label prediction; the second text feature extraction module, for the data analysis results, uses a long short-term memory network (LSTM) to capture the temporal information and long-term dependencies in the data analysis results, can effectively handle the sequence problems in the data analysis results, and extracts deep text features to provide a strong text representation for the subsequent retrieval label generation.

[0098] S1-7: According to the data analysis results of several historical data, use the reinforcement learning algorithm to construct a distributed storage policy generation model and generate several historical distributed storage policy generation experiences.

[0099] The distributed storage policy generation model is constructed based on the Multi-Objective Proximal Policy Optimization (MOPPO)-Conditional Generative Adversarial Network (cGAN) algorithm, and the distributed storage policy generation model includes a distributed storage policy generation module constructed based on the MOPPO algorithm and an adversarial training module constructed based on the cGAN algorithm connected in sequence. The distributed storage policy generation module is provided with a set of objective functions, an experience replay pool, an Actor network, a Critic network, and an agent. The adversarial training module is provided with a generator and a discriminator.

[0100] The Actor network of the distributed storage policy generation module is responsible for outputting the probability distribution of the actions that should be taken in a given state. The goal is to learn an optimal policy, that is, to maximize the long-term cumulative reward. In the continuous action space, the Actor network usually outputs a mean and an optional variance parameter to describe the probability distribution of the actions. The Critic network is responsible for evaluating the value of a given state, that is, predicting the expected return that can be obtained starting from this state and following the current policy, and usually outputs a scalar value representing the value of the state or the state-action value. The experience replay pool is used to store historical experiences for reuse during the training process. The set of objective functions includes functions that define multiple data processing objectives, including minimizing the distributed storage cost, minimizing the distributed storage response time, and maximizing the distributed storage efficiency, etc. The generator of the adversarial training module is used to generate policies, and the discriminator is used to distinguish the generated policies from the optimal policies. Through this adversarial training process, the distributed storage policy generation module can learn more effective policies, and at the same time, the adversarial training module helps ensure the diversity and quality of the policies, so as to find a better optimal solution in the multi-objective optimization problem.

[0101] Based on the analysis results of a number of historical data, using a reinforcement learning algorithm, a distributed storage policy generation model is constructed, and a number of historical distributed storage policy generation experiences are generated, including the following steps:

[0102] S1-7-1: Use the MOPPO-GAN algorithm to construct an initial distributed storage policy generation model; the initial distributed storage policy generation model includes an initial distributed storage policy generation module and an initial adversarial training module;

[0103] S1-7-2: Set a set of objective functions, an experience replay pool, an Actor network, a Critic network, and an agent for the initial distributed storage policy generation module;

[0104] S1-7-3: Take the distributed storage policy generation problem as the simulation environment of the initial distributed storage policy generation module, and set an action space and a state space for the agent;

[0105] S1-7-4: Based on any objective function in the set of objective functions, according to the analysis results of a number of historical data, pre-train the initial distributed storage policy generation module to obtain a pre-trained distributed storage policy generation module, and generate a number of historical distributed storage policies and corresponding historical distributed storage policy generation experiences;

[0106] S1-7-5: Optimize and train the initial generator of the initial adversarial training module according to the analysis results of a number of historical data and the corresponding historical distributed storage policies to obtain an optimized generator, and generate a number of generated distributed storage policies;

[0107] S1-7-6: Optimize and train the initial discriminator of the initial adversarial training module according to a number of historical distributed storage policies and the corresponding generated distributed storage policies to obtain an optimized discriminator, and generate a number of historical discrimination results;

[0108] S1-7-7: Use the Critic network of the pre-trained distributed storage policy generation module to obtain the rewards of a number of generated distributed storage policies, and optimize the Actor network of the pre-trained distributed storage policy generation module according to a number of rewards to obtain an optimized Actor network;

[0109] S1-7-8: Optimize the Critic network of the pre-trained distributed storage policy generation module according to the rewards of a number of generated distributed storage policies and the corresponding historical discrimination results to obtain an optimized Critic network;

[0110] S1-7-9: Traverse all the objective functions in the objective function set, repeat the above adversarial training steps, and obtain an optimized distributed storage policy generation module with an optimized Actor network and an optimized Critic network, and an optimized adversarial training module with an optimized discriminator and an optimized discriminator;

[0111] S1-7-10: Integrate the optimized distributed storage policy generation module and the optimized adversarial training module to obtain a final distributed storage policy generation model, and store a number of historical distributed storage policy generation experiences in the experience replay pool;

