Disorder factor self-organizing bidirectional computing data security processing method and system

Abstract

The present application relates to the field of information security and data encryption technology, and discloses a disordered factor self-organizing bidirectional calculation data security processing method and system, which sequentially comprises data vectorization, double-layer Levy segmentation, hyper-chaotic transformation, non-bijection decomposition and distributed storage, original data is converted into a set of disordered factors and distributed to multiple geographically distributed edge nodes by constructing an irreversible disturbance model, and when a legal recovery credential is available, approximate data recovery is completed through an optimization and reorganization mechanism with a gradient regular term. By constructing a factor decomposition mechanism with irreversible disturbance, combining distributed mapping storage with geographical distribution, and using a limited recovery algorithm with regular modulation, the physical isolation of data, recoverability after disturbance and long-term confidentiality are ensured without relying on keys, and the overall security and recovery robustness of the system in the face of node failure, data leakage and quantum cracking risks in the edge computing scenario are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of information security and data encryption technology, specifically to a method and system for secure processing of data using unordered factor self-organizing bidirectional computation. Background Technology

[0002] As network computing systems continue to migrate to the edge, the generation, processing, and application of data are gradually moving away from central servers, exhibiting highly distributed characteristics. Especially in scenarios such as smart cities, telemedicine, and the Industrial Internet of Things (IIoT), a large amount of sensitive information is generated directly at edge nodes and temporarily stored, placing higher demands on data privacy and availability. Under this trend, how to achieve secure, efficient, and controllable data persistence and recovery mechanisms has become a pressing technical challenge that needs to be addressed in edge computing applications.

[0003] Current technologies generally rely on traditional encryption algorithms to protect edge data, with the core being the control of data access and decryption through key mechanisms. However, this approach inherently carries a single point of failure; if the key is stolen or mismanaged, the overall system security will be compromised. Furthermore, the centralized scheduling architecture's insufficient recovery capabilities in the event of node failures or attacks can easily lead to unrecoverable data. In addition, recovery mechanisms for disturbed or incomplete data are still geared towards precise reconstruction, lacking resilient solutions to maintain data availability under complex conditions, thus limiting recovery capabilities.

[0004] Faced with emerging attack methods such as quantum computing, the vulnerabilities of traditional security models are becoming increasingly apparent, and existing encryption systems based on mathematical complexity are gradually facing the risk of being breached. Therefore, building innovative mechanisms that do not require key dependencies, support irreversible perturbations, and enable controlled recovery under limited conditions has become a key direction for improving the security and reliability of edge data. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for secure processing of self-organized bidirectional computational data in disordered factors, which solves the problems of weak data storage security, strong key dependence, and uncontrollable recovery of disturbed data in edge computing environments.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for securely processing self-organized bidirectional computational data of disordered factors, comprising the following steps:

[0007] S1. Data vectorization: Converting raw data into a decimal integer vector;

[0008] S2. Primary Lévy Partition: Randomly partition the decimal vector based on the Lévy flight distribution to generate a first-level sub-vector set;

[0009] S3, Hyperchaotic Transformation: Nonlinear mapping of the first-level sub-vector set is performed on a three-dimensional tensor coupled chaotic system to output a chaotic mapped vector set;

[0010] S4. Secondary Lévy partitioning: Perform a second Lévy distribution partitioning on the chaotic mapping vector set to generate a second-level sub-vector set;

[0011] S5. Non-bijective decomposition: Apply an irreversible function transformation to the second-level sub-vector set to generate an unordered factor set that satisfies the entropy threshold;

[0012] S6. Distributed storage: Distribute the unordered factor set to geographically distributed edge nodes;

[0013] S7. Restricted Recovery: When valid recovery credentials are verified, optimized recombination is performed based on the unordered factor set.

[0014] Preferably, the Levy flight distribution in step S2 adopts a stable distribution with parameter α = 1.5, and its step size generation formula is:

[0015]

[0016] Where μ and v are independent and identically distributed standard normal random variables, α is the shape parameter of the Lévy distribution, and s is the Lévy flight step size;

[0017] The randomly selected division points in step S2 are determined through modular arithmetic. Dynamically determined, where s k Let m be the step size of the Lévy flight generated in the kth iteration, m be the dimension of the original decimal vector, and p be the step size of the Lévy flight. i Let i be the i-th dynamic segmentation point.

[0018] Preferably, the dynamic equations of the three-dimensional tensor-coupled chaotic system in step S3 are as follows:

[0019]

[0020] Where γ > 3.7, W is the dynamic weight matrix, the chaotic sequence is generated iteratively using the fourth-order Runge-Kutta method, x, y, and z are state variables, and γ is the control parameter. 1 is the dissipation coefficient, 28 is the critical parameter, and 10 is the damping coefficient.

