Search

CN-122020691-A - Disaster recovery data sealing method and device based on space-time entropy coding

CN122020691ACN 122020691 ACN122020691 ACN 122020691ACN-122020691-A

Abstract

The application discloses a disaster recovery data sealing method and device based on space-time entropy coding, which comprises the steps of obtaining a weighted data generation time sequence of data to be sealed, carrying out multi-scale fuzzy entropy analysis to obtain a dynamic time entropy value vector, collecting physical position information of a storage environment where the data to be sealed is located, mapping the physical position information into a normalized space vector through a graph attention network, inputting the dynamic time entropy value vector and the normalized space vector into a space-time entropy field model for continuous time coupling to generate space-time fingerprints, carrying out mixed-diffusion-scrambling joint coding on the data to be sealed by using the space-time fingerprints as initial seeds and control parameters, driving a deflection tent chaotic mapping function for resisting quantum attack to generate sealed memory intermediate data, and embedding a true random seed and a polarization redundancy check block constructed based on the space-time fingerprints into the sealed memory intermediate data to generate a sealed memory. The application improves the safety and the integrity of disaster recovery data in the long-term sealing and storing process.

Inventors

  • XU SHENGWANG

Assignees

  • 深圳市数存科技有限公司

Dates

Publication Date
20260512
Application Date
20260407

Claims (10)

