A multi-party secure computing method for data asset sharing

Abstract

This invention relates to the field of computer data security protection and multi-party secure computation technology, and discloses a multi-party secure computation method for data asset sharing, comprising: collecting subjective and objective evaluation data of participating nodes to calculate a comprehensive trust utility value; calculating a mask correction factor based on large prime numbers and the comprehensive trust utility value; constructing a secret sharing polynomial, and generating an adaptive secret sharing share by modulo-adding the mask correction factor and the polynomial solution values; each participating node calculating the intermediate ciphertext result, stripping the mask correction factor when secret reconstruction is triggered, and recovering the original private data asset through Lagrange interpolation. This invention transforms the node state into a finite-field mask correction factor, realizing adaptive blinding of the secret sharing share, protecting node privacy and resisting interference from abnormal nodes, while maintaining the algebraic closed loop of the reconstruction process, and improving the stability and efficiency of multi-party secure computation.

Description

Technical Field

[0001] This invention belongs to the field of computer data security protection and multi-party secure computation technology, and in particular relates to a multi-party secure computation method for data asset sharing. Background Technology

[0002] In the process of data asset sharing and collaborative computing, secure multi-party computation is an important technology for protecting private data. This technology usually uses cryptographic methods such as secret sharing, obfuscated circuits, and homomorphic encryption to enable all participants to jointly complete the computation of common functions without disclosing the original private data assets. This achieves the goal of making the data available but the original content invisible, and provides security protection for the data throughout the entire collaborative computing process.

[0003] As computing networks become increasingly heterogeneous and dynamic, different participating nodes exhibit significant differences in computing power, network latency, and historical success rates, leading to short-term fluctuations in their operational status. Traditional secret-sharing protocols typically assume a fixed set of participating nodes and relatively uniform trust levels among them. Masking and threshold reconstruction are executed according to predetermined algebraic rules and do not change with the real-time operational status of the nodes. When some nodes experience a degradation in their computing status due to network jitter, insufficient computing power, or the entry of non-real business data, the fixed threshold construction rules fail to reflect the current risk status of each node in a timely manner, potentially reducing the effectiveness of multi-party collaborative computing protocols. Inefficiency can be a serious problem, and in severe cases, it can even lead to failure in encrypted reconstruction. To strengthen the protection of abnormal nodes, existing technologies usually introduce node evaluation mechanisms or directly raise the secret reconstruction threshold. However, focusing only on the status monitoring of node hardware and operating environment is insufficient to meet the complex network collaboration requirements. At the level of control and evaluation methods, existing mechanisms also have obvious protection defects. For example, Chinese invention patent with authorization announcement number CN116633525B discloses a secure sharing method for industrial IoT data, which uses blockchain to record discrete binary evaluations to update node reputation values ​​and uses them as threshold conditions for access control.

[0004] Therefore, how to efficiently integrate multi-source heterogeneous state assessment data and enable the secret sharing share to be adaptively transformed and adjusted according to the real-time credibility of nodes has become the technical problem to be solved by this invention. Summary of the Invention

[0005] To address the problems in the background art, the technical solution of the present invention is as follows: A multi-party secure computation method for data asset sharing, comprising: Step S1: Obtain the data asset sharing task instruction, determine the N participating nodes in the data asset sharing task, and collect the subjective knowledge assessment data and objective operation log data of each participating node. The subjective knowledge assessment data includes true membership, false membership, uncertain membership and their credibility components, and the objective operation log data includes transmission delay, calculation success rate and protocol response delay fluctuation. Step S2: Based on the subjective knowledge evaluation data aggregated by the variable weight operator, the subjective reputation status of the data is obtained. Combined with the objective weights determined by the entropy weight method, the comprehensive weight of the attribute is calculated, and the comprehensive credibility utility value of each participating node is obtained. Step S3: Based on the comprehensive reliable utility value, calculate the mask correction factor corresponding to each participating node in the finite field composed of preset large prime numbers; Step S4: Allocate discrete evaluation points, construct a secret sharing polynomial with a reconstruction threshold of a preset threshold, and add the mask correction factor modulo superimposed on the solution value of the secret sharing polynomial at the discrete evaluation points to generate an adaptive secret sharing share. In step S5, each participating node calculates the intermediate ciphertext result based on the adaptive secret sharing share. When secret reconstruction is triggered, the number of participating nodes that have reached the preset threshold is selected. The corresponding mask correction factor is calculated and stripped according to the node identity. The calculation value of the secret sharing polynomial at the discrete evaluation point is restored. The original private data asset is restored by calculating through Lagrange interpolation.

[0006] Preferably, in step S2, the aggregation of subjective knowledge assessment data to obtain the subjective reputation status of the aggregated computing includes: step S21, using the Einstein variable weight extended power geometric operator to calculate the similarity and support between the intelligent and trustworthy numbers in each individual value, calculating the variable weight coefficient based on the support, wherein the variable weight coefficient corresponding to the subjective knowledge assessment data that deviates from the group central tendency is reduced; using the Einstein multiplicative conormative operator to calculate the intersection of true membership degrees, using the Einstein additive conormative operator to calculate the union of false membership degrees and uncertain membership degrees, simultaneously aggregating the trustworthy components, and outputting the aggregated computing subjective reputation status of each participating node.