[0021] Preferably, the irreversible function transformation in step S5 includes:

[0022] Perform a circular left shift x << < s for each subvector i Displacement s i Taken from a chaotic sequence;

[0023] Applying a random mask matrix M iPerform bitwise XOR ⊙M i ;

[0024] Combination operations are

[0025] Preferably, step S5 further includes entropy feedback control:

[0026] The formula for calculating the information entropy of the disordered factor is as follows:

[0027]

[0028] Where H is the entropy value, v is the data value, p(v) is the probability distribution, and log2 is the logarithmic basis;

[0029] When H > 0.01 bits, return to S4 for further splitting.

[0030] Preferably, the optimized recombination in step S7 is achieved by solving the following formula:

[0031]

[0032] in, To optimize the variables, f is a non-bijective function, F i Here are known parameters, and N is a scalar parameter. For gradient operators, Let ||·||1 be the L2 norm, ||·||1 be the L1 norm, and 0.1 be the regularization coefficient.

[0033] The calculation terminates when the recovery error rate (ERR) is ≥ 37%, where:

[0034]

[0035] Where m is a scalar parameter. To recover the data, x j This is the original data.

[0036] A second aspect of the present invention provides a data security processing system for unordered factor self-organized bidirectional computation, used to execute the aforementioned data security processing method for unordered factor self-organized bidirectional computation, comprising:

[0037] The data preprocessing module is used to convert raw data into a standard digital vector format;

[0038] The chaos calculation module is used to perform Lévy flight distribution segmentation and hyperchaotic transformation processing;

[0039] The non-bijective decomposition module is used to perform irreversible function transformations on data fragments;

[0040] Edge storage networks are used to manage the distributed storage and location obfuscation of fragmented data.

[0041] Data recovery engine for performing limited-optimization reorganization based on storage fragmentation.

[0042] Preferably, the chaos calculation module includes:

[0043] The Lévy parameter configuration unit is used to dynamically configure the α=1.5 parameter of the Lévy flight distribution, controlling the randomness characteristics of data segmentation;

[0044] Tensor coupling operation unit, used for real-time calculation of tensor coupling terms in three-dimensional chaotic system;

[0045] A chaos iteration accelerator for hardware acceleration of the fourth-order Runge-Kutta method for solving chaotic differential equations.

[0046] Preferably, the edge storage network includes:

[0047] Fragment tag encryption unit, used to generate unpredictable UUID tags based on quantum random numbers;

[0048] The correlation analyzer is used to detect and block the storage of logically related factors on the same physical node;

[0049] The dynamic migration engine is used to periodically trigger random migrations of fragment physical locations.

[0050] Preferably, the data recovery engine includes:

[0051] The gradient calculation unit is used for hardware-accelerated gradient norm calculation, enabling rapid evaluation of the regularization term;

[0052] An error monitor is used to monitor the recovery error rate in real time, and the calculation is forcibly terminated when the ERR is greater than or equal to 37%.

[0053] The projection optimizer is used to perform steepest descent iterations within a constrained solution space to find local optima.

[0054] This invention provides a method and system for securely processing self-organized bidirectional computational data of disordered factors. It offers the following advantages:

[0055] 1. This invention adopts a geographically distributed mapping storage technology scheme based on a perturbed, disordered factor set, which achieves physical decoupling and logical isolation of data among multiple edge nodes. This achieves the technical effect of redundancy and resilience in different network topologies. Compared with the problem that existing centralized architectures cannot cope with local node failures, this solution effectively avoids the risk of single point of failure and improves the availability of data in regional disaster scenarios.

[0056] 2. This invention employs a variational optimization model with L1 sparse regularization and L2 error constraints to achieve limited-precision reconstruction of perturbed data. It achieves the technical effect of performing data recovery in a high-interference environment without relying on the original plaintext or key. Compared with the existing recovery mechanism based on key-based accurate restoration, this invention solves the problems of complex key management and insufficient uniqueness of recovery path, and enhances the elasticity of data recovery under uncontrollable conditions.

[0057] 3. By constructing an irreversible non-bijective perturbation function system and combining it with a mapping index mechanism, this invention ensures at the structural layer that data fragments cannot be deduced from their original form. This achieves the technical effect that even if an attacker obtains all edge node data, they cannot restore the valid content. Compared with existing methods that rely on the strength of encryption algorithms, this invention solves the shortcomings of traditional public key systems being easily cracked under quantum computing models, and significantly improves long-term data confidentiality and resistance to algorithm leakage. Attached Figure Description

[0058] Figure 1 This is a flowchart of the method of the present invention;

[0059] Figure 2 This is a flowchart of the system framework of the present invention;

[0060] Figure 3 This is a framework diagram of the chaos calculation module of the present invention;

[0061] Figure 4 This is a framework diagram of the edge storage network of the present invention;

[0062] Figure 5 This is a framework diagram of the data recovery engine of this invention. Detailed Implementation

[0063] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] Please see the appendix Figure 1 This invention provides a method for securely processing self-organized bidirectional computational data of disordered factors, comprising the following steps:

[0065] S1. Data vectorization: Converting raw data into a decimal integer vector;

[0066] This step involves converting the raw data into a decimal integer vector for subsequent encryption processing. This process includes extracting the original information from the data and transforming it into a computationally pleasing standard format to facilitate the execution of subsequent encryption algorithms.