  1. 1. The disaster recovery data sealing method based on space-time entropy coding is characterized by comprising the following steps: Acquiring a weighted data generation time sequence of data to be sealed, and performing multi-scale fuzzy entropy analysis on the weighted data generation time sequence to obtain a dynamic time entropy value vector; Collecting physical position information of a storage environment where the data to be sealed are located, and mapping the physical position information into normalized space vectors through a graph attention network; inputting the dynamic time entropy vector and the normalized space vector into a space-time entropy field model for continuous time coupling, and generating space-time fingerprints; The space-time fingerprint is used as an initial seed and a control parameter, and a skewed tent chaotic mapping function for resisting quantum attack is driven to carry out mixed coding of confusion, diffusion and scrambling on the data to be sealed, so that sealed memory intermediate data is generated; Embedding the true random seed and the polarization redundancy check block constructed based on the space-time fingerprints into the middle data of the sealed memory to generate the sealed memory.
  2. 2. The method of claim 1, wherein the obtaining the weighted data generation time series of the data to be sealed and the multi-scale fuzzy entropy analysis of the weighted data generation time series to obtain the dynamic time entropy value vector comprise: extracting event occurrence time stamps from system logs and metadata of data to be sealed, giving different weights according to operation types, and generating a weighted data generation time sequence; adopting empirical mode decomposition to decompose the weighted data generation time sequence into a plurality of eigenmode function components; respectively calculating fuzzy entropy of each eigenmode function component in the scale factor range to form a multi-scale fuzzy entropy matrix; calculating the maximum Lyapunov exponent and the associated dimension of the weighted data generation time sequence as nonlinear dynamics characteristic parameters; Inputting the multi-scale fuzzy entropy matrix and the nonlinear dynamics characteristic parameter into a pre-trained characteristic fusion neural network, and outputting a dynamic time entropy value vector.
  3. 3. The method of claim 1, wherein the collecting physical location information of the storage environment in which the data to be sealed is located and mapping the physical location information into normalized spatial vectors through a graph attention network comprises: collecting physical position information of a storage environment in which the data to be sealed are located, wherein the physical position information comprises a logic storage node identifier, a cabinet global number and geographic longitude and latitude coordinates; Inputting the logic storage node identifier and the cabinet global number into a physical unclonable function circuit to generate a unique response code of the equipment; converting the geographical longitude and latitude coordinates into east-north-sky local Cartesian coordinates, and fusing attitude angles output by an inertial measurement unit to form a six-dimensional space state vector; Constructing an initial storage topological graph by taking a logic storage node as a graph vertex and a cabinet connection relationship as an edge; taking the unique response code of the equipment as a node attribute and taking the six-dimensional space state vector as an edge attribute; Updating the initial storage topological graph according to the node attribute and the edge attribute to obtain an updated storage topological graph; and inputting the updated storage topological graph to a graph attention network, aggregating neighbor node information through a multi-head attention mechanism, updating embedded representation of each node, and outputting a normalized space vector.
  4. 4. The method of claim 1, wherein said inputting the dynamic temporal entropy vector and the normalized spatial vector into a spatio-temporal entropy field model for continuous time coupling generates a spatio-temporal fingerprint, comprising: splicing the dynamic time entropy value vector and the normalized space vector into a joint input tensor; Constructing a space-time entropy field model, wherein the space-time entropy field model consists of a space-time encoder based on a neural ordinary differential equation and an entropy-space coupling layer; inputting the combined input tensor to a space-time encoder of the space-time entropy field model, and solving a neural ordinary differential equation by using the combined input tensor as an initial value through the space-time encoder in a latent space to obtain a latent variable track; performing space-time cross attention weighting on the latent variable track through the entropy-space coupling layer to obtain a coupling characteristic tensor; Inputting the coupling characteristic tensor into a cyclic neural network, and extracting the hidden state of the target time step; carrying out hash processing on the hidden state to generate a hash abstract; and splicing the hash abstract with the characteristic value of the Fisher information matrix of the latent variable track to generate a space-time fingerprint.
  5. 5. The method of claim 1, wherein the driving the quantum attack resistant skewed tent chaotic mapping function to perform the joint scrambling of the data to be sealed by using the spatiotemporal fingerprint as an initial seed and a control parameter, generating sealed memory intermediate data comprises: the data to be sealed is subjected to variable length blocking according to a content-aware Rabin fingerprint algorithm, and a data block sequence and a blocking index table are generated; The space-time fingerprint is used as an initial seed and a control parameter to drive a deflection tent chaotic mapping function for resisting quantum attack to generate a chaotic sequence; carrying out nonlinear function post-processing on the chaotic sequence to obtain a post-processed chaotic sequence; homogenizing the post-processed chaotic sequence, and extracting according to the bit to generate a chaotic key stream; Generating a reversible substitution box, a diffusion mask and three-dimensional chaotic scrambling mapping control parameters based on the chaotic key stream; Performing byte substitution obfuscation on each data block in the sequence of data blocks based on the reversible substitution box, generating an obfuscated sequence of data blocks; Performing cyclic left shift and exclusive-or operation on the confused data block sequence based on the diffusion mask to generate a diffused data block sequence; And performing three-dimensional space scrambling on the diffused data block sequence based on the three-dimensional chaotic scrambling mapping control parameters to generate sealed memory intermediate data.