[0007] Preferably, in step S2, calculating the comprehensive weight of attributes includes: step S22, calculating the subjective weight of each subjective evaluation indicator based on the single-valued intelligent credibility number using the stepwise priority evaluation method; calculating the objective weight of each objective operation indicator based on the objective operation log data using the entropy weight method; and using the Uninorm combination operator to calculate the ratio of the sum of the cross product and dual product of the subjective weight and the objective weight, and normalizing it to derive the comprehensive weight of attributes mapping each attribute dimension.

[0008] Preferably, in step S2, the calculation of the comprehensive credibility utility value of each participating node includes: step S23, using the set pair analysis method, converting the subjective reputation status and objective operation log data of the aggregate computing into a connection number space constructed by identity difference degree, uncertainty difference degree, and opposition difference degree; constructing a multi-attribute compromise ranking decision matrix, calculating the weighted Euclidean distance of each participating node from the relative positive ideal solution and the relative negative ideal solution in the connection number space, and deriving the comprehensive credibility utility value corresponding to each participating node.

[0009] Preferably, in step S4, constructing the secret sharing polynomial and generating the adaptive secret sharing share includes: step S41, constructing a basic secret sharing polynomial based on the original private data asset, a preset large prime number p-order finite field, random coefficients, and a reconstruction threshold as a preset threshold; adding the mask correction factor to the solution value of the basic secret sharing polynomial at the discrete evaluation point, and deriving the adaptive secret sharing share corresponding to the participating node under the modular calculation of the finite field.

[0010] Preferably, in step S5, stripping the mask correction factor and restoring the original private data assets includes: step S51, when a secret reconstruction request is triggered, selecting a set of participating nodes whose number reaches a preset threshold; each selected participating node submits its own adaptive secret sharing share; based on the identity of the participating node that submitted the share, calculating the mask correction factor of the participating node, deducting the mask correction factor from the adaptive secret sharing share, and restoring the solution value of the basic secret sharing polynomial at the discrete evaluation point under modular calculation in a finite field; constructing a Lagrange interpolation basis function based on the restored solution value, solving and restoring the original private data assets at the zero point position.

[0011] Preferably, when a participating node experiences a communication failure or its overall trust utility value degrades to a zero threshold, the mask correction factor corresponding to that participating node is automatically reset to zero, and the adaptive secret sharing share of that participating node degrades to the solution value of the basic secret sharing polynomial; the remaining participating nodes that meet the preset threshold complete the share stripping and interpolation calculation based on their respective mask correction factors, and the solution restores the original private data assets that maintain the algebraic closed loop.

[0012] Preferably, in step S5, the intermediate ciphertext result decryption by each participating node includes: step S52, each participating node uses its corresponding adaptive secret sharing share as input, calculates the homomorphic addition and homomorphic multiplication ciphertext overlap result based on the preset multiplication triplet; without decrypting the adaptive secret sharing share of each participating node, the intermediate ciphertext result representing the objective function of the data asset sharing task is interactively generated, and the status log of each participating node is updated.

[0013] Preferably, after generating the adaptive secret sharing share in step S4, the system further includes step S42, where each participating node performs homomorphic zero-knowledge proof verification on the received adaptive secret sharing share; when the verification passes, the computation response delay and computation success status are written into the objective operation log database for updating the objective operation log data in subsequent data asset sharing tasks.

[0014] Compared with existing technologies, the multi-party secure computation method for data asset sharing proposed in this invention has the following advantages: 1. In the multi-party secure computation of data asset sharing, the true membership, false membership, uncertain membership and their credibility in the subjective knowledge assessment data are expressed by the single-valued intelligent credibility number. The aggregation weight is adjusted according to the similarity between the assessment data by the Einstein variable weight extended power geometric operator, so that the weight of abnormal assessment data that deviates from the group is reduced accordingly. This can reduce the interference of abnormal values ​​on the node credibility assessment results, avoid the traditional fixed weight method being too sensitive to local abnormal data, and maintain the continuity and accuracy of the subjective reputation state calculation of the aggregated computing when there is uncertain noise and local fluctuations in the assessment data.

[0015] 2. Subjective expert knowledge and objective operational logs are used together for node credibility assessment, and the subjective reputation status of the computing platform and objective log data are unified into the contact number space for comprehensive judgment. This allows status information from different sources and in different forms to participate in node credibility ranking. This approach can take into account both expert knowledge and the actual operation of nodes, and improve the accuracy of multi-source heterogeneous status information conversion and node ranking without disclosing the private operational details of nodes. This provides a clear and quantifiable credibility basis for the subsequent allocation of secret sharing shares.

[0016] 3. By determining the mask correction factor corresponding to each participating node through comprehensive trusted utility value, the real-time trusted state of the node can directly participate in the adjustment of the adaptive secret sharing share, avoiding the separation of node state assessment and secret sharing process. When the trusted state of the node changes, the corresponding secret sharing share can be adjusted accordingly. During threshold reconstruction, the corresponding mask correction factor is stripped according to the node identity identifier, thereby ensuring that Lagrange interpolation can accurately restore the original private data assets and maintain the algebraic closed loop of the secret sharing reconstruction process. Attached Figure Description

