Data storage management method and platform for extended warranty services

Abstract

The application discloses a data storage management method and platform for warranty service, relates to the technical field of safe data management, and comprises the following steps: performing wavelet packet transformation and modal decomposition on a device running data set to obtain a time-frequency domain feature matrix; performing random mask encryption on the time-frequency domain feature matrix by using a quantum chaotic encryption algorithm to output an encrypted device state vector; inputting the encrypted device state vector into a federal global health model; applying differential privacy law to a privacy learning layer for distributed feature learning to obtain device feature latent variables; and applying a secure multi-party computation protocol to a global aggregation layer for cross-node parameter fusion to form a federal consensus parameter. The application enhances the anti-attack capability and privacy protection strength of data in the transmission and storage process by using the quantum chaotic encryption algorithm and the federal global health model, and improves the intelligent level and privacy security guarantee capability of the data storage management method.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of secure data management, in particular to a data storage management method and platform for warranty service. BACKGROUND

[0002] With the rapid development of Internet of Things (IoT) technology, the degree of intelligence of devices is increasing, which leads to the generation of massive device operation data. Device operation data can effectively analyze device health status and predict potential failures, providing accurate decision-making basis for warranty service. Existing storage management methods collect device operation data and perform centralized storage and simple encryption processing, which to some extent realizes the traceability and basic security of data, and realizes the basic needs of data retention and retrieval in warranty business.

[0003] The existing data storage management method mainly faces two major challenges: first, the lack of efficient and accurate feature extraction and encryption mechanism in data processing, which leads to the vulnerability of device operation data in transmission and storage process, affecting the credibility of warranty service. Second, the value assessment of data is too static, which fails to fully consider the value fluctuation of data over time and the influence of external conditions on data value, leading to unreasonable allocation of resources in warranty service. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a data storage management method for warranty service to solve the problems of lack of effective encryption mechanism and static value assessment.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a data storage management method for warranty service, which includes performing wavelet packet transform and modal decomposition on a device operation data set to obtain a time-frequency domain feature matrix; applying a quantum chaotic encryption algorithm to perform random mask encryption on the time-frequency domain feature matrix to output an encrypted device state vector;

[0008] The encrypted device state vector is input into a federal global health model, the privacy learning layer applies differential privacy method for distributed feature learning to obtain device feature latent variables; the global aggregation layer applies secure multi-party computation protocol for cross-node parameter fusion to form federal consensus parameters; the device feature latent variables and the federal consensus parameters are subjected to attention weighting and normalized activation to output a device health assessment report;

[0009] Risk level weights, access frequency values and legal compliance weights of the equipment health assessment report are extracted, and dynamic weighted fusion and time sequence decay correction are performed to generate a multi-dimensional value evaluation vector; the multi-dimensional value evaluation vector is anchored by a block chain to form an extended warranty smart contract;

[0010] The extended warranty smart contract is subjected to semantic analysis and instruction mapping to generate storage strategy instructions; according to the storage strategy instructions, the encrypted equipment state vector is migrated and stored to form a storage location update record; the storage location update record is subjected to Merkle tree storage to obtain a storage management audit report.

[0011] As a preferred scheme of the data storage management method for extended warranty service, the output encrypted equipment state vector specifically includes the following steps,

[0012] Wavelet packet transformation is performed on the equipment operation data set to obtain a multi-scale energy spectrum feature; the multi-scale energy spectrum feature is subjected to modal decomposition to obtain a time-frequency domain feature matrix;

[0013] The quantum chaotic encryption algorithm is applied to the time-frequency domain feature matrix for phase space mapping to form a chaotic disturbance base, and the chaotic disturbance base is subjected to random mask encryption to output the encrypted equipment state vector.

[0014] As a preferred scheme of the data storage management method for extended warranty service, the federal global health model is specifically constructed as follows,

[0015] A privacy learning layer is built by a differential privacy neural network, and a global aggregation layer is built by a federal average architecture;

[0016] The privacy learning layer and the global aggregation layer are integrated and stacked in layers to build the federal global health model.

[0017] As a preferred scheme of the data storage management method for extended warranty service, the device feature latent variable is specifically obtained as follows,

[0018] The encrypted equipment state vector is input into the federal global health model, and the privacy learning layer performs gradient quantization on the encrypted equipment state vector to form an encrypted gradient parameter;

[0019] The encrypted gradient parameter is subjected to Gaussian noise disturbance and distributed feature learning by applying differential privacy law to obtain the device feature latent variable.

[0020] As a preferred scheme of the data storage management method for extended warranty service, the federal consensus parameter is specifically formed as follows,

[0021] The global aggregation layer performs aggregation calculation on the device feature latent variable to form an encrypted aggregation feature;

[0022] According to a secure multi-party computation protocol, the encrypted aggregation feature is fused across nodes to form a federated consensus parameter.

[0023] As a preferred scheme of the data storage management method for extended service, the output device health assessment report specifically includes the following steps,

[0024] The device feature latent variable and the federated consensus parameter are weighted by the multi-head attention mechanism, and the health state feature is output.

[0025] The health state feature is normalized and activated to generate a health degree score, and the health degree score is structured and coded to output a device health assessment report.