  6. 6. The method of claim 1, wherein the embedding the true random seed and the polarization redundancy check block constructed based on the spatiotemporal fingerprint into the sealed-memory intermediate data, generating a sealed-memory, comprises: Generating a true random seed by a quantum random number generator based on a photon path interference principle; performing polarization coding on the middle data of the sealed memory based on the true random seed and the space-time fingerprint to construct a polarization redundancy check block; Packaging metadata blocks by using the hash digest of the true random seed, the polarization redundancy check block and polarization code freezing bit configuration information dynamically determined by the space-time fingerprint; And embedding the metadata block into a preset check area of the intermediate data of the sealed memory to obtain the sealed memory.
  7. 7. The method of claim 6, wherein said polarization encoding said sealed-memory intermediate data based on said true random seed and said spatiotemporal fingerprint, constructing a polarization redundancy check block, comprises: Dividing the middle data of the sealed memory into a plurality of data segments according to a preset block length, and distributing a corresponding segment identifier for each data segment; Performing secure hash operation on the space-time fingerprints to generate fingerprint hash values, and splitting the fingerprint hash values into a first subkey and a second subkey, wherein the first subkey is used for frozen bit selection, and the second subkey is used for coding disturbance; Taking the first subkey as a pseudo-random seed, and dynamically selecting a frozen bit position set from a standard polarization code mother matrix through a deterministic replacement algorithm; bit rearrangement is carried out on the true random seeds to generate auxiliary freezing masks, and the freezing bit position sets and the auxiliary freezing masks are subjected to bit exclusive OR to obtain freezing bit masks; updating a frozen bit mark of the standard polarization code mother matrix based on the frozen bit mask to obtain a polarization code generation matrix; performing bit exclusive OR on the second subkey and each data segment to obtain a disturbance data segment; Performing polarization coding based on the polarization code generation matrix and the disturbance data segments to obtain polarization coding data of each data segment; Splicing the polarization coding data of each data segment according to the segment identifier sequence to form an original check sequence; and performing bit exclusive OR operation on the original check sequence and a part of bit sequences of the true random seeds to generate a polarized redundancy check block.
  8. 8. The method of claim 7, wherein performing polarization encoding based on the polarization code generation matrix and the disturbance data segment to obtain polarization encoded data for each data segment comprises: dividing the disturbance data segment into a plurality of information sub-blocks according to the information bit length of the polarization code, and generating a sub-block serial number label for each information sub-block; performing Galois field multiplication on each information sub-block by utilizing the polarization code generation matrix to obtain an initial check symbol sequence; intercepting a random mask subsequence equal to an initial check symbol sequence from the true random seed by taking the sub-block sequence number tag as an address pointer; performing bit exclusive or on the initial check symbol sequence and the random mask subsequence to generate a mask check symbol; adding a cyclic redundancy checksum to the mask check symbol to form sub-block level polarization encoded data; And splicing the sub-block level polarization coding data according to the sequence of the sub-block serial numbers to obtain the polarization coding data of each data segment.
  9. 9. The method of claim 1, wherein embedding the true random seed and the polarization redundancy check block constructed based on the spatiotemporal fingerprint into the sealed-memory intermediate data, after generating a sealed-memory, further comprises: When an unpacking instruction is received, acquiring a real-time six-dimensional space state vector of a current unpacking environment and a unique response code of equipment, and constructing a current space-time context; inputting the current space-time context into a space-time entropy field model which is the same as the sealing stage, and generating a current space-time fingerprint; extracting hash digests of original space-time fingerprints, hash digests of true random seeds and polarization code freezing bit configuration information from the metadata blocks of the sealed memory; Carrying out the same hash operation on the current space-time fingerprint, and comparing the current space-time fingerprint with a hash abstract of the original space-time fingerprint; If the hash values are inconsistent, triggering a hardware-level self-destruction mechanism, clearing a key cache in the security chip, a sealed memory copy in the main memory and a log buffer area, and outputting an irreversible self-destruction signal; if the hash values are consistent, reconstructing an anti-quantum decryption key based on the current space-time fingerprint and the hash abstract of the true random seed, and performing inverse chaotic decoding on the intermediate data of the sealed memory to obtain decoded data; Initializing a polarization decoder by utilizing the frozen bit configuration information, executing successive elimination list decoding on the decoded data by combining the polarization redundancy check block, recovering the original data to be sealed, and outputting a seal-releasing success mark.
  10. 10. A disaster recovery data sealing device based on space-time entropy coding, characterized in that the device comprises: the analysis module is used for acquiring a weighted data generation time sequence of the data to be sealed, and carrying out multi-scale fuzzy entropy analysis on the weighted data generation time sequence to obtain a dynamic time entropy value vector; The mapping module is used for acquiring physical position information of a storage environment where the data to be sealed are located and mapping the physical position information into normalized space vectors through a graph attention network; The coupling module is used for inputting the dynamic time entropy vector and the normalized space vector into a space-time entropy field model for continuous time coupling, and generating space-time fingerprints; The encoding module is used for driving a deflection tent chaotic mapping function for resisting quantum attack to carry out mixed encoding of confusion, diffusion and scrambling on the data to be sealed by taking the space-time fingerprint as an initial seed and a control parameter, so as to generate sealed memory intermediate data; and the embedding module is used for embedding the true random seed and the polarization redundancy check block constructed based on the space-time fingerprints into the middle data of the sealed memory body to generate the sealed memory body.