[0067] First, the original data X = {x1, x2, ..., x} m The data needs to be quantified, and each data item x needs to be converted into a numerical value. i Convert to a decimal integer. This process is preferably based on specific encoding rules (such as ASCII or other encoding formats) to ensure that the converted data can be used as input for subsequent encryption operations.

[0068] The formula derivation process is as follows:

[0069] Vectorization of raw data: Let the raw data be X = {x1, x2, ..., x...} m}, where x i Let x be the i-th element of the original data. i Converted to decimal integers using standard encoding, vector X is obtained. int ={x′1,x′2,...,x′ m}, where each x′ i For x i The converted decimal value.

[0070] Numericalization steps: For example, for character data, if ASCII encoding is used, then each character x... i The transformation relationship is: x′ i =ASCII(x i For binary data, a direct conversion rule is used to ensure that each data point x i It is mapped to the corresponding decimal integer.

[0071] This vectorization process converts each item of data into a numerical form, ensuring that subsequent chaotic encryption operations can be performed correctly.

[0072] Implementation process description:

[0073] During vectorization, the original data is converted into integer values ​​one by one according to pre-defined encoding rules (such as ASCII, UTF-8, etc.). These integer values ​​will serve as input for subsequent algorithms, ensuring that the data can be successfully encrypted through mathematical models and algorithmic processes.

[0074] In physical implementation, data is converted into integer vectors through hardware or software encoding / decoding modules, completing the data preprocessing process. After data conversion, the encryption step size can be generated using a random number generator or chaotic sequence to achieve random data partitioning and encryption.

[0075] This implementation effectively transforms raw data into a standardized digital format, laying the foundation for subsequent encryption processing. This vectorization process ensures that subsequent mathematical models and encryption algorithms can adapt to various types of input data, improving the applicability and scalability of the encryption system.

[0076] S2, Primary Lévy Partition: Randomly partition the decimal vector based on the Lévy flight distribution to generate a first-level sub-vector set;

[0077] In one embodiment of the present invention, the decimal integer vector X = x1, x2, ..., x2 obtained in step S1 is further... m The primary subvector segmentation is achieved by generating random step sizes and calculating dynamic indexes based on the Lévy flight distribution.

[0078] First, a stable Lévy distribution with parameter α = 1.5 is used to generate a random step-size sequence. The step-size generation formula is as follows:

[0079]

[0080] in, Let be independent and identically distributed standard normal random variables; α = 1.5 represents the shape parameter of the Lévy distribution; s is the Lévy flight step size.

[0081] Subsequently, the step-size sequence is accumulated, and the split point position is determined by modulo operation. The formula for calculating the split point is as follows:

[0082]

[0083] Where, p i s represents the index position of the i-th split point. k Let m be the step size of the Levi flight generated for the kth time, and m be the dimension of the original decimal vector.

[0084] To ensure the validity and uniqueness of the split points, the calculated split points {p1, p2, ...} are sorted and duplicates are removed, resulting in the final split point sequence P = {p1, p2, ..., p...}. r},satisfy:

[0085] 0 <p1<p2<…<p r <m;

[0086] Based on the sequence of split points p, the input vector X is divided into r+1 subvectors:

[0087]

[0088] Where p0 = -1 represents a virtual starting point; p r+1 =m-1 represents the virtual endpoint; each subvector X (i)The data segments are complete and do not overlap; X (i) This is the i-th sub-vector segment; This means combining all the sub-vector segments to form the complete original vector X, satisfying the conditions of no overlap and no omission.

[0089] The aforementioned primary Lévy partitioning process is executed by the chaos computation module in the system. This module consists of the following three physical units:

[0090] Levy parameter configuration unit: used to set and lock the shape parameter α = 1.5, control the long-tail characteristics of the step distribution, and determine the sparsity of the jump distribution;

[0091] Tensor-coupled operation unit: In conjunction with the distributed generation logic, it performs tensor-level modeling of the interaction between step-size sequences and time series, and provides a mapping interface for step-size sequence accumulation and propagation;

[0092] Chaos Iteration Accelerator: Accelerates the computation of the coupled process of chaotic state variables and distribution changes through the fourth-order Runge-Kutta method, providing high concurrency generation support.

[0093] The Levy parameter configuration unit provides a system-level control interface for coordinating with the main control scheduling module to set distributed parameters. The tensor coupling operation unit is nested within the multi-channel chaotic tensor intermediate processing layer, synchronously constructing the index state in conjunction with the real-time input step flow.

[0094] The chaotic iteration accelerator integrates a hardware floating-point arithmetic unit to perform real-time differential advancement of the index update path under the influence of the step size, providing periodic drive for step size accumulation and split point calculation.

[0095] The entire chaos calculation module is directly coupled to the front-end data preprocessing module of the system through a logical connection interface. It receives vector X as input data stream and outputs the segmented sub-vector set to the subsequent non-bijective decomposition module.

[0096] This implementation structure ensures the dynamic representation of the Lévy distribution characteristics during the segmentation process, enhances the unpredictability of subvectors in terms of dimension, position, and structure, and strengthens the system's resistance to statistical recombination attacks.