[0026] As a preferred scheme of the data storage management method for extended service, the formation of the extended service smart contract specifically includes the following steps,

[0027] The device health assessment report is singular value decomposed to extract risk level weights, access frequency values and legal compliance weights;

[0028] The risk level weights, access frequency values and legal compliance weights are dynamically weighted and fused by the entropy weight method, and are time-decay corrected by a time decay factor to generate a multi-dimensional value evaluation vector;

[0029] The multi-dimensional value evaluation vector is hashed and anchored to the blockchain to form an extended service smart contract.

[0030] As a preferred scheme of the data storage management method for extended service, the storage location update record specifically includes the following steps,

[0031] The extended service smart contract is semantically parsed to form a storage hierarchical parameter, and the storage hierarchical parameter is instruction mapped to generate a storage strategy instruction;

[0032] According to the storage strategy instruction, the encrypted device state vector is migrated across the chain and stored to form a distributed storage credential, and the distributed storage credential is hashed and recorded to form a storage location update record.

[0033] As a preferred scheme of the data storage management method for extended service, the storage management audit report specifically includes the following steps,

[0034] The storage location update record is hashed and block-sampled to obtain a hash node set;

[0035] Performing a Merkel tree storage on the hash node set, and obtaining a storage management audit report.

[0036] In a second aspect, the application provides a data storage management system for extended service, comprising: a quantum encryption module, configured to perform wavelet packet transform and modal decomposition on a device running data set to obtain a time-frequency domain feature matrix; and perform random mask encryption on the time-frequency domain feature matrix by using a quantum chaotic encryption algorithm to output an encrypted device state vector.

[0037] A health assessment module, configured to input the encrypted device state vector into a federal global health model, perform distributed feature learning by using a differential privacy method at a privacy learning layer to obtain a device feature latent variable, perform cross-node parameter fusion by using a secure multi-party computation protocol at a global aggregation layer to form a federal consensus parameter, and perform attention weighting and normalized activation on the device feature latent variable and the federal consensus parameter to output a device health assessment report.

[0038] A contract generation module, configured to extract a risk level weight, an access frequency prediction value and a legal compliance weight of the device health assessment report, and perform dynamic weighted fusion and time sequence attenuation correction to generate a multi-dimensional value evaluation vector; and perform blockchain anchoring on the multi-dimensional value evaluation vector to form an extended service smart contract.

[0039] A storage execution module, configured to perform semantic analysis and multi-objective optimization on the extended service smart contract to generate a storage strategy instruction, perform migration storage on the encrypted device state vector according to the storage strategy instruction to form a storage location update record, and perform a Merkel tree storage on the storage location update record to obtain a storage management audit report.

[0040] The application has the following beneficial effects: the wavelet packet transform and the quantum chaotic encryption algorithm are used for time-frequency domain feature extraction and random mask encryption, which significantly enhances the anti-attack ability and privacy protection strength of data in the transmission and storage process, thereby ensuring the data credibility of the extended service. Meanwhile, the federal global health model is used for distributed feature learning and cross-node parameter fusion, which realizes precise evaluation of the device health state and collaborative optimization of the consensus parameter, enhances the rationality of resource allocation of the extended service, and further improves the intelligent level and privacy security protection ability of the data storage management method. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0042] Fig. 1Flow chart of a data storage management method for extended service.

[0043] Fig. 2 Schematic diagram of a data storage management system for extended service.

[0044] Fig. 3 Flow chart of quantum chaotic encryption.

[0045] Fig. 4 Flow chart of federal consensus parameter generation. DETAILED DESCRIPTION

[0046] In order to make the above objectives, features and advantages of the present application more apparent, specific embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0047] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details given herein, that the present application can be practiced with other than the described embodiments, and that variations from the particular embodiments described herein can be made and still be within the scope of the present application.

[0048] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is separate or alternative to other embodiments.

[0049] Reference Figs. 1-4 For one embodiment of the present application, the embodiment provides a data storage management method for extended service, comprising the following steps:

[0050] S1, wavelet packet transform and modal decomposition are performed on the equipment operation data set to obtain a time-frequency domain feature matrix; a quantum chaotic encryption algorithm is applied to perform random mask encryption on the time-frequency domain feature matrix to output an encrypted equipment state vector.

[0051] Specifically, the following steps are included,

[0052] S1.1, the equipment operation data set is collected and preprocessed.

[0053] In specific operation,

[0054] The equipment operation data set includes vibration signals, temperature parameters and operating conditions; the vibration signals are collected by an acceleration sensor, the temperature parameters are collected by a temperature and humidity sensor, and the operating conditions are collected by an operating monitoring software (such as SCADA).

[0055] Next, the device operation data set is preprocessed. Further, for the vibration signal, baseline correction is performed through zero mean processing to eliminate baseline drift of the vibration signal, and wavelet denoising is applied for interference suppression to improve the purity of the vibration signal. Linear interpolation is used for resolution resampling to ensure time consistency.

[0056] For the temperature parameter, missing data is completed using adjacent point interpolation to restore the continuity of the temperature parameter, and moving average is used for fluctuation smoothing to weaken random disturbance. At the same time, the range is normalized to eliminate the influence of dimension.