[0017] Figure 1 This is a flowchart of the adaptive secret sharing share generation and reconstruction process of the present invention; Figure 2 This is the state transition diagram of the node operation state evolution involved in this invention. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0019] A method for secure multi-party computation for data asset sharing includes: Step S1: Obtain the data asset sharing task instruction, determine the N participating nodes in the data asset sharing task, and collect the subjective knowledge assessment data and objective operation log data of each participating node. The subjective knowledge assessment data includes true membership, false membership, uncertain membership and their credibility components, and the objective operation log data includes transmission delay, calculation success rate and protocol response delay fluctuation. Step S2: Based on the subjective knowledge evaluation data aggregated by the variable weight operator, the subjective reputation status of the data is obtained. Combined with the objective weights determined by the entropy weight method, the comprehensive weight of the attribute is calculated, and the comprehensive credibility utility value of each participating node is obtained. Step S3: Based on the comprehensive reliable utility value, calculate the mask correction factor corresponding to each participating node in the finite field composed of preset large prime numbers; Step S4: Allocate discrete evaluation points, construct a secret sharing polynomial with a reconstruction threshold of a preset threshold, and add the mask correction factor modulo superimposed on the solution value of the secret sharing polynomial at the discrete evaluation points to generate an adaptive secret sharing share. In step S5, each participating node calculates the intermediate ciphertext result based on the adaptive secret sharing share. When secret reconstruction is triggered, the number of participating nodes that have reached the preset threshold is selected. The corresponding mask correction factor is calculated and stripped according to the node identity. The calculation value of the secret sharing polynomial at the discrete evaluation point is restored. The original private data asset is restored by calculating through Lagrange interpolation.

[0020] Preferably, in step S2, the aggregation of subjective knowledge assessment data to obtain the subjective reputation status of the aggregated computing includes: step S21, using the Einstein variable weight extended power geometric operator to calculate the similarity and support between the intelligent and trustworthy numbers in each individual value, calculating the variable weight coefficient based on the support, wherein the variable weight coefficient corresponding to the subjective knowledge assessment data that deviates from the group central tendency is reduced; using the Einstein multiplicative conormative operator to calculate the intersection of true membership degrees, using the Einstein additive conormative operator to calculate the union of false membership degrees and uncertain membership degrees, simultaneously aggregating the trustworthy components, and outputting the aggregated computing subjective reputation status of each participating node.

[0021] Preferably, in step S2, calculating the comprehensive weight of attributes includes: step S22, calculating the subjective weight of each subjective evaluation indicator based on the single-valued intelligent credibility number using the stepwise priority evaluation method; calculating the objective weight of each objective operation indicator based on the objective operation log data using the entropy weight method; and using the Uninorm combination operator to calculate the ratio of the sum of the cross product and dual product of the subjective weight and the objective weight, and normalizing it to derive the comprehensive weight of attributes mapping each attribute dimension.

[0022] Preferably, in step S2, the calculation of the comprehensive credibility utility value of each participating node includes: step S23, using the set pair analysis method, converting the subjective reputation status and objective operation log data of the aggregate computing into a connection number space constructed by identity difference degree, uncertainty difference degree, and opposition difference degree; constructing a multi-attribute compromise ranking decision matrix, calculating the weighted Euclidean distance of each participating node from the relative positive ideal solution and the relative negative ideal solution in the connection number space, and deriving the comprehensive credibility utility value corresponding to each participating node.

[0023] Preferably, in step S4, constructing the secret sharing polynomial and generating the adaptive secret sharing share includes: step S41, constructing a basic secret sharing polynomial based on the original private data asset, a preset large prime number p-order finite field, random coefficients, and a reconstruction threshold as a preset threshold; adding the mask correction factor to the solution value of the basic secret sharing polynomial at the discrete evaluation point, and deriving the adaptive secret sharing share corresponding to the participating node under the modular calculation of the finite field.

[0024] Preferably, in step S5, stripping the mask correction factor and restoring the original private data assets includes: step S51, when a secret reconstruction request is triggered, selecting a set of participating nodes whose number reaches a preset threshold; each selected participating node submits its own adaptive secret sharing share; based on the identity of the participating node that submitted the share, calculating the mask correction factor of the participating node, deducting the mask correction factor from the adaptive secret sharing share, and restoring the solution value of the basic secret sharing polynomial at the discrete evaluation point under modular calculation in a finite field; constructing a Lagrange interpolation basis function based on the restored solution value, solving and restoring the original private data assets at the zero point position.

[0025] Preferably, when a participating node experiences a communication failure or its overall trust utility value degrades to a zero threshold, the mask correction factor corresponding to that participating node is automatically reset to zero, and the adaptive secret sharing share of that participating node degrades to the solution value of the basic secret sharing polynomial; the remaining participating nodes that meet the preset threshold complete the share stripping and interpolation calculation based on their respective mask correction factors, and the solution restores the original private data assets that maintain the algebraic closed loop.

[0026] Preferably, in step S5, the intermediate ciphertext result decryption by each participating node includes: step S52, each participating node uses its corresponding adaptive secret sharing share as input, calculates the homomorphic addition and homomorphic multiplication ciphertext overlap result based on the preset multiplication triplet; without decrypting the adaptive secret sharing share of each participating node, the intermediate ciphertext result representing the objective function of the data asset sharing task is interactively generated, and the status log of each participating node is updated.

[0027] Preferably, after generating the adaptive secret sharing share in step S4, the system further includes step S42, where each participating node performs homomorphic zero-knowledge proof verification on the received adaptive secret sharing share; when the verification passes, the computation response delay and computation success status are written into the objective operation log database for updating the objective operation log data in subsequent data asset sharing tasks.