[0112] S2: Use the cloud data center to receive real-time data, and use the data analysis model to perform data analysis based on the real-time data to obtain real-time data analysis results, including the following steps:

[0113] S2-1: Use the cloud data center to receive real-time data and preprocess the real-time data to obtain preprocessed real-time data;

[0114] S2-2: Parse the preprocessed real-time data to obtain preprocessed real-time image data, preprocessed real-time text data, and preprocessed real-time audio data;

[0115] S2-3: Use the first text feature extraction module of the data analysis model to extract the first real-time text data feature of the preprocessed real-time text data;

[0116] S2-4: Use the image feature extraction module of the data analysis model to extract the real-time image data feature of the preprocessed real-time image data;

[0117] S2-5: Use the audio feature extraction module of the data analysis model to extract the real-time audio data feature of the preprocessed real-time audio data;

[0118] S2-6: According to the preset attention weight value, use the attention weight module of the data analysis model to perform weighted fusion on the first real-time text data feature, real-time image data feature, and real-time audio data feature to obtain a real-time weighted fusion feature;

[0119] S2-7: According to the real-time weighted fusion feature, use the data analysis module of the data analysis model to perform data analysis to obtain real-time data analysis results;

[0120] S3: According to the real-time data and the corresponding real-time data analysis results, use the retrieval label generation model to perform retrieval label generation to obtain real-time retrieval labels, including the following steps:

[0121] S3-1: Use the second text feature extraction module of the retrieval tag generation model to extract the second real-time text data features of the real-time data analysis results;

[0122] S3-2: Use the multi-modal feature extraction module of the retrieval tag generation model to extract the real-time key features of the real-time weighted fusion features corresponding to the preprocessed real-time data;

[0123] S3-3: According to the second real-time text data features and the real-time key features, use the retrieval tag generation model of the retrieval tag generation model to perform retrieval tag generation to obtain real-time retrieval tags;

[0124] S4: According to the real-time data analysis results, use the distributed storage strategy generation model to perform distributed storage strategy generation to obtain a real-time distributed storage strategy, including the following steps:

[0125] S4-1: Analyze the real-time data analysis results to obtain several real-time data analysis states, and according to the several real-time data analysis states, update the state space of the agent in the distributed storage strategy generation module of the distributed storage strategy generation model to obtain an updated state space;

[0126] S4-2: Randomly extract several historical distributed storage strategy generation experiences from the experience replay pool of the distributed storage strategy generation module, and generate several possible distributed storage actions according to the several historical distributed storage strategy generation experiences;

[0127] S4-3: According to the several possible distributed storage actions, update the action space of the agent in the distributed storage strategy generation module to obtain an updated action space;

[0128] S4-4: Select a real-time objective function from the objective function set of the distributed storage strategy generation module, and based on the real-time objective function, use the agent to control the Critic network to generate the real-time value of all possible distributed storage actions in the updated action space for each real-time data analysis state in the updated state space;

[0129] S4-5: According to the several real-time values, use the agent to control the Actor network to generate the probability distribution of all possible distributed storage actions corresponding to each real-time data analysis state;

[0130] S4-6: Take the possible distributed storage action with the highest probability distribution in the updated action space as the execution distributed storage action corresponding to the real-time data analysis state;

[0131] S4-7: Integrate the execution distributed storage actions of all real-time data analysis states in the updated state space to obtain a real-time distributed storage strategy;

[0132] S5: According to the real-time distributed storage strategy, fragment and encrypt the real-time data to obtain several encrypted data fragments and real-time storage requests, and use the consensus blockchain network to conduct consensus on the real-time storage requests, including the following steps:

[0133] S5-1: According to the real-time data fragmentation decision in the real-time distributed storage strategy and in combination with the fragment replica mechanism, fragment the real-time data to obtain several real-time data fragments including replica fragments;

[0134] S5-2: According to the real-time random encryption decision in the real-time distributed storage strategy, use a pseudo-random number generator to generate a random key, and according to the random key, use a key derivation function to generate an encryption seed;

[0135] S5-3: According to the encryption seed, use a dynamic encryption algorithm to dynamically encrypt several real-time data fragments to obtain several encrypted data fragments, generate a real-time storage request, and send the real-time storage request to the consensus blockchain network;

[0136] S5-4: Take the consensus node that receives the real-time storage request in the consensus blockchain network as the primary node, and based on the primary node, use a consensus algorithm to conduct consensus on the real-time storage request, including the following steps:

[0137] S5-4-1: Take the consensus node that receives the real-time storage request in the consensus blockchain network as the primary node, and use the primary node to send the real-time storage request to other consensus nodes;