[0057] For the operating condition, NTP time stamp protocol is applied for time alignment to ensure time consistency, and Z-Score standardization is used for format unification to output the preprocessed device operation data set.

[0058] S1.2, wavelet packet transform is performed on the preprocessed device operation data set to obtain multi-scale energy spectrum features; modal decomposition is performed on the multi-scale energy spectrum features to obtain a time-frequency domain feature matrix.

[0059] In specific operation,

[0060] db4 wavelet basis function is used to perform wavelet packet transform on the preprocessed device operation data set. Further, the preprocessed device operation data set is decomposed into multiple layers (such as 5 layers) to obtain multi-layer frequency domain spectrum, and the multi-layer frequency domain spectrum is separated by frequency band and normalized to generate sub-band coefficients. Energy integration is performed on each sub-band coefficient to form a frequency spectrum energy distribution. Frequency band fusion and feature extraction are performed on the frequency spectrum energy distribution to obtain multi-scale energy spectrum features.

[0061] Subsequently, the multi-scale energy spectrum features are modal decomposed using Hilbert transform. Further, the multi-scale energy spectrum features are frequency separated, and the first L (such as L=5) instantaneous frequency components are extracted as dominant frequency components. Bandpass filtering is performed on the dominant frequency components to eliminate irrelevant frequency interference to generate time-frequency modal components. The time-frequency modal components are parameterized sampled to obtain time-frequency characteristic parameters, and the time-frequency characteristic parameters are matrix reconstructed to generate a time-frequency domain feature matrix. The row elements in the time-frequency domain feature matrix correspond to time parameters, and the column elements represent time-frequency characteristic parameters of different modes.

[0062] S1.3, quantum chaotic encryption algorithm is applied to the time-frequency domain feature matrix for phase space mapping to form a chaotic disturbance base. The chaotic disturbance base is randomly masked and encrypted to output an encrypted device state vector.

[0063] In specific operation,

[0064] The quantum chaotic encryption algorithm is applied to the phase space mapping of the time-frequency domain feature matrix. Further, the time-frequency domain feature matrix is input into a trusted execution environment (such as Intel SGX) to ensure zero exposure during the encryption process, and a quantum random number generator (QRNG) is used for chaotic initialization. For example, the energy entropy value and the center frequency of the frequency domain feature matrix are extracted, the energy entropy value is set as the chaotic sensitive factor, and the center frequency is set as the phase space boundary parameter to construct the Lorenz chaotic factor. Then, the Lorenz chaotic factor is mapped in the phase space, and further, the Lorenz chaotic factor is trajectory fitted to generate a chaotic state trajectory, which is discretized to obtain a chaotic parameter sequence. The chaotic parameter sequence is subjected to amplitude modulation and phase disturbance to output a chaotic disturbance base.

[0065] It should be noted that trajectory fitting refers to the process of state iteration and curve fitting of the Lorenz chaotic factor by numerical integral fitting method; phase disturbance refers to the process of phase shift and nonlinear superposition of the chaotic parameter sequence.

[0066] Then, the chaotic disturbance base is randomly masked and encrypted, and further, a random mask factor is used to superimpose the mask on the chaotic disturbance base to obtain a mask chaotic matrix. The mask chaotic matrix is subjected to hash encryption and dimension compression to generate encrypted mask data. The encrypted mask data is completed by bilinear interpolation to obtain a standard ciphertext vector. The standard ciphertext vector is associated with the device identifier to generate an encrypted device state vector.

[0067] It should be noted that the random mask factor is defined based on the dynamic sensitivity parameters of the chaotic disturbance base, and the exemplary value range is [0.02, 0.05]; the hash encryption refers to the process of one-way hash encryption of the mask chaotic matrix; the device identifier association refers to the process of matching and storing the standard ciphertext vector and the device unique code (collected by FIDO2 authentication software).

[0068] S2, input the encrypted device state vector into the federal global health model, the privacy learning layer applies differential privacy method for distributed feature learning to obtain device feature latent variable; the global aggregation layer applies secure multi-party computation protocol for cross-node parameter fusion to form federal consensus parameters; the device feature latent variable and the federal consensus parameters are subjected to attention weighting and normalized activation to output a device health assessment report.

[0069] Specifically, the following steps are included,

[0070] S2.1, construct and train the federal global health model.

[0071] In specific operation,

[0072] In the TensorFlow framework, the differential privacy neural network is called by the noise_multiplier parameter and initialized, for example, the privacy budget is set to 0.5, the privacy relaxation factor is set to 1e-5, and the regularization coefficient is set to 0.3; the differential privacy neural network is followed by a normalization activation layer for feature standardization to eliminate dimensional differences, and a GELU activation function is used for gradient normalization to complete the construction of the privacy learning layer; the federated average architecture is called by the FedAvg parameter and initialized, for example, the learning rate is set to 0.01, the maximum communication round is set to 100, and the client sampling rate is set to 0.8; the federated average architecture is embedded with a secure multi-party computation protocol to protect parameter aggregation privacy, and a multi-head attention mechanism is connected to perform dynamic weight allocation to complete the construction of the global aggregation layer.