[0028] Example 1: In a cross-institutional financial and medical data asset sharing computing network, the computing network receives a data asset sharing computing task instruction, determines the set of nodes participating in the computing as P1, P2, and P3, the total number of participating nodes is N=3, the private value of the data asset to be shared is S=500, and a finite field is used. Large prime number p=997, secret reconstruction threshold Due to network jitter and heterogeneous operating environments, participating nodes experience transient fluctuations in data quality and response latency. The system collects subjective knowledge assessment data and objective operation log data from each participating node. The subjective assessment indicators include data quality attribute C1 and security attribute C2, represented by a single-valued trusted intelligence number. Where node P1 is evaluated as a value under attribute C1. Under attribute C2, Node P2 is evaluated as follows under attribute C1: Under attribute C2, Node P3 has an evaluation value of C1. Under attribute C2, The protocol response latency fluctuations in the objective operation log data are characterized by real-time protocol response latency records: 12ms for node P1, 25ms for node P2, and 45ms for node P3.

[0029] To suppress abnormal numerical interference caused by local bias in expert subjective evaluation, the system employs the Einstein variable weight extended power geometric operator. The subjective knowledge assessment data of each node are weighted and aggregated, and a preset basic weight is set. , The variable weight control parameter α=1; where, the similarity between the two individual values ​​of the confidence number is calculated by weighting the absolute differences of their true membership, false membership, uncertain membership, and their respective corresponding confidence components, and the difference between the value 1 and the average of the absolute differences of each component is taken as the similarity measure between the two; for a specific evaluation data, the support obtained is the sum of the similarities between this data and all other evaluation data, and the variable weight coefficient is calculated by normalizing the proportion of the product of the support and the preset basic weight in the weighted sum of all support. Taking node P1 as an example, the individual value of the confidence number n 11 With n 12 The similarity between them is denoted as Calculated Export support Solving for the variable weight coefficients , Using Einstein's multiplication conormator With Einstein's addition covariant operator The calculated true membership degree of the set-based subjective reputation of node P1 is obtained. False membership degree Uncertain membership degree The three confidence components are simultaneously weighted and summed using the same set of variable weight coefficients v1 and v2 to obtain the aggregate confidence value. Calculations show that node P1's subjective reputation status is: Node P2 is Node P3 is The variable weights are dynamically adjusted based on support levels to suppress subjective knowledge assessment data that deviates from the norm.

[0030] For the weight allocation of subjective and objective multi-source state information, the stepwise priority evaluation method is used to solve the subjective weight vector. The objective weight vector is calculated based on objective operation log data using the entropy weight method. By using the Uninorm combination operator After fusion and normalization, the attribute comprehensive weight vector is obtained. When solving for subjective weights using the stepwise priority evaluation method, the individual value of the confidence index corresponding to each subjective evaluation indicator is used as the basis for priority evaluation. Decision experts determine the most important optimal attribute and the least important worst attribute from the set of evaluation attributes. Then, using integer scales from 1 to 9, they construct preference vectors for the optimal attribute to other attributes, and preference vectors for other attributes to the worst attribute. The system then establishes an optimization mathematical model with the objective of minimizing the maximum value of the absolute value of the deviation between the weight ratio of each attribute and its corresponding scale. Under the constraint that the sum of the weights is 1 and all terms are non-negative, the subjective weight vector is solved. Subsequently, based on set pair analysis, the subjective reputation status is transformed into a connection number space, and the connection number is denoted as... Where 'a' is the product of true membership and confidence, and 'c' is the product of false membership and confidence. Solve for the number of connections at node P1 Number of connections at node P2 Number of connections at node P3 For objective operation log data, normalization is performed according to the value direction of each objective operation indicator, and the data is incorporated into the connection number space constructed by identity difference, uncertainty difference, and opposition difference according to the corresponding attribute dimensions. During the construction of the decision matrix, the system expands the real and imaginary parts of the connection numbers of each node under each attribute into an extended evaluation matrix. The maximum value of each attribute column is extracted as the positive ideal solution benchmark, and the minimum value as the negative ideal solution benchmark. Then, the extended evaluation matrix is ​​column-weighted using attribute comprehensive weights, and the weighted Euclidean distance between the weighted connection number vector of each node and the positive and negative ideal solution vectors is calculated. Furthermore, by calculating the ratio of the relative distance of each node to the negative ideal solution to the sum of the distances to the positive and negative ideal solutions, the normalized comprehensive credibility utility value is obtained. A multi-attribute compromise ranking decision matrix, i.e., the MARCOS decision matrix, is constructed within the connection number space in conjunction with the attribute comprehensive weights. The comprehensive credibility utility values ​​of participating nodes P1, P2, and P3 are calculated as follows: , , ,satisfy The order relationship is used to complete the node trust classification without disclosing the node's private operation log data.

[0031] During the secret share generation phase, the system directly maps the comprehensive trust utility value to a mask correction factor in a finite field. Sum of utility values According to the formula The mask correction factor of node P1 is calculated. Masking correction factor for node P2 Mask correction factor for node P3 Constructing a fundamental secret shared polynomial Select random coefficients Discrete evaluation points x1=1, x2=2, and x3=3 are assigned, and the values ​​of each discrete evaluation point are used as the position identifiers of the participating nodes. The numerical values ​​of the solution of the basic secret sharing polynomial at each discrete evaluation point are calculated. , , The mask correction factor is modulo-added to the corresponding solution value to derive the adaptive secret sharing share of node P1. Adaptive secret sharing share of node P2 Adaptive secret sharing share of node P3 This approach achieves adaptive blinding of node trusted states to finite-field polynomial masks. In this process, the mask correction factor serves as a dynamic bias carrier of node trusted states in the transmission and collaborative computing links at both the physical and logical levels. Its core protection mechanism lies in the fact that the random coefficients of the basic secret sharing polynomial guarantee the absolute security of private data in the sense of information theory, while the mask correction factor in the finite field serves as an explicit mapping of node operating state and reputation level at the algebraic level. When there is non-collusive passive eavesdropping or single-point state deviation in the network, this dynamic bias decouples the final computing share held by each node from the basic secret sharing polynomial. This allows any node participating in collaborative computing to strip off the bias through the system's centralized authentication module or local collaboration, provided that it passes state assessment and identity verification. This establishes a lightweight defense barrier on the communication link that relies on dynamic authentication of node states.