[0097] S3, Hyperchaotic Transformation: Nonlinear mapping of the first-level sub-vector set is performed through a three-dimensional tensor coupled chaotic system to output a chaotic mapped vector set;

[0098] In an embodiment of the present invention, after the initial Lévy segmentation in step S2 is completed, the resulting sub-vector set {X(i)}i=0r is introduced into a three-dimensional tensor-coupled chaotic system to perform a nonlinear mapping operation to generate a chaotic mapping vector set.

[0099] First, a continuous-time dynamic model of the system is established, described as follows:

[0100]

[0101] Where x, y, and z are state variables; γ>3.7 is a control parameter; and W is a dynamic weight matrix constructed from the input sub-vectors. This indicates a tensor multiplication operation, used to enhance coupling between system variables; coefficients 10 and 28 are used in conjunction with... These are the damping factor, critical control parameter, and dissipation coefficient, respectively.

[0102] The construction process of the dynamic weight tensor W is as follows:

[0103] W (i) =reshape(X) (i) ,d1×d2×d3);

[0104] Among them, X (i) Let d1, d2, d3 be the i-th input subvector; d1, d2, d3 are tensor dimension parameters, satisfying d1·d2·d3=|X (i) |, to ensure the integrity of tensor reconstruction; W (i) This refers to the dynamic tensor used in the current iteration.

[0105] The above system of differential equations is numerically integrated using the fourth-order Runge-Kutta method, with the specific discrete iterative form as follows:

[0106]

[0107] Where, x n y n and z n Let x be the value of the state variable during the nth iteration. n+1 y n+1 and z n+1 For the (n+1)th iteration, the state variable update value is given, and k1, l2, and m1 represent the Runge-Kutta intermediate slope values ​​of the differential equations for x, y, and z, respectively.

[0108] The intermediate variable is calculated as follows:

[0109]

[0110] Among them, f x ,f y ,f z Let l1, l2, l3, and l4 represent the function terms applied to x, y, and z in the system equations, respectively, where h is the time step, and l1, l2, l3, and l4 are the function terms applied to x, y, and z respectively. The four estimated slopes, m1, m2, m3, and m4, are for... The four estimated slopes, k1, k2, k3, and k4, are for... The four estimated slopes.

[0111] After each iteration, the system obtains a sequence of state variables. The combined result of the three-channel chaotic mapping is constructed as follows:

[0112]

[0113] Among them, Y (i) Let represent the chaotic output after mapping the i-th subvector, which serves as the input for subsequent non-bijective decomposition steps. and Let be the sequence of state variables of the i-th subvector at the time of iteration termination, and concat(·) is the concatenation operation.

[0114] The above transformation process is completed at the system level by the chaos calculation module, whose module structure is as follows:

[0115] Levy parameter configuration unit: provides α=1.5 setting to support front-stage random segmentation;

[0116] Tensor-coupled computation unit: Constructs a weight tensor W based on the input subvectors and performs... Operations;

[0117] Chaos Iteration Accelerator: Based on hardware implementation of the fourth-order Runge-Kutta algorithm, used to solve the equations of continuous-time dynamic model in real time.

[0118] The modules are connected in parallel via a data bus to ensure efficient transmission between input vector switching and tensor generation, and to support the continuous evolution of state variables of chaotic systems in high-dimensional space.

[0119] The tensor-coupled chaotic mapping mechanism in this embodiment realizes a high-dimensional nonlinear mapping of the primary segmentation vector, effectively improving the irreversibility and complexity of data in the transformation space and enhancing the defense against statistical reconstruction attacks.

[0120] S4. Secondary Lévy Partition: Perform a second Lévy distribution partition on the chaotic mapping vector set to generate a second-level sub-vector set;

[0121] In an embodiment of the present invention, after completing the chaotic mapping in step S3, the system obtains a set of chaotic mapping vectors. This vector set has a complex structure, increased dimensionality, and exhibits high nonlinearity and high entropy.

[0122] To further enhance the irreversibility and randomness of the data, a secondary Lévy distribution partition needs to be performed on the vector set before entering the non-bijective decomposition to form a second-level sub-vector set, which serves as the basic input for data scrambling in the next stage.

[0123] First, a Lévy flight distribution model is established, and a probabilistic control mechanism for secondary segmentation is defined. The Lévy distribution is a heavy-tailed distribution, and its probability density function is defined as follows:

[0124]

[0125] Where β is the offset parameter, set to zero to indicate a symmetrical distribution; μ is the offset starting point, generally set to the position where the vector is smallest and most segmentable; c is the scale parameter, affecting the distribution width, and can be combined with Y. (i) Length adaptive setting

[0126] The system is configured with a Lévy distributed controller, for each Y (i) The process of performing a random sliding window segmentation operation based on this distribution is as follows:

[0127] Initialize variable index j = 0, and set vector length to N = |Y (i) |

[0128] Generate a random length l from the Lévy distribution. j ~L(s; α, β), perform the rounding operation.