[0073] The privacy learning layer and the global aggregation layer are fused using residual connections to obtain integrated feature representations; the KL divergence is used to align the distribution of the integrated feature representations to obtain regularized features; the regularized features are activated by the Softmax function to generate attention allocation weights; the privacy learning layer and the global aggregation layer are stacked according to the attention allocation weights to complete the construction of the federated global health model.

[0074] Next, the federated global health model is trained, and further, the historical encrypted device state vector is divided into a sample set, a training set, and a validation set; on the sample set, the data augmenter is used for sample expansion and feature enhancement to form enhanced samples; on the training set, the AdamW optimizer is used to fine-tune the parameters of the enhanced samples, and the gradient clipping is applied simultaneously to constrain the gradient, obtaining updated federated global health model parameters; on the validation set, the cross-entropy loss function is applied to quantify the loss of the updated federated global health model parameters, obtaining the validation loss; when the validation loss exceeds the convergence threshold for consecutive rounds (e.g., 5 times), the training is terminated, and the trained federated global health model is output synchronously.

[0075] It should be noted that the convergence threshold is defined based on the volatility of the historical validation loss, and the example value range is [0.001, 0.005].

[0076] S2.2, the privacy learning layer applies the differential privacy method for distributed feature learning to obtain device feature latent variables.

[0077] In specific operations,

[0078] The encrypted device state vector is input into the federal global health model through a secure data channel (such as a TLS 1.3 encrypted channel), and a privacy learning layer applies a multi-layer differential privacy neural network to gradient quantization of the encrypted device state vector. Further, the first layer performs feature dimension reduction mapping on the encrypted device state vector to obtain encrypted low-dimensional features, and performs L2 gradient clipping on the encrypted low-dimensional features to generate normalized gradients. The second layer performs sliding window cumulative integration on the normalized gradients to obtain aggregated gradient parameters. The exponential weighted smoothing is performed on the aggregated gradient parameters to output stable gradient parameters. The third layer performs homomorphic encryption and gradient sparsification on the stable gradient parameters to form encrypted gradient parameters.

[0079] It should be noted that L2 gradient clipping refers to the process of gradient scaling and norm constraint on the encrypted device state vector; homomorphic encryption refers to the process of encrypting and securely aggregating stable gradient parameters using Paillier encryption algorithm.

[0080] The encrypted gradient parameters are subjected to Gaussian noise disturbance using differential privacy law. Further, dynamic noise multipliers are used to inject and scale Gaussian noise into the encrypted gradient parameters to obtain noise disturbance gradients. The noise disturbance gradients are subjected to homomorphic decryption reconstruction to output global gradient update representation.

[0081] It should be noted that the dynamic noise multiplier is defined based on the privacy budget consumption rate of the historical encrypted gradient parameters, and the exemplary value range is [0.5, 1.2]; homomorphic decryption reconstruction refers to the process of private key decryption and gradient normalization of noise disturbance gradients.

[0082] Then, distributed feature learning is performed on the global gradient update representation. Further, feature projection transformation is performed on the global gradient update representation to obtain low-dimensional gradient embedding representation. The low-dimensional gradient embedding representation is subjected to distribution alignment and sparse coding to obtain regularized feature representation. The regularized feature representation is subjected to non-linear conversion by Swish activation function to output device feature latent variable.

[0083] It should be noted that sparse coding refers to the process of feature selection and coefficient sparsification of low-dimensional gradient embedding.

[0084] S2.3, the global aggregation layer applies a secure multi-party computation protocol to cross-node parameter fusion to form a federal consensus parameter.

[0085] In specific operation,

[0086] The global aggregation layer performs aggregation calculations on the device feature latent variables to form encrypted aggregated features. Further, it extracts the encrypted feature vectors from the device feature latent variables and performs normalization processing to obtain standard encryption parameters. Local weighted aggregation is then performed on the standard encryption parameters to obtain locally aggregated ciphertext. The locally aggregated ciphertext is then distributed-key-sharded to obtain encrypted shard data. Finally, the encrypted shard data is aggregated using a Lagrange basis function to obtain encrypted aggregated vectors. The specific mathematical format is as follows.

[0087] ;

[0088] in, Represents an encrypted aggregate vector. Indicates an index for encrypted fragmented data. Indicates the total number of encrypted data fragments. Indicates the first The weight of each encrypted data fragment Indicates the first The parameter update amount for each encrypted data segment;

[0089] It should be noted that the weights of the encrypted fragment data are defined based on the stability contribution rate of the historical encrypted aggregate vector, with an exemplary value range of [0.1, 0.9]; the parameter update amount is obtained by performing a feature space transformation operation on the encrypted fragment data.

[0090] The encrypted aggregated vector is subjected to Gaussian filtering and dimensionality compression to form encrypted aggregated features. According to the secure multi-party computation protocol, the encrypted aggregated features are fused across nodes. Furthermore, the encrypted aggregated features are decrypted with a threshold to obtain collaborative decryption nodes. The collaborative decryption nodes are dynamically weighted and averaged to obtain node fusion decryption parameters. The node fusion decryption parameters are cross-validated to form federated consensus parameters.

[0091] It should be noted that threshold decryption refers to the process of using Shamir secret sharing to shard the encrypted aggregated features to t nodes and then performing collaborative decryption on the nodes; cross-validation consensus refers to using zero-knowledge proofs to verify the gradient distribution of the node fusion decryption parameters in order to improve the global convergence of the node fusion decryption parameters.