[0032] When data asset sharing computing triggers a secret reconstruction request, a threshold value that meets the requirements is selected. Given sets of nodes P1 and P2, each node submits its own adaptive secret sharing share. and The system determines the comprehensive credible utility value and the sum of utility values ​​corresponding to the node during share generation based on the node's identity identifier. It then calculates the mask correction factor corresponding to the node according to the aforementioned mask correction factor calculation formula, subtracts this mask correction factor from the adaptive secret sharing share, and restores the calculated value of the basic secret sharing polynomial at the corresponding discrete evaluation point under finite field modular arithmetic. , The reconstructed value is obtained at x=0 using the Lagrange interpolation basis functions, and the interpolation coefficients are calculated. , Substitute into the solution to reconstruct private values The reconstruction result is completely consistent with the original private value; if in the extreme boundary condition, when node P3 loses communication, the system treats the comprehensive trust utility value of the participating node as a zero threshold; when the comprehensive trust utility value of node P3 itself degenerates to a zero threshold, the same result is obtained. After substituting this value into the mask correction factor calculation formula, its mask correction factor automatically returns to zero. The adaptive secret sharing share degradation is the numerical solution of the basic secret sharing polynomial at the corresponding discrete evaluation point, i.e. At this point, nodes P1 and P3 are selected for collaborative reconstruction. After calculating and stripping the corresponding mask correction factor according to the node identity, the result is restored. , Calculate the interpolation coefficients , Substitute into the solution to reconstruct private values The entire computation process maintains an algebraic closed loop, ensuring the accuracy of algebraic solutions and adaptive protection capabilities for multi-party secure computation under dynamic node state evolution. For abnormal nodes that are disconnected or whose reputation degrades to the zero threshold, the mechanism of setting their mask correction factor to zero is to strip the node's trusted weighted state at the system scheduling level and mark the node as unschedulable. When threshold reconstruction is performed among normal node sets, the reconstruction scheduling strategy automatically filters out abnormal nodes in the zero-trust state and selects only the share set that reaches the threshold from healthy nodes with a comprehensive trust utility value greater than zero for mask stripping and interpolation calculation. If the historical storage share of a node that has been zeroed is called, the original solution value of the basic secret shared polynomial is directly used to participate in the reorganization, thereby maintaining the consistency of the system scheduling protocol and algebraic restoration rules during topology state switching.

[0033] The in-situ mapping of the single-valued trusted number evaluation state, the set pair analysis connection number compromise ordering, and the finite field polynomial mask correction factor enables the lossless transformation of multi-source heterogeneous uncertain states into dynamic perturbation blinding values ​​of finite field secret shares. While maintaining the closed loop of Lagrange interpolation algebra reconstruction, it resolves the structural contradiction between dynamic trusted state control and cryptographic secret reconstruction.

[0034] Example 2: This example uses a distributed computing node cluster with an 8-core CPU and 16GB of memory to conduct a multi-party secure computing test for data asset sharing. Data is transmitted between nodes via 1Gbps Ethernet. The test input is taken from a cross-institutional medical and financial collaborative computing log dataset. Gaussian white noise with a signal-to-noise ratio of 20dB and an outlier disturbance with an amplitude of 15% are superimposed at the input to form the reputation assessment deviation corresponding to network jitter and non-real business data entry. During the calibration of the sampling period and the variable weight control parameter α, when the standard deviation of the latency fluctuation in the monitored log is greater than 15ms, the variable weight control parameter α is set to 1 and the sampling period is set to 50ms, so that the status data of the participating nodes are entered into the subsequent evaluation calculation according to this period.

[0035] The experiment included the experimental group of this invention and three control groups. Control group one used the classic Shamir secret sharing protocol without a dynamic mask correction mechanism; control group two used a weighted arithmetic mean operator instead of the Einstein variable-weight extended power geometric operator and did not use the Uninorm associative operator; control group three used the Uninorm associative operator but did not perform the connection number space transformation and multi-attribute compromise sorting decision matrix calculation. During the test, the outlier evaluation interference rate was set to increase in increments of 0%, 15%, 30%, and 45%, and the adjustment range of network transmission latency fluctuation was increased from 10ms to 150ms. The actual recorded latency fluctuations of the experimental group of this invention under each working condition were 12ms, 30ms, 45ms, and the upper limit of the adjustment range, respectively. The system collected subjective knowledge evaluation data from participating nodes P1, P2, and P3. In conjunction with objective operational log data, each participating node shares its own adaptive secret data. As input, the overlapping results of homomorphic addition and homomorphic multiplication ciphertexts are calculated based on preset multiplication triples. Homomorphic addition performs modular addition on the adaptive secret sharing shares within a finite field, while homomorphic multiplication completes the share operation through multiplication triples. The fixed offset introduced by the mask correction factor is congruently aligned in the corresponding linear reconstruction. The resulting result serves as the intermediate ciphertext result corresponding to the data asset sharing task. After triggering the secret reconstruction request, the corresponding mask correction factor is determined and stripped based on the identity of the participating nodes. Then, Lagrange interpolation is used to reconstruct the private values. The experiment used reconstruction success rate, absolute computational deviation, and protocol convergence time as evaluation indicators. The absolute computational deviation was obtained by repeated reconstruction tests for 100 consecutive times, and the arithmetic absolute difference between the reconstruction results and the original private values ​​in the modulus mapping interval was averaged.