[0138] S5-4-2: According to the Byzantine Fault Tolerance (IPBFT) consensus algorithm, use the primary node to broadcast a pre-prepare message to other consensus nodes and conduct a legality verification on the real-time storage request;

[0139] S5-4-3: If the legality verification passes, use the primary node to broadcast a prepare message containing the voting information of the primary node to other consensus nodes and write the prepare message into the message log;

[0140] S5-4-4: Based on all consensus nodes, exchange confirmation messages. If the primary node receives more than the quantity threshold of confirmation messages, the consensus is successful, and use the primary node to otherwise convert the real-time storage request and the corresponding real-time retrieval label into a data block and chain the data block, otherwise the consensus fails;

[0141] S6: If the consensus is successful, according to the real-time distributed storage strategy, use a distributed storage network to conduct distributed storage on several encrypted data fragments and their real-time retrieval labels, including the following steps:

[0142] S6-1: If the consensus is successful, according to the real-time random sending decision of the real-time distributed storage policy, send several encrypted data shards and their real-time retrieval tags to several storage nodes of the distributed storage network;

[0143] S6-2: Use all storage nodes to locally store the received encrypted data shards and their real-time retrieval tags, and send real-time storage success information to other storage nodes;

[0144] S6-3: If a storage node receives several real-time storage success messages exceeding the preset quantity threshold, end the distributed storage step; otherwise, continue with distributed storage.

[0145] Embodiment 2:

[0146] As Figure 2 shown, this embodiment provides a blockchain-based cloud data distributed storage system for implementing the cloud data distributed storage method. The system is set in the cloud data center and includes an initialization unit, a data analysis unit, a retrieval tag generation unit, a policy generation unit, a data consensus unit, and a distributed storage unit.

[0147] The initialization unit is used to deploy a distributed storage network and a consensus blockchain network in the cloud data center, and construct a data analysis model, a retrieval tag generation model, and a distributed storage policy generation model;

[0148] The data analysis unit is used to use the cloud data center to receive real-time data, and according to the real-time data, use the data analysis model to perform data analysis to obtain a real-time data analysis result;

[0149] The retrieval tag generation unit is used to generate retrieval tags according to the real-time data and the corresponding real-time data analysis result by using the retrieval tag generation model to obtain real-time retrieval tags;

[0150] The policy generation unit is used to generate a distributed storage policy according to the real-time data analysis result by using the distributed storage policy generation model to obtain a real-time distributed storage policy;

[0151] The data consensus unit is used to slice and encrypt the real-time data according to the real-time distributed storage policy to obtain several encrypted data shards and a real-time storage request, and use the consensus blockchain network to perform consensus on the real-time storage request;

[0152] The distributed storage unit is used, after the consensus is successful, to perform distributed storage on several encrypted data shards and their real-time retrieval tags according to the real-time distributed storage policy by using the distributed storage network.

[0153] The present invention discloses a method and system for distributed storage of cloud data based on blockchain. By deploying a distributed storage network and a consensus blockchain network, the risk of single-point failure is reduced, the overall security is enhanced, and centralized attacks can be resisted. By using random encryption and a consensus mechanism, it is ensured that the data owner has full control over the data, effectively preventing data leakage and unauthorized access. The distributed storage network can be flexibly expanded without a large amount of hardware investment, adapting to the rapid growth of data volume and changing requirements. Moreover, through the distributed storage strategy and the use of a distributed network, the hardware and maintenance costs are reduced, achieving a more economical data storage solution. By combining the consensus mechanism and the distributed storage strategy, the consistency and high reliability of data among different nodes are guaranteed, reducing the risk of data loss and damage. Through the data analysis model and the retrieval tag generation model, automated data analysis and retrieval tag generation are realized, setting customized retrieval tags for each data, providing a basis for subsequent data retrieval, and improving the user experience. Through the distributed storage strategy generation model, the distributed storage strategy is adjusted according to the data characteristics, improving the practicability and efficiency of distributed storage and ensuring the normal progress of data storage work.

[0154] The present invention is not limited to the above optional embodiments, and anyone can obtain other various forms of products under the inspiration of the present invention. The above specific embodiments should not be construed as limiting the protection scope of the present invention. The protection scope of the present invention should be defined by the claims, and the specification can be used to interpret the claims.