[0092] S2.4. Perform attention-weighted and normalized activation on the latent variables of equipment characteristics and federated consensus parameters, and output the equipment health assessment report.

[0093] In specific operations,

[0094] The device feature latent variable and the federated consensus parameter are weighted by using the multi-head attention mechanism, and normalized activation is performed to generate a health score. Further, the device feature latent variable is linearly projected to generate a query vector, and the federated consensus parameter is simultaneously subjected to bilinear mapping and feature decoupling to obtain a key vector and a value vector. The query vector, the key vector, and the value vector are weighted by using a scaling dot product formula to obtain a weighted health feature, and residual connection is performed on the weighted health feature to output a health state feature. The health state feature is normalized to eliminate dimensional differences, and a GELU activation function is applied for nonlinear transformation to generate a preliminary health score, and the specific mathematical formula is as follows,

[0095] ;

[0096] wherein, is the preliminary health score, is the query vector; is the key vector, denotes a transposition operation, denotes the value vector, denotes a scaling factor;

[0097] It should be noted that the scaling factor is defined based on the dimension of the key vector, and the example value range is [64, 256].

[0098] The health score is compressed to the interval [0, 1] by using a Sigmoid function to form a standard health score. At the same time, the standard health score is subjected to trend analysis by using a sliding window statistical method. Further, the standard health score is subjected to sliding difference processing to obtain the mean and standard deviation of the standard health score, and the mean and standard deviation of the standard health score are subjected to exponential moving average to generate a trend feature vector. According to the trend feature vector, the standard health score is subjected to abnormal pattern recognition. For example, when the trend feature vector exceeds an abnormal threshold, the corresponding standard health score is defined as an abnormal health state, and is integrated to obtain an abnormal fluctuation pattern set.

[0099] It should be noted that the abnormal threshold is defined based on the dynamic fluctuation range of the historical trend feature vector, and the example value range is [0.15, 0.85].

[0100] The abnormal fluctuation pattern set and the standard health score are structured and encoded by using the Protobuf serialization protocol. Further, the abnormal fluctuation pattern and the standard health score are subjected to feature splicing and binary encoding to obtain compressed health data stream. The compressed health data stream is subjected to Base64 transcoding to obtain a transmittable health data packet. The transmittable health data packet is subjected to JSON structured packaging to generate a device health evaluation report.

[0101] It should be noted that the Protobuf serialization protocol is called through the Java language code library; Base64 transcoding refers to the process of ASCII character mapping of compressed health data stream.

[0102] S3, extract the risk level weight, access frequency value and legal compliance weight of the equipment health evaluation report, and dynamically weight fusion and time sequence decay correction to generate a multi-dimensional value evaluation vector; the multi-dimensional value evaluation vector is anchored by the blockchain to form an extended warranty smart contract.

[0103] Specifically, the following steps are included,

[0104] S3.1, singular value decomposition is performed on the equipment health evaluation report to extract the risk level weight, access frequency value and legal compliance weight.

[0105] In specific operation,

[0106] The equipment health evaluation report is dimensionally compressed and missing value filled to output a standardized health evaluation matrix; then the standardized health evaluation matrix is subjected to eigenvalue decomposition to decompose the standardized health evaluation matrix into left singular values and right singular values, and the first k left singular values are extracted for integration to obtain a left singular vector set;

[0107] The first left singular vector in the left singular vector set and the standard health degree score are weighted and fused to obtain a risk feature vector, which is subjected to Softmax normalization to obtain the risk level weight; the second left singular vector and the access timestamp are linearly combined to obtain an access time sequence feature, and the access time sequence feature is subjected to Min-Max scaling to generate an access frequency value; the third left singular vector is projected by inner product with the compliance label to obtain a compliance association feature, and the compliance association feature is subjected to Sigmoid activation to output the legal compliance weight;

[0108] It should be noted that the access timestamp is collected from the device running log; the compliance label is extracted from the device historical compliance record (collected by Splunk enterprise version software).

[0109] S3.2, apply the entropy weight method to dynamically weight fusion of the risk level weight, access frequency value and legal compliance weight, and perform time sequence decay correction through a time decay factor to generate a multi-dimensional value evaluation vector.

[0110] In specific operation,

[0111] The risk level weight, access frequency value and legal compliance weight are subjected to dynamic weight fusion to obtain a multi-index weighted matrix; the entropy weight of the multi-index weighted matrix is calculated to obtain the information entropy value of each index, and the specific mathematical formula is as follows,

[0112] ;

[0113] wherein, denotes the index of the indicator, the information entropy value of the first indicator, denotes the total number of devices, denotes the index of the device, denotes the entropy weight value of the first device on the first indicator;

[0114] The information entropy value is time-series decay corrected by a time decay factor, for example, the information entropy value in the last 7 days is assigned a high decay weight (such as 0.7), and the information entropy value of more than 30 days in history is assigned a low decay weight (such as 0.3); according to the assigned decay weight, the information entropy value is weighted and aggregated to obtain a time-series weighted entropy value;

[0115] It should be noted that the time decay factor is defined based on the dynamic fluctuation range of the historical information entropy value, and the exemplary value range is [0.1, 0.9].