[0036] Under the lowest interference rate condition of the sample group of this invention, the delay fluctuation is 12ms, and the comprehensive reliability utility value of participating node P1 is... The overall trust utility value of participating node P3 is 0.800. The value is 0.400, corresponding to a mask correction factor. The value is 221. After calculating the intermediate ciphertext result and performing secret reconstruction using the above adaptive secret sharing share, the reconstruction success rate is 100.0%, the absolute calculation deviation is 0, the protocol convergence time is 18.2ms, and after increasing the interference rate to the next level, the delay fluctuation is 30ms. It is 0.802. It is 0.385. With a value of 213, after the same share calculation, mask stripping, and Lagrange interpolation process, the reconstruction success rate was 100.0%, the absolute calculation deviation was 0, and the protocol convergence time was 19.5ms. When the interference rate was further increased, the delay fluctuation reached 45ms. It is 0.800. It is 0.400. The value is 221, the reconstruction power is still 100.0%, the absolute calculation deviation is 0, and the protocol convergence time is 21.4ms.

[0037] Under the highest interference rate condition, the delay fluctuation of node P3 reaches the upper limit of the aforementioned adjustment range. The Einstein variable-weighted extended power geometric operator, based on similarity... The support received Decrease, corresponding variable weight coefficient Consequently, the number of contacts was lowered; After entering the multi-attribute compromise ranking decision matrix, the comprehensive credible utility value of node P1 is... The overall reliability utility value of node P3 is 0.835. A value of 0.000 corresponds to a mask correction factor. When the value is 0, the adaptive secret sharing share of node P3 degenerates into the solution value of the basic secret sharing polynomial at the corresponding discrete evaluation point, i.e. After mask stripping and Lagrange interpolation, the reconstruction is obtained. Consistent with the original private values, under this condition, the reconstruction success rate is 100.0%, the absolute calculation deviation is 0, and the protocol convergence time is 24.8ms.

[0038] When conducting a control test under the same interference conditions, control group one used the classic Shamir secret sharing protocol. and Both are 0.333. The result was 0, the reconstruction success rate was 62.4%, the absolute calculation bias was 142, and the protocol convergence time was 58.6ms. Control group two used a weighted arithmetic mean operator to process the subjective knowledge evaluation data and did not use the Uninorm combination operator. It is 0.650. It is 0.520. The value was 288, the reconstruction success rate was 78.5%, the absolute computational bias was 45, the protocol convergence time was 38.1ms, and the control group used the three-retained Uninorm binding operator, but did not execute the uncertainty difference term in the connection number space. The transformation and subsequent complete compromise sorting yielded the following results. It is 0.710. It is 0.480. The value was 265, the reconstruction success rate was 89.2%, the absolute calculation deviation was 12, and the protocol convergence time was 29.3ms.

[0039] In the above test process, the subjective knowledge evaluation data carried by the single-valued trusted data is aggregated by the Einstein variable weight extension power geometric operator and participates in the calculation of attribute comprehensive weight and comprehensive trusted utility value together with the objective operation log data. The comprehensive trusted utility value is further used to determine the mask correction factor under the finite domain. As the node state and network latency change, the adaptive secret sharing share participates in the calculation of intermediate ciphertext results according to the corresponding mask correction factor, and completes mask stripping and Lagrange interpolation in the secret reconstruction stage. The reconstruction power, absolute calculation deviation and protocol convergence time of the sample group and each control group of this invention thus present the above experimental results.

[0040] Example 3: In this example, a computing network consisting of 3 participating nodes receives a data asset sharing computing instruction, determines the set of participating nodes as P1, P2, and P3, the total number of nodes N=3, the private value of the data asset to be shared is S=500, and the finite field is... Large prime number p=997, secret reconstruction threshold The discrete evaluation points are set to x1=1, x2=2, and x3=3 respectively. For the calibration of the variable weight control parameter α, the system collects the reliable data of each participating node under the data quality attribute C1 and the security attribute C2. Using Einstein's multiplication conormal operator With Einstein's addition covariant operator Calculate the similarity between nodes and support Then calculate the sample variance of the support sequence. When the support variance When, the variable weight control parameter is taken as α=1; when At that time, according to the formula Dynamic correction was applied, with support values ​​ranging from 0 to 1. After 500 outlier perturbation experiments, it was determined that when the evaluation group's opinion was consistent, the natural fluctuation of the support sequence sample variance was within the range of 0.01 to 0.04. Therefore, 0.05 was taken as the boundary between normal fluctuation and outlier perturbation. When the variance exceeded 0.05, an amplification factor of 10 caused the correction parameter to increase stepwise with increasing dispersion to the range of 2 to 3. In the experiment, the negative impact of outlier data on the aggregation results was suppressed by 75%. Taking P1 as an example, its evaluation value under attribute C1 was... The evaluation value under attribute C2 is Calculations yielded , Strain weighting coefficients , After aggregation, the true membership degree of node P1 in the aggregated subjective reputation state is calculated. False membership degree Uncertain membership degree The credibility aggregation component is Similarly, the subjective reputation state of node P2 is: Node P3 is .