[0129] Extract sub-segments Generate new subvectors

[0130] Update index j = j + l j Repeat steps S2 to S4 until j ≥ N.

[0131] The above operations can be denoted as a piecewise function:

[0132] wherel j ~L(s; α=1.5);

[0133] Among them, Y (i) This is the i-th chaotic mapping vector output from the previous stage. For the j-th secondary subvector generated in this stage, segment(·) is the piecewise function, dividing the source vector by length l. j Extract the segment.

[0134] To avoid information fragmentation due to excessively small segment lengths or impact on diversity due to excessively large segment lengths, the system introduces a threshold constraint:

[0135]

[0136] Among them, l min With l max The system is a configurable parameter, preferably automatically adjusted by the data preprocessing module based on the dimensions of the original data.

[0137] The physical implementation of this process is accomplished by the interface component between the chaos calculation module and the non-bijective decomposition module, including:

[0138] The distribution generation unit connects the Lévy parameter configuration unit and the tensor output port, and is used to output a random length sequence that meets the parameter settings;

[0139] The segmented controller is connected to the mapping result cache unit and is used to dynamically extract sub-vectors according to their length and write them to the intermediate storage array.

[0140] The output reconstruction department is used to perform unified structured management of the extracted results and generate secondary sub-vector sets.

[0141] Through this step, the system can further perform high-dimensional perturbations based on the enhanced chaotic structure, construct a multi-scale sub-vector set with heavy-tailed distribution characteristics, provide highly diverse data inputs for the next non-bijective transformation, and effectively improve the irreversibility and encryption strength of the overall system.

[0142] S5. Non-bijective decomposition: Apply an irreversible function transformation to the second-level sub-vector set to generate an unordered factor set that satisfies the entropy threshold.

[0143] In this embodiment, the secondary sub-vector set obtained in step S4 is completed. The inputs are sequentially fed into a non-bijective transform logic, which performs multiple irreversible perturbation operations, ultimately outputting an unordered set of factors. As the core input of the encryption system.

[0144] First, for each input subvector Assigning chaotic control sequences Where K represents the number of perturbation iterations, the value of which depends on the system configuration and is generally not less than 3.

[0145] The chaotic control sequence comes from the three-channel chaotic state vector Y output in step S3. (i) The result obtained after submap transformation is:

[0146]

[0147] in, For the i-th chaotic sequence, the k-th term, Let be the cyclic displacement of the i,j subvector in the k-th round of operation, with a value range of [0,7]. mod(·,8) is the modulo function to ensure that the displacement does not exceed the single-byte limit.

[0148] Subsequently, the system generates a random mask matrix for each round of operations. Its generation is jointly determined by a pseudo-random generator (PRG) and a perturbation key seed, ensuring that each execution is unique.

[0149] The combined non-bijective perturbation function is defined as follows:

[0150]

[0151] in, To represent each byte of the subvector being shifted left bit by bit The ⊙ symbol represents a bitwise AND operation or a bitwise mask. To represent the bitwise XOR operation, used to fuse the results of multiple rounds of perturbation, f(·) is the final non-bijective perturbation function.

[0152] To ensure the irreversibility of the output, all All have the same They share the same dimensions and are generated independently in each round, making them non-reusable.

[0153] After the perturbation is complete, the system enters the entropy feedback module to calculate each disordered factor vector. Shannon entropy:

[0154]

[0155] in, Let v be the Shannon entropy value of the unordered factor vector, v be the byte value, and p(v) be the probability of the value v appearing in the unordered factor vector.

[0156] The entropy value is set by the system threshold H. min Perform a determination. If the following conditions are met:

[0157]

[0158] Among them, H min The minimum entropy threshold set for the system is preferably set to 0.01 bits.

[0159] If the disturbance is considered valid, the output will be... If the condition is not met, return to step S4 and re-execute the secondary Levy split.

[0160] This process is completed by the non-bijective decomposition module. Through this step, the source data subvectors are subjected to strong nonlinear perturbation and structural scrambling to ensure that the generated results do not have an invertible mapping relationship. At the same time, the output quality is dynamically controlled by the entropy feedback mechanism, providing a high-intensity, high-obfuscation input foundation for subsequent embedding encoding or channel mapping.

[0161] S6. Distributed storage: Distributes unordered factor sets to geographically distributed edge nodes;

[0162] After completing step S5, the system obtains an unordered set of factors. This collection is highly obfuscated and irreversible, making it suitable for implementing disaster recovery distributed storage in edge environments.

[0163] First, establish a geographic node topology model G = (V, E), where:

[0164]

[0165] Where V is the set of all edge storage nodes, E is the set of reachable communication links between nodes, and M is the total number of edge nodes.

[0166] Secondly, through the node allocation function This function maps unordered factors to specific edge nodes. It relies on hash indexes and node state feedback to make decisions, and is defined as follows:

[0167]

[0168] Where h(·) is a hash function, taking arbitrary length input data and outputting a fixed-length index, v * The edge node of the selected storage target.