[0116] Then the multi-indicator weighted matrix, the time-series weighted entropy value and the time decay factor are normalized and linearly combined to generate a preliminary evaluation vector; the preliminary evaluation vector is subjected to feature dimension integration to form a multi-dimensional value evaluation vector.

[0117] S3.3, the multi-dimensional value evaluation vector is hashed and anchored to the blockchain to form an extended warranty smart contract.

[0118] In specific operation,

[0119] The multi-dimensional value evaluation vector is hashed using the SHA-256 algorithm, and further, the multi-dimensional value evaluation vector is serialized and encoded, for example, the numerical field of the multi-dimensional value evaluation vector is encoded as 0, and the classification label field of the multi-dimensional value evaluation vector is encoded as 1, to obtain a binary data stream; the binary data stream is hashed to generate a fixed length (such as 256 bits) hash value; the fixed length hash value is converted to hexadecimal to obtain a hash encryption string, which is used as the unique fingerprint of the multi-dimensional value evaluation vector and can uniquely identify the data integrity;

[0120] It should be noted that the hash digest refers to the process of one-way hash transformation of the binary data stream.

[0121] Then the hash encrypted string is anchored to the blockchain through a smart contract (such as an Ethereum contract written in Solidity), and further, the key metadata of the hash encrypted string (such as the evaluation vector version and the generation time) are recorded through the smart contract. The PBFT consensus mechanism is triggered synchronously to write the key metadata into the distributed ledger of the blockchain, and the blockchain transaction hash record is output.

[0122] It should be noted that the PBFT consensus mechanism is a practical Byzantine fault tolerance algorithm, which determines the storage of key metadata through a three-phase protocol (preparation, preparation, and submission).

[0123] To enhance traceability, the blockchain transaction hash record and the multi-dimensional value evaluation vector are stored synchronously through an offline database (such as MongoDB), and a chain-on-chain and off-chain index is established. According to the chain-on-chain and off-chain index, the blockchain transaction hash record is digitally signed and bound to prevent malicious tampering from end to end, and a secure blockchain transaction record is generated. The secure blockchain transaction record is executed through contract logic encapsulation to generate an extended smart contract.

[0124] It should be noted that contract logic encapsulation refers to the process of mapping business rules to secure blockchain transaction records.

[0125] S4, semantic analysis and instruction mapping of the extended smart contract are performed to generate storage strategy instructions; according to the storage strategy instructions, the encrypted device state vector is migrated and stored to form a storage location update record; the storage location update record is executed through Merkle tree evidence to obtain a storage management audit report.

[0126] Specifically, the following steps are included,

[0127] S4.1, semantic analysis of the extended smart contract is performed to form storage hierarchical parameters, and instruction mapping of the storage hierarchical parameters is performed to generate storage strategy instructions.

[0128] In specific operation,

[0129] Semantic analysis of the extended smart contract is performed, and further, state variable data of the extended smart contract is extracted and the weekly access frequency of the state variable data is counted. According to the weekly access frequency, cold and hot identification is performed, for example, state variable data with low weekly access frequency (such as <10 times) is identified as cold data, and state variable data with high weekly access frequency (such as ≥100 times) is identified as hot data, and cold and hot data identification is output.

[0130] The synchronous timing sampling is performed on the insurance smart contract to obtain a contract access timestamp; the sliding window integration is performed on the contract access timestamp to generate a contract access interval; the density threshold is applied to the contract access interval to perform density interval division, for example, when the contract access interval is greater than the density threshold, the high-frequency density interval is divided, and when the contract access interval is less than the density threshold, the low-frequency density interval is divided; the low-frequency density interval and the high-frequency density interval are classified and weightedly fused to obtain an activity distribution; and the hot and cold data identifiers and the activity distribution are integrated to obtain a storage grading parameter.

[0131] It should be noted that the density threshold is defined based on the dynamic volatility rate of the historical contract access interval, and the exemplary value range is [100 ms, 5 s].

[0132] Then, the storage grading parameter is mapped to an instruction, and further, the storage grading parameter is pre-defined for an operation type, for example, the hot data is defined as a memory operation type, the cold data is defined as a disk operation type, the low-frequency density interval is defined as a batch processing operation, and the high-frequency density interval is defined as a real-time processing operation; the pre-defined operation type is mapped to a command template to generate an instruction rule set; and the instruction rule set is formatted and packaged to output a storage strategy instruction.

[0133] It should be noted that the command template mapping refers to the process of template selection and parameter binding of the pre-defined operation type.

[0134] S4.2, according to the storage strategy instruction, the encrypted device state vector is migrated and stored across the chain to form a distributed storage credential, the distributed storage credential is recorded by hashing to form a storage location update record.

[0135] In specific operations,

[0136] The encrypted device state vector is migrated from a source chain (such as Ethereum) to a target chain (such as Polkadot) through a cross-chain protocol (such as IBC protocol), and further, the encrypted device state vector is locked on the source chain, and the encrypted device state vector is marked for integrity, and a marked encrypted data packet is output; then, the marked encrypted data packet is migrated to the target chain through a cross-chain relay, and the inter-chain state synchronization is synchronized to ensure complete data transmission;

[0137] It should be noted that the inter-chain state synchronization refers to the process of confirming the data integrity of the target chain for the marked encrypted data packet.