Claims

1. A cloud data distributed storage method based on blockchain, characterized in that: The steps include: In the cloud data center, deploy a distributed storage network and a consensus blockchain network, and build a data analysis model, a retrieval tag generation model, and a distributed storage strategy generation model; Use cloud data centers to receive real-time data, and use data analysis models to perform data analysis based on the real-time data to obtain real-time data analysis results; According to the real-time data and the corresponding real-time data analysis results, a retrieval tag generation model is used to generate retrieval tags to obtain real-time retrieval tags; According to the real-time data analysis results, a distributed storage strategy generation model is used to generate a distributed storage strategy to obtain a real-time distributed storage strategy; According to the real-time distributed storage strategy, the real-time data is sharded and encrypted to obtain several encrypted data shards and real-time storage requests, and the consensus blockchain network is used to reach consensus on the real-time storage requests; If the consensus is successful, a distributed storage network is used to distribute and store several encrypted data shards and their real-time retrieval tags according to the real-time distributed storage strategy.

2. According to a blockchain-based cloud data distributed storage method according to claim 1, it is characterized by: In the cloud data center, a distributed storage network and a consensus blockchain network are deployed, and a data analysis model, a retrieval tag generation model, and a distributed storage strategy generation model are constructed, including the following steps: In the cloud data center, identify several data nodes participating in the distributed storage of cloud data, and extract the historical behavior data of each data node; According to the historical behavior data of each data node, the identity allocation method is used to divide several data nodes into several consensus nodes and several storage nodes; Using blockchain technology, several consensus nodes are connected in a distributed manner to obtain a distributed storage network, and several storage nodes are connected in a distributed manner to obtain a consensus blockchain network; Collecting some historical data, and preprocessing some historical data to obtain some preprocessed historical data; Based on some pre-processed historical data, a deep learning algorithm is used to build a data analysis model and generate some historical data analysis results; Based on a number of historical data and corresponding historical data analysis results, a natural language processing algorithm is used to construct a search tag generation model and generate a number of historical search tags; Based on the analysis results of several historical data, a distributed storage strategy generation model is constructed using a reinforcement learning algorithm, and several historical distributed storage strategy generation experiences are generated.

3. A cloud data distributed storage method based on blockchain according to claim 2, characterized in that: According to the historical behavior data of each data node, the identity allocation method is used to divide several data nodes into several consensus nodes and several storage nodes, including the following steps: According to the preset node behavior rules, the consensus behavior evaluation matrix of the data nodes is generated, and the information entropy of the consensus behavior evaluation matrix is ​​obtained; According to the information entropy, the reputation value of each data node is obtained, and the node reputation value update strategy of the data node is set; According to the reputation value of the data nodes, several data nodes are sorted in descending order, and the first M data nodes are used as consensus nodes, and the remaining data nodes are used as storage nodes.

4. A cloud data distributed storage method based on blockchain according to claim 3, characterized in that: The data analysis model is constructed based on the LSTM-FPN-CNN-DBN algorithm, and the data analysis model includes a first text feature extraction module constructed based on the LSTM algorithm, an image feature extraction module constructed based on the FPN algorithm, an audio feature extraction module constructed based on the CNN algorithm, an attention weight module constructed based on the Attention mechanism, and a data analysis module constructed based on the DBN algorithm. The first text feature extraction module, the image feature extraction module, and the audio feature extraction module are all connected to the attention weight module, and the attention weight module is connected to the data analysis module; The retrieval tag generation model is constructed based on the BERT-LSTM-CRF algorithm, and the retrieval tag generation model includes a multimodal feature extraction module, a second text feature extraction module constructed based on the LSTM algorithm, and a retrieval tag generation model constructed based on the CRF algorithm; The distributed storage strategy generation model is constructed based on the MOPPO-cGAN algorithm, and the distributed storage strategy generation model includes a distributed storage strategy generation module constructed based on the MOPPO algorithm and an adversarial training module constructed based on the cGAN algorithm, which are connected in sequence. The distributed storage strategy generation module is provided with an objective function set, an experience replay pool, an Actor network, a Critic network and an intelligent agent, and the adversarial training module is provided with a generator and a discriminator.

5. A cloud data distributed storage method based on blockchain according to claim 4, characterized in that: Using a cloud data center to receive real-time data, and using a data analysis model to perform data analysis based on the real-time data to obtain real-time data analysis results, includes the following steps: Using a cloud data center to receive real-time data and pre-process the real-time data to obtain pre-processed real-time data; Parsing the preprocessed real-time data to obtain preprocessed real-time image data, preprocessed real-time text data, and preprocessed real-time audio data; Using a first text feature extraction module of the data analysis model, extracting a first real-time text data feature of the preprocessed real-time text data; Using an image feature extraction module of a data analysis model, extracting real-time image data features of the pre-processed real-time image data; Using the audio feature extraction module of the data analysis model to extract real-time audio data features of the pre-processed real-time audio data; According to a preset attention weight value, using an attention weight module of a data analysis model, weighted fusion is performed on the first real-time text data feature, the real-time image data feature, and the real-time audio data feature to obtain a real-time weighted fusion feature; According to the real-time weighted fusion features, the data analysis module of the data analysis model is used to perform data analysis to obtain real-time data analysis results.