[0169] The system further analyzes each Compile segmented index labels It contains the original subvector number, perturbation round, and mapping node position, used for subsequent lookup and recovery. The tag structure is as follows:

[0170]

[0171] Among them, ID i,j The atomic vector sequence number is represented by K, which is the perturbation round number, consistent with the perturbation number in step S5. * The identifier of the edge node mapped by the current subvector.

[0172] Then via encrypted communication channel The data is distributed to the corresponding edge nodes, and the system protects the channel using an end-to-end security protocol (such as DTLS or TLS) to ensure that the data is not tampered with or eavesdropped on during transmission.

[0173] Through the above process, the system can achieve distributed splitting and storage of highly obfuscated data without relying on a central control node, effectively improving the system's robustness against deletion attacks, traffic monitoring, and node failures.

[0174] In actual deployment, the system can further incorporate the node load factor λ(v) m Dynamic remapping is performed to achieve load balancing control. This factor is defined as:

[0175]

[0176] The system is based on λ(v) m It determines whether to perform node replacement or multi-replica redistribution, thereby realizing dynamic avoidance of high-risk nodes and replica disaster recovery mechanism.

[0177] By implementing this step, disordered data fragments after irreversible disturbances can be mapped to a distributed edge storage cluster in a structured manner, providing a storage mechanism with encryption, discreteness, and traceability, meeting the comprehensive needs for edge data anti-tampering and disaster recovery in high-security scenarios.

[0178] S7. Restricted Recovery: When validating a valid recovery credential, perform optimized reorganization based on an unordered factor set.

[0179] After completing S6, the system stores the scrambled unordered factor set and its index labels through edge nodes. To recover the original data, the system first verifies the recovery credentials, including access authorization, data hash verification, and node consistency verification, to ensure that the requester has legitimate permissions.

[0180] After the credentials are verified, the system enters the optimization and reorganization process, with the goal of resolving the unordered factor set. Reconstructing an approximate original data vector This is achieved by solving the following objective function:

[0181]

[0182] in, To optimize variables, F i The parameters are known and stored in the edge nodes, where N is a scalar parameter. For gradient operators, it means that for... The first-order gradient is used to constrain the smoothness of the reconstruction. Let f be the L2 norm, ||·||1 be the L1 norm, 0.1 be the regularization coefficient, and f be a non-bijective function.

[0183] The above optimization problem adopts a variational solution strategy, using an alternating minimization method for iterative approximation, and leveraging the computing resources of edge nodes to accelerate the optimization process in parallel.

[0184] The system evaluates the currently recovered data after each iteration. The recovery error rate (ERR) between the original data and the original data is defined as follows:

[0185]

[0186] Where m is a scalar parameter. To recover the data, x jThis is the original data.

[0187] If ERR ≥ 37%, the system determines that the current data cannot be effectively recovered, terminates the optimization process, and sends a recovery failure signal.

[0188] Through this implementation method, the system can reconstruct the original input data within a limited error range, effectively realizing the "conditionally reversible" recovery function for high-security scenarios. At the same time, the irreversibility of the perturbation function ensures that unauthorized recovery cannot obtain usable information.

[0189] In specific deployments, support for multi-round, multi-replica joint reconstruction strategies can be further expanded to enhance fault tolerance in the event of partial node loss, or the reconstruction of trusted blockchain record paths can be introduced to improve operational traceability.

[0190] The data security processing system for unordered factor self-organized bidirectional computation described below can be referred to in correspondence with the data security processing method for unordered factor self-organized bidirectional computation described above.

[0191] Please see the appendix Figure 2 A data security processing system for unordered factor self-organizing bidirectional computation, used to execute the aforementioned data security processing method for unordered factor self-organizing bidirectional computation, including:

[0192] The data preprocessing module is used to convert raw data into a standard digital vector format;

[0193] The chaos calculation module is used to perform Lévy flight distribution segmentation and hyperchaotic transformation processing;

[0194] The non-bijective decomposition module is used to perform irreversible function transformations on data fragments;

[0195] Edge storage networks are used to manage the distributed storage and location obfuscation of fragmented data.

[0196] A data recovery engine for performing limited optimized reorganization based on storage fragmentation.

[0197] Please see the appendix Figure 3 The chaos calculation module includes:

[0198] The Lévy parameter configuration unit is used to dynamically configure the α=1.5 parameter of the Lévy flight distribution, controlling the randomness characteristics of data segmentation;

[0199] Tensor coupling operation unit, used for real-time calculation of tensor coupling terms in three-dimensional chaotic system;

[0200] A chaos iteration accelerator for hardware acceleration of the fourth-order Runge-Kutta method for solving chaotic differential equations.

[0201] Please see the appendix Figure 4Edge storage networks include:

[0202] Fragment tag encryption unit, used to generate unpredictable UUID tags based on quantum random numbers;

[0203] The correlation analyzer is used to detect and block the storage of logically related factors on the same physical node;

[0204] The dynamic migration engine is used to periodically trigger random migrations of fragment physical locations.

[0205] Please see the appendix Figure 5 The data recovery engine includes:

[0206] The gradient calculation unit is used for hardware-accelerated gradient norm calculation, enabling rapid evaluation of the regularization term;

[0207] An error monitor is used to monitor the recovery error rate in real time, and the calculation is forcibly terminated when the ERR is greater than or equal to 37%.