[0138] In the process of cross-chain migration, the marked encrypted data packet is stored by storing strategy instruction, and further, the storage shard parameter of the storage strategy instruction is extracted, the storage shard parameter is used for performing shard processing on the marked encrypted data packet, and an encrypted data shard set is obtained; the storage location registration is performed on the encrypted data shard set, the shard storage location information is obtained; the blockchain notarization is performed on the shard storage location information, and a distributed storage credential is generated;

[0139] It should be noted that the storage location registration refers to the process of storage node binding and location information recording of the encrypted data shard set.

[0140] Then the distributed storage credential is recorded by hashing, and further, the distributed storage credential is mapped by hashing to obtain a location credential hash value; the location credential hash value is timestamped to form a time location marker credential, and the time location marker credential is archived on the chain to obtain a storage location update record.

[0141] S4.3, the storage location update record is hashed and sampled randomly, a hash node set is obtained, and the hash node set is executed by the Merkle tree notarization to obtain a storage management audit report.

[0142] In specific operation,

[0143] The storage location update record is hashed and sampled randomly, a hash node set is obtained, and the hash node set is executed by the Merkle tree notarization to obtain a storage management audit report.

[0144] Then the hash node set is executed by the Merkle tree notarization, and further, the hash node set is sorted in descending order to obtain an ordered hash node, the ordered hash node is used as a leaf node of the Merkle tree, and the leaf node of the Merkle tree is recursively stacked to build a complete Merkle tree structure; the complete Merkle tree structure is written into the blockchain through the smart contract (such as Ethereum Solidity contract) to generate a Merkle tree notarization record;

[0145] The storage location update record and the Merkle tree notarization record are mapped by field to obtain an associated field set, the associated field set is audited and signed to form an audit intermediate record, and the audit intermediate record is executed by report field filling to output a storage management audit report.

[0146] The embodiment also provides a data storage management system for extended warranty service, comprising: a quantum encryption module, configured to perform wavelet packet transform and modal decomposition on a device running data set to obtain a time-frequency domain feature matrix; and a quantum chaotic encryption algorithm is used for performing random mask encryption on the time-frequency domain feature matrix to output an encrypted device state vector;

[0147] a health assessment module configured to input the encrypted device state vector into a federated global health model, apply differential privacy law for distributed feature learning at a privacy learning layer, and obtain a device feature latent variable; apply a secure multi-party computation protocol for cross-node parameter fusion at a global aggregation layer, and form a federated consensus parameter; and perform attention weighting and normalized activation on the device feature latent variable and the federated consensus parameter, and output a device health assessment report;

[0148] a contract generation module configured to extract a risk level weight, an access frequency prediction value, and a legal compliance weight of the device health assessment report, and perform dynamic weighted fusion and time sequence attenuation correction, to generate a multi-dimensional value evaluation vector; and perform blockchain anchoring on the multi-dimensional value evaluation vector, to form an extended warranty smart contract;

[0149] a storage execution module configured to perform semantic analysis and multi-objective optimization on the extended warranty smart contract, to generate a storage strategy instruction; perform migration storage on the encrypted device state vector according to the storage strategy instruction, to form a storage location update record; and perform Merkle tree storage evidence on the storage location update record, to obtain a storage management audit report.

[0150] The embodiment also provides a computer device suitable for the data storage management method for extended warranty services, including a memory and a processor; the memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions to implement the data storage management method for extended warranty services proposed in the above embodiment.

[0151] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is configured to perform wired or wireless communication with external terminals. The wireless communication can be achieved through WIFI, a carrier network, NFC (near field communication), or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball, or touchpad arranged on the shell of the computer device. The input device can also be an external keyboard, touchpad, or mouse, etc.

[0152] The embodiment also provides a storage medium on which a computer program is stored, the computer program being executed by a processor to implement the data storage management method for warranty service proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk, or an optical disk.

[0153] To sum up, the present application significantly enhances the anti-attack ability and privacy protection strength of data in the transmission and storage process by wavelet packet transform and quantum chaotic encryption algorithm for time-frequency domain feature extraction and random mask encryption, thereby guaranteeing the data credibility of the warranty service. At the same time, through the distributed feature learning and cross-node parameter fusion of the federal global health model, the precise evaluation of the device health state and the collaborative optimization of the consensus parameters are realized, the rationality of the resource allocation of the warranty service is enhanced, and the intelligent level and privacy security protection ability of the data storage management method are improved.