[0041] In the attribute weight calculation, the individual value of the confidence index corresponding to each subjective evaluation indicator is used as the input for stepwise priority evaluation to obtain the subjective weight vector. In the objective operation log data, the real-time protocol response latency of nodes P1, P2, and P3 are 12ms, 25ms, and 45ms, respectively. The objective weight vector is calculated using the entropy weight method. The system uses the Uninorm combination operator. By integrating subjective and objective weights, taking attribute C1 as an example, the following calculations were performed: After normalization, the attribute comprehensive weight vector is obtained. Subsequently, the subjective reputation status and objective operation log data of the centralized computing system are transformed into a connection number space composed of identity difference degree, uncertainty difference degree, and opposition difference degree; wherein, the connection number corresponding to the subjective reputation status of the centralized computing system is denoted as... Identity Difference Opposites Difference Uncertainty about the degree of difference After the objective operation log data is normalized according to the corresponding objective operation indicators, it is transformed into the same connection number space and mapped to the corresponding attribute dimensions to calculate the connection number of node P1. The number of connections of node P2 The number of connections of node P3 In distance calculation, the identity difference, uncertainty difference, and opposition difference of the connection number are treated as three components in a 3D Euclidean orthogonal space. For each node, the corresponding component is subtracted from the relative positive ideal solution and the relative negative ideal solution component by component, the squares are summed, and the square root is taken to obtain the weighted Euclidean distance in the real number domain. Based on this, a multi-attribute compromise ranking decision matrix is ​​constructed, and the comprehensive credibility utility values ​​of nodes P1, P2, and P3 are calculated as follows: , , .

[0042] In the adaptive secret sharing share generation process, the total sum of comprehensive trustworthy utility values ​​is: ,according to Calculate the mask correction factor for each participating node in the finite field to obtain , , Constructing a basic secret shared polynomial random coefficients The calculated numerical values ​​are obtained at the discrete evaluation points x1=1, x2=2, and x3=3. , , The mask correction factor is modulo-superimposed on the corresponding solution value in a finite field to obtain the adaptive secret sharing share of node P1. Adaptive secret sharing share of node P2 Adaptive secret sharing share of node P3 Each participating node performs homomorphic zero-knowledge proof verification on the received adaptive secret sharing share. After verification, it performs homomorphic addition and homomorphic multiplication calculations using its own adaptive secret sharing share as input. During the offline preprocessing stage, random multiplication triplet secret sharing shares that satisfy the finite field multiplication relationship are pre-distributed to the participating nodes. During the homomorphic addition process, each participating node performs finite field modular addition on the input share locally. During the homomorphic multiplication process, each node subtracts the first and second shares of the pre-set multiplication triplet from the two input shares to be multiplied, broadcasts the corresponding difference share, and collaboratively reconstructs the plaintext difference constant. Then, the plaintext difference is linearly combined with the pre-set multiplication triplet share locally to obtain the multiplication output share. The fixed offset formed by the mask correction factor enters the difference calculation along with the input share and participates in the linear recombination of the multiplication triplet along with the plaintext difference. It maintains congruence correspondence under finite field modular calculation, thereby obtaining the ciphertext overlap result of homomorphic addition and homomorphic multiplication, which serves as the intermediate ciphertext result of the data asset sharing task.

[0043] When a secret reconfiguration is triggered in a data asset sharing computing network, a threshold for secret reconfiguration is selected. Nodes P1 and P2 respectively submit adaptive secret sharing shares. and The system determines the comprehensive trust utility value of the corresponding participating node based on the node's identity identifier, and then calculates the result by combining the total utility value of 1.800 and the large prime number 997 according to the mask correction factor calculation procedure. and After subtracting the mask correction factor from the corresponding adaptive secret sharing share, the numerical solution of the fundamental secret sharing polynomial at the discrete evaluation point is recovered under finite field modular arithmetic, resulting in... , Calculate the Lagrange interpolation basis function coefficients at the zero point x=0, and obtain , Based on this, the target private value is reconstructed. When the overall trust utility value of node P3 degrades to the zero threshold At that time, its mask correction factor is zeroed according to the same calculation procedure. The corresponding adaptive secret-sharing share degradation is the numerical value of the solution of the basic secret-sharing polynomial at the discrete evaluation point, i.e. When selecting nodes P1 and P3 for reconstruction, the mask correction factor is removed to obtain... , The Lagrange interpolation basis function coefficients are respectively , Finally obtained .

[0044] Example 4: This example combines Figures 1 to 2 This paper describes a multi-party secure computation method for data asset sharing, such as... Figure 1 As shown, in step S1, the data asset sharing task instruction is obtained, N participating nodes are determined, and subjective knowledge assessment data and objective operation log data of each participating node are collected. The subjective knowledge assessment data includes true membership, false membership, uncertain membership, and their credibility components. The objective operation log data includes transmission delay, computation success rate, and protocol response delay fluctuations. In step S2, the subjective knowledge assessment data is aggregated using the variable weight operator to obtain the subjective reputation status of the data processing system. Combined with the objective weight calculation attribute comprehensive weight determined based on the entropy weight method, the comprehensive credibility utility value of each participating node is calculated. In step S3, based on the comprehensive credibility utility value, the following calculations are performed: In step S4, the mask correction factor corresponding to each participating node in a finite field composed of a preset large prime number is calculated. Discrete evaluation points are allocated, and a secret sharing polynomial with a reconstruction threshold of a preset threshold is constructed. The mask correction factor is modulo-superimposed on the solution value of the secret sharing polynomial at the discrete evaluation point to generate an adaptive secret sharing share. In step S5, each participating node solves the intermediate ciphertext result based on the adaptive secret sharing share. When secret reconstruction is triggered, participating nodes with a number reaching the preset threshold are selected. The corresponding mask correction factor is calculated and stripped according to the node identity, and the solution value of the secret sharing polynomial at the discrete evaluation point is restored. The original private data asset is restored by calculating through Lagrange interpolation.