[0208] The projection optimizer is used to perform steepest descent iterations within a constrained solution space to find local optima.

[0209] The system in this embodiment can be used to execute the above method embodiments, and its principle and technical effect are similar, so they will not be described again here.

[0210] Test Example: Comparative Experiment of the Data Disturbance Storage and Restricted Recovery Method Based on the Invention and AES-256 Encryption Technology

[0211] Experimental objective:

[0212] The data perturbation storage and restricted recovery method proposed in this invention is compared with the traditional AES-256 encryption algorithm in terms of anti-attack capability, encryption speed, and performance under quantum computing attacks.

[0213] Experimental environment and equipment:

[0214] Hardware environment:

[0215] Processor: Intel i910 core 3.6GHz;

[0216] Memory: 64GB DDR4;

[0217] Storage: 1TB SSD;

[0218] Operating system: Linux Ubuntu 20.04;

[0219] Software environment:

[0220] Implementation of this invention: Perturbation storage, regularization optimization recovery algorithm and quantum attack simulation are implemented using Python;

[0221] AES-256 encryption algorithm: Encryption and decryption operations are performed using the OpenSSL library.

[0222] Experimental procedure:

[0223] Experimental data preparation:

[0224] Dataset: Use a 1GB text file containing randomly generated character data.

[0225] Experimental plan:

[0226] The 1GB of data was processed in two ways:

[0227] Option 1: Use the perturbation storage and restricted recovery method of the present invention for data storage and recovery.

[0228] Option 2: Use the AES-256 encryption algorithm to encrypt and store the data.

[0229] Step 1: Data Encryption and Storage

[0230] Solution 1 (Method of the present invention):

[0231] Convert the original data into a decimal integer vector.

[0232] The data is randomly partitioned based on the Lévy distribution.

[0233] Nonlinear mapping of subvector sets is performed using a three-dimensional tensor-coupled chaotic system.

[0234] Perform a quadratic Lévy distribution partitioning on the chaotic mapping vector set.

[0235] Applying an irreversible function to the second-order sub-vector set generates an unordered factor set.

[0236] The generated unordered factor set is distributed to multiple geographically distributed edge nodes (simulating a distributed storage environment).

[0237] Option 2 (AES-256 encryption):

[0238] A 1GB file was encrypted using the AES-256 encryption algorithm, and the encrypted data was stored.

[0239] Step 2: Attack Simulation

[0240] Resistance to brute-force attacks:

[0241] For Option 1, attempting to brute-force the disturbed data will not result in the original data being recovered.

[0242] For scheme 2, a brute-force attack is simulated to attempt to decrypt the ciphertext data. Theoretically, this would require 2^256 attempts.

[0243] Anti-AI pattern recognition attack:

[0244] Option 1: Use a machine learning model (e.g., a deep neural network) to attempt to identify and predict patterns in the perturbed data. However, due to the use of an irregular Lévy flight distribution, the model cannot identify any patterns.

[0245] Option 2: Use machine learning models for pattern recognition. Due to its statistical properties, AES encrypted data can be cracked through pattern recognition.

[0246] Resistant to quantum computing attacks:

[0247] Option 1: Simulate a quantum computing attack to test the quantum computer's ability to break the perturbation model based on mathematical irreversibility (non-bijective perturbation function system). According to current research on quantum computing attacks, attacks relying on quantum computing cannot break the perturbation model of this invention.

[0248] Option 2: Use Shor's algorithm to simulate a quantum computing attack, and AES-256 encryption will be cracked.

[0249] Step 3: Recovery and Decryption

[0250] Solution 1 (Method of the present invention):

[0251] Optimized reorganization is performed based on an unordered factor set using valid recovery credentials.

[0252] By using an optimization model with gradient regularization, perturbed data can be reconstructed with limited accuracy to recover results that are close to the original data.

[0253] Option 2 (AES-256 encryption):

[0254] Use the AES-256 decryption key to decrypt the encrypted data.

[0255] Experimental data and results analysis:

[0256]

[0257] Experimental conclusion:

[0258] Resistance to brute-force attacks: The method of this invention has an resistance to brute-force attacks of "∞" because the perturbed data is irreversible, and attackers cannot recover the original data. In contrast, AES-256 requires 2^256 attempts, which is extremely time-consuming and relies heavily on key management.

[0259] Anti-AI pattern recognition attack: The method of this invention uses an irregular Lévy flight distribution, which is not easily recognized by machine learning models and avoids statistical cracking problems; while AES-256, due to its statistical characteristics, may be cracked when attackers use AI for pattern recognition.

[0260] Quantum computing resistance: The method of this invention is based on irreversible perturbation and quantum computing resistance design, which cannot be cracked by quantum computers; while AES-256 is vulnerable to Shor's algorithm in a quantum computing environment.

[0261] Encryption / Recovery Speed: The encryption and recovery speed of the method of this invention is superior to AES-256. Encrypting a 1GB file takes only 0.5 seconds, while AES-256 takes 2 seconds, representing a 4-fold improvement.