Description

Disaster recovery data sealing method and device based on space-time entropy coding Technical Field The application relates to the technical field of data safety storage, in particular to a disaster recovery data sealing method and device based on space-time entropy coding. Background With the rapid development of information technology, data has become a core production factor. However, the risks of data loss or leakage rise dramatically due to the increasing threat of natural disasters, network attacks, human error operations, etc. Traditional disaster recovery or disaster recovery technologies, such as regular snapshot and remote copy, guarantee the usability of data to a certain extent, but lack a deep guarantee mechanism for data integrity and security when facing long-term sealing and storing requirements, and are difficult to cope with complex and changeable threat scenes. Disclosure of Invention The application mainly aims to provide a disaster recovery data sealing method and device based on space-time entropy coding, and aims to solve the technical problems of poor data integrity and security in a long-term sealing process in the prior art. In order to achieve the above objective, the present application provides a disaster recovery data sealing method based on space-time entropy coding, which comprises: the step of obtaining the weighted data generation time sequence of the data to be sealed, and carrying out multi-scale fuzzy entropy analysis on the weighted data generation time sequence to obtain a dynamic time entropy value vector, comprises the following steps: extracting event occurrence time stamps from system logs and metadata of data to be sealed, giving different weights according to operation types, and generating a weighted data generation time sequence; adopting empirical mode decomposition to decompose the weighted data generation time sequence into a plurality of eigenmode function components; respectively calculating fuzzy entropy of each eigenmode function component in the scale factor range to form a multi-scale fuzzy entropy matrix; calculating the maximum Lyapunov exponent and the associated dimension of the weighted data generation time sequence as nonlinear dynamics characteristic parameters; Inputting the multi-scale fuzzy entropy matrix and the nonlinear dynamics characteristic parameter into a pre-trained characteristic fusion neural network, and outputting a dynamic time entropy value vector. In an embodiment, the obtaining the weighted data generation time sequence of the data to be sealed and performing multi-scale fuzzy entropy analysis on the weighted data generation time sequence to obtain a dynamic time entropy value vector includes: extracting event occurrence time stamps from system logs and metadata of data to be sealed, giving different weights according to operation types, and generating a weighted data generation time sequence; adopting empirical mode decomposition to decompose the weighted data generation time sequence into a plurality of eigenmode function components; respectively calculating fuzzy entropy of each eigenmode function component in the scale factor range to form a multi-scale fuzzy entropy matrix; calculating the maximum Lyapunov exponent and the associated dimension of the weighted data generation time sequence as nonlinear dynamics characteristic parameters; Inputting the multi-scale fuzzy entropy matrix and the nonlinear dynamics characteristic parameter into a pre-trained characteristic fusion neural network, and outputting a dynamic time entropy value vector. In an embodiment, the collecting physical location information of the storage environment where the data to be sealed is located, and mapping the physical location information into a normalized space vector through a graph attention network includes: collecting physical position information of a storage environment in which the data to be sealed are located, wherein the physical position information comprises a logic storage node identifier, a cabinet global number and geographic longitude and latitude coordinates; Inputting the logic storage node identifier and the cabinet global number into a physical unclonable function circuit to generate a unique response code of the equipment; converting the geographical longitude and latitude coordinates into east-north-sky local Cartesian coordinates, and fusing attitude angles output by an inertial measurement unit to form a six-dimensional space state vector; Constructing an initial storage topological graph by taking a logic storage node as a graph vertex and a cabinet connection relationship as an edge; taking the unique response code of the equipment as a node attribute and taking the six-dimensional space state vector as an edge attribute; Updating the initial storage topological graph according to the node attribute and the edge attribute to obtain an updated storage topological graph; and inputting the updated storage topological graph