[0154] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A method for data storage management for warranty services, characterized by: comprising, performing wavelet packet transform and modal decomposition on the equipment operation data set to obtain a time-frequency domain feature matrix; applying a quantum chaotic encryption algorithm to perform random mask encryption on the time-frequency domain feature matrix to output an encrypted equipment state vector; inputting the encrypted equipment state vector into a federal global health model, and applying differential privacy law in a privacy learning layer for distributed feature learning to obtain equipment feature latent variables; applying a secure multi-party computation protocol in a global aggregation layer for cross-node parameter fusion to form a federal consensus parameter; performing attention weighting and normalization activation on the equipment feature latent variables and the federal consensus parameter to output an equipment health assessment report; extracting risk level weights, access frequency values and legal compliance weights from the equipment health assessment report, and performing dynamic weighted fusion and time sequence decay correction to generate a multi-dimensional value evaluation vector; anchoring the multi-dimensional value evaluation vector to a blockchain to form an extended warranty smart contract; performing semantic analysis and instruction mapping on the extended warranty smart contract to generate storage strategy instructions; migrating and storing the encrypted equipment state vector according to the storage strategy instructions to form a storage location update record; and performing Merkle tree evidence on the storage location update record to obtain a storage management audit report.

2. The data storage management method for warranty service of claim 1, wherein: The output encrypted equipment state vector specifically includes the following steps, performing wavelet packet transform on the equipment operation data set to obtain a multi-scale energy spectrum feature; and performing modal decomposition on the multi-scale energy spectrum feature to obtain a time-frequency domain feature matrix; applying a quantum chaotic encryption algorithm to perform phase space mapping on the time-frequency domain feature matrix to form a chaotic disturbance base, and performing random mask encryption on the chaotic disturbance base to output an encrypted equipment state vector.

3. The data storage management method for warranty service of claim 2, wherein: The federal global health model is specifically constructed as follows, building a privacy learning layer through a differential privacy neural network, and building a global aggregation layer through a federal average architecture; performing hierarchical integration stacking on the privacy learning layer and the global aggregation layer to construct a federal global health model.

4. The data storage management method for warranty service of claim 1, wherein: The equipment feature latent variables are obtained by specifically including the following steps, inputting the encrypted equipment state vector into the federal global health model, and performing gradient quantization on the encrypted equipment state vector in the privacy learning layer to form encrypted gradient parameters; applying differential privacy law to the encrypted gradient parameters to perform Gaussian noise disturbance and distributed feature learning to obtain equipment feature latent variables.

5. The data storage management method for warranty service of claim 4, wherein: The federal consensus parameter is formed by specifically including the following steps, performing aggregation calculation on the equipment feature latent variables in the global aggregation layer to form encrypted aggregated features; performing cross-node parameter fusion on the encrypted aggregated features according to a secure multi-party computation protocol to form a federal consensus parameter.

6. The data storage management method for warranty service of claim 1, wherein: The output equipment health assessment report specifically includes the following steps, applying a multi-head attention mechanism to perform attention weighting on the equipment feature latent variables and the federal consensus parameter to output health state features; performing normalization activation on the health state features to generate a health degree score, and structuring the health degree score to output an equipment health assessment report.

7. The data storage management method for warranty service of claim 1, wherein: The extended warranty smart contract is formed by specifically including the following steps, performing singular value decomposition on the equipment health assessment report to extract risk level weights, access frequency values and legal compliance weights; The entropy weight method is applied to dynamically weight and fuse the risk level weight, access frequency value and legal compliance weight, and time decay factor is used for time decay correction to generate a multi-dimensional value evaluation vector; The multi-dimensional value evaluation vector is hashed and anchored by the blockchain to form an extended warranty smart contract.

8. The data storage management method for warranty service of claim 7, wherein: The storage location update record specifically includes the following steps, The extended warranty smart contract is semantically analyzed to form a storage hierarchical parameter, and the storage hierarchical parameter is instruction mapped to generate a storage strategy instruction; According to the storage strategy instruction, the encrypted device state vector is migrated and stored across the chain to form a distributed storage voucher, and the distributed storage voucher is hashed to form a storage location update record.

9. The data storage management method for warranty service of claim 1, wherein: The storage management audit report is obtained, specifically including the following steps, The storage location update record is hashed and block-sampled to obtain a hash node set; The hash node set is executed by the Merkle tree to obtain a storage management audit report.

10. A data storage management system for warranty service based on the data storage management method for warranty service according to any one of claims 1 to 9, characterized in that: It includes, The quantum encryption module is used to perform wavelet packet transform and modal decomposition on the device running data set to obtain a time-frequency domain feature matrix; The quantum chaotic encryption algorithm is applied to randomly mask and encrypt the time-frequency domain feature matrix to output an encrypted device state vector; The health assessment module is used to input the encrypted device state vector into the federal global health model, and the privacy learning layer applies differential privacy method for distributed feature learning to obtain device feature latent variables; The global aggregation layer applies secure multi-party computation protocol for cross-node parameter fusion to form a federal consensus parameter; The device feature latent variables and the federal consensus parameter are subjected to attention weighting and normalized activation to output a device health assessment report; The contract generation module is used to extract the risk level weight, access frequency prediction value and legal compliance weight of the device health assessment report, and to dynamically weight and fuse and time decay correction to generate a multi-dimensional value evaluation vector; The multi-dimensional value evaluation vector is anchored by the blockchain to form an extended warranty smart contract; The storage execution module is used to semantically analyze and multi-objective optimize the extended warranty smart contract to generate a storage strategy instruction; according to the storage strategy instruction, the encrypted device state vector is migrated and stored to form a storage location update record; the storage location update record is executed by the Merkle tree to obtain a storage management audit report.

Images