6. A cloud data distributed storage method based on blockchain according to claim 5, characterized in that: According to the real-time data and the corresponding real-time data analysis results, the search tag generation model is used to generate the search tag to obtain the real-time search tag, including the following steps: Using a second text feature extraction module of the retrieval tag generation model, extracting a second real-time text data feature of the real-time data analysis result; Use the multimodal feature extraction module of the retrieval tag generation model to extract the real-time key features of the real-time weighted fusion features corresponding to the pre-processed real-time data; According to the second real-time text data feature and the real-time key feature, a retrieval tag generation model is used to generate a retrieval tag to obtain a real-time retrieval tag.

7. A cloud data distributed storage method based on blockchain according to claim 6, characterized in that: According to the real-time data analysis results, a distributed storage strategy generation model is used to generate a distributed storage strategy to obtain a real-time distributed storage strategy, including the following steps: Parsing the real-time data analysis results to obtain a number of real-time data analysis states, and updating the state space of the intelligent agent of the distributed storage strategy generation module in the distributed storage strategy generation model according to the number of real-time data analysis states to obtain an updated state space; Randomly extract a number of historical distributed storage strategy generation experiences from the experience replay pool of the distributed storage strategy generation module, and generate a number of possible distributed storage actions based on the number of historical distributed storage strategy generation experiences; According to a number of possible distributed storage actions, the action space of the agent of the distributed storage strategy generation module is updated to obtain an updated action space; Select a real-time objective function from the objective function set of the distributed storage strategy generation module, and based on the real-time objective function, use the agent to control the Critic network to generate the real-time value of all possible distributed storage actions in the updated action space for each real-time data analysis state in the updated state space; Based on several real-time values, use intelligent agents to control the Actor network and generate the probability distribution of all possible distributed storage actions corresponding to each real-time data analysis state; The possible distributed storage action with the highest probability distribution in the updated action space is used as the execution distributed storage action of the corresponding real-time data analysis state; Integrate the distributed storage actions of all real-time data analysis states in the updated state space to obtain a real-time distributed storage strategy.

8. A cloud data distributed storage method based on blockchain according to claim 7, characterized in that: According to the real-time distributed storage strategy, the real-time data is sharded and encrypted to obtain several encrypted data shards and real-time storage requests, and a consensus blockchain network is used to reach a consensus on the real-time storage requests, including the following steps: According to the real-time data sharding decision in the real-time distributed storage strategy and in combination with the sharding replica mechanism, the real-time data is sharded to obtain a number of real-time data shards including replica shards; According to the real-time random encryption decision in the real-time distributed storage strategy, a pseudo-random number generator is used to generate a random key, and according to the random key, an encryption seed is generated using a key derivation function; According to the encrypted seed, a dynamic encryption algorithm is used to dynamically encrypt several real-time data shards to obtain several encrypted data shards, generate real-time storage requests, and send the real-time storage requests to the consensus blockchain network; The consensus node that receives the real-time storage request in the consensus blockchain network is used as the master node, and based on the master node, a consensus algorithm is used to reach a consensus on the real-time storage request.

9. A cloud data distributed storage method based on blockchain according to claim 8, characterized in that: If the consensus is successful, the distributed storage network is used to perform distributed storage of several encrypted data shards and their real-time retrieval tags according to the real-time distributed storage strategy, including the following steps: If the consensus is successful, several encrypted data shards and their real-time retrieval tags are sent to several storage nodes of the distributed storage network according to the real-time random sending decision of the real-time distributed storage strategy; Use all storage nodes to locally store the received encrypted data shards and their real-time retrieval tags, and send real-time storage success information to other storage nodes; If the storage node receives a number of real-time storage success messages exceeding a preset number threshold, the distributed storage step is terminated; otherwise, the distributed storage is continued.

10. A cloud data distributed storage system based on blockchain, used to implement the cloud data distributed storage method according to any one of claims 1 to 9, characterized in that: The system is arranged in a cloud data center, and the system includes an initialization unit, a data analysis unit, a retrieval tag generation unit, a strategy generation unit, a data consensus unit and a distributed storage unit.

Images