[0262] The experiment demonstrates that the perturbation storage and restricted recovery method proposed in this invention has significant advantages in terms of anti-attack capability, encryption speed, and resistance to future quantum computing environments.

[0263] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for securely processing self-organized bidirectional computational data of disordered factors, characterized in that, Includes the following steps: S1. Data vectorization: Converting raw data into a decimal integer vector; S2. Primary Lévy Partition: The decimal integer vector is randomly partitioned based on the Lévy flight distribution to generate a first-level sub-vector set; S3, Hyperchaotic Transformation: Nonlinear mapping of the first-level sub-vector set is performed on a three-dimensional tensor coupled chaotic system to output a chaotic mapped vector set; S4. Secondary Lévy partitioning: Perform a second Lévy distribution partitioning on the chaotic mapping vector set to generate a second-level sub-vector set; S5. Non-bijective decomposition: Apply an irreversible function transformation to the second-level sub-vector set to generate an unordered factor set that satisfies the entropy threshold; The irreversible function transformation in step S5 includes: Perform a circular left shift on each subvector Displacement Taken from a chaotic sequence; Applying a random mask matrix Perform bitwise XOR ; Combination operations are ; The S5 step also includes entropy feedback control: The formula for calculating the information entropy of the disordered factor is as follows: ； in, The entropy value. For data values, For probability distribution, It is a logarithmic base; when When the bit is reached, return to S4 for further splitting; S6. Distributed storage: Distribute the unordered factor set to geographically distributed edge nodes; S7. Restricted Recovery: When the recovery credential is verified to be valid, an optimized reorganization is performed based on the unordered factor set.

2. The data security processing method for self-organized bidirectional computation of disordered factors according to claim 1, characterized in that, The Levi flight distribution in step S2 uses parameters. For a stable distribution with a step size of 1.5, the formula for generating the step size is: ； in, and Let them be independent and identically distributed standard normal random variables. Let Lévy's distribution be the shape parameter. For Levi's flight stride; The random segmentation points in step S2 are determined by modulo operation. Dynamically determined, among which, For the first The second-generated Levi flight stride The dimension of the original decimal vector. For the first A dynamic segmentation point.

3. The data security processing method for self-organized bidirectional computation of disordered factors according to claim 1, characterized in that, The dynamic equations of the three-dimensional tensor-coupled chaotic system in step S3 are as follows: ； in, >3.7, The dynamic weight matrix is ​​used to iteratively generate chaotic sequences using the fourth-order Runge-Kutta method. , , For state variables, For control parameters, Here is the dissipation coefficient. For critical parameters, is the damping coefficient.

4. The data security processing method for self-organized bidirectional computation of disordered factors according to claim 1, characterized in that, The optimized recombination in step S7 is achieved by solving the following equation: ； in, To optimize variables, It is a non-bijective function. Given parameters, For scalar parameters, For gradient operators, It is the L2 norm. It is an L1 norm. The regularization coefficient is used. When the recovery error rate The calculation terminates at a certain time, where: ； in, For scalar parameters, To recover the data, This is the original data.

5. A data security processing system for self-organizing bidirectional computation of disordered factors, characterized in that: The data security processing method for unordered factor self-organizing bidirectional computation as described in any one of claims 1-4 includes: The data preprocessing module is used to convert raw data into a standard digital vector format; The chaos calculation module is used to perform Lévy flight distribution segmentation and hyperchaotic transformation processing; The non-bijective decomposition module is used to perform irreversible function transformations on data fragments; Edge storage networks are used to manage the distributed storage and location obfuscation of fragmented data. A data recovery engine for performing limited optimized reorganization based on storage fragmentation.

6. The data security processing system for self-organizing bidirectional computation of disordered factors according to claim 5, characterized in that, The chaos calculation module includes: The Levy parameter configuration unit is used to dynamically configure the Levy flight distribution. =1.5 parameter, controls the randomness of data segmentation; Tensor coupling operation unit, used for real-time calculation of tensor coupling terms in three-dimensional chaotic system; A chaos iteration accelerator for hardware acceleration of the fourth-order Runge-Kutta method for solving chaotic differential equations.

7. The data security processing system for self-organizing bidirectional computation of disordered factors according to claim 5, characterized in that, The edge storage network includes: Fragment tag encryption unit, used to generate unpredictable UUID tags based on quantum random numbers; The correlation analyzer is used to detect and block the storage of logically related factors on the same physical node; The dynamic migration engine is used to periodically trigger random migrations of fragment physical locations.

8. The data security processing system for self-organizing bidirectional computation of disordered factors according to claim 5, characterized in that, The data recovery engine includes: The gradient calculation unit is used for hardware-accelerated gradient norm calculation, enabling rapid evaluation of the regularization term; An error monitor is used to monitor the recovery error rate in real time, and the calculation is forcibly terminated when the ERR is greater than or equal to 37%. The projection optimizer is used to perform steepest descent iterations within a constrained solution space to find local optima.

Images