[0045] like Figure 2 As shown, the process enters the initialization state by obtaining task instructions and deploys the pre-calibration process. After collecting subjective and objective data, it switches to the state evaluation state to calculate the comprehensive reliable utility value. In the state evaluation state, it enters the share generation state after extracting factors to generate an adaptive secret shared share. After zero-knowledge proof verification, it enters the collaborative computing state to calculate the intermediate ciphertext result. When a reconstruction request is triggered, it enters the secret reconstruction state to restore the original private data assets and completes the calculation process by updating the running log. When the utility value degrades in the state evaluation state or when communication is lost in the collaborative computing state, it switches to the unschedulable state to automatically reset the mask correction factor to zero and filters abnormal nodes to enter the secret reconstruction state to restore the original private data assets.

[0046] Example 5: Before the data asset sharing computing task begins, the node initialization unit deploys a pre-calibration process for participating nodes P1, P2, and P3. The total number of participating nodes is N=3. The system collects log data corresponding to historical computing success rate, transmission delay, and protocol response latency fluctuations within 100 consecutive 50ms sampling periods to form initial objective operation log data and calculates the basic objective weight vector. When a new participating node is added to the computation network, or when the mean delay of an existing node deviates from the preset baseline by more than 20ms, a 10-second sliding time window variance check is initiated, the entropy weights are recalculated, and the attribute comprehensive weight vector under the Uninorm combined operator is updated. In a finite field with a large prime number p=997 and a reconstruction threshold, Dynamic reconstruction of the node's multidimensional state baseline is completed under the condition of remaining unchanged. The 50ms sampling period matches the average round-trip time of Ethernet packets. 100 consecutive periods are collected to form 5s baseline data, which is used to filter out transient impulse noise. When the node delay deviates by more than 20ms, the network congestion criterion is reached. A 10s sliding time window covers the communication recovery period to reduce repeated entropy weight calculations in the critical state. The false membership increase threshold of 0.300 is determined based on the confidence interval of expert evaluation. When the false membership jump exceeds 0.300, it is used as the judgment threshold for triggering the node's comprehensive credible utility value to return to zero.

[0047] During the collaborative computation process of the multi-party secret sharing protocol, the node status monitoring unit collects subjective knowledge assessment data and objective operation log data from each participating node in real time. The subjective knowledge assessment data uses a single-valued trusted data set. The system represents the transformation of subjective reputation status and objective operation log data into a connection space. When node P3 experiences a sudden increase in protocol response latency due to network jitter, and the increase in the pseudo-membership degree f is greater than 0.300, the Einstein variable weight extended power geometric operator is activated. The calculated support Monotonically decays to less than 0.100 for strain weighting coefficients The overall reliability utility value of node P3 approaches 0. Subsequently, the value drops to the zero threshold of 0.000. At this point, the total utility value still uses the fixed baseline calibration constant of 1.800 within the current computational task's lifetime, and the mask correction factor for node P3 is obtained according to the established calculation relationship: After the mask correction factor is reduced to zero, the adaptive secret sharing share of node P3... Degradation is the numerical solution of the basic secret shared polynomial at the corresponding discrete evaluation point. The remaining trusted nodes P1 and P2 determine their respective comprehensive trusted utility values ​​according to their respective identities, and calculate their respective mask correction factors in conjunction with the benchmark calibration constant. After mask stripping is completed, the original data asset private value S=500 is restored through Lagrange interpolation.

[0048] During the lifecycle of the current computation task, the sum of utility values ​​used by each participating node to generate adaptive secret sharing shares is fixed as the benchmark calibration constant. During initial distribution, a correspondence is established between the node's identity identifier and the comprehensive trusted utility value used by that node to calculate the mask correction factor and the benchmark calibration constant. After secret reconstruction is triggered, the system obtains the corresponding comprehensive trusted utility value based on the identity identifier of the participating node submitting the share, and calculates the corresponding mask correction factor in combination with the benchmark calibration constant. Then, it is stripped from the adaptive secret sharing shares. When it is detected that the comprehensive trusted utility value of a specific node degenerates to the zero threshold, only the mask correction factor of that node is set to zero, and its adaptive secret sharing share degenerates to the solution value of the basic secret sharing polynomial. The comprehensive trusted utility values ​​of other participating nodes and the calculation parameters of their corresponding mask correction factors remain unchanged.

[0049] After the secret reconstruction verification is passed, the system writes the calculation response latency and calculation success status to the objective operation log database for subsequent data asset sharing tasks to update the objective operation log data.

[0050] Example 6: After the distributed data asset sharing network is expanded, the system configures participating node sets P1, P2, and P3 according to the preset computing tasks. The total number of participating nodes N=3. The private value of the data asset to be shared is S=500, the finite field large prime number p=997, and the secret reconstruction threshold is... The discrete evaluation points are x1=1, x2=2, and x3=3. In the attribute weight calculation, the confidence number of the single value corresponding to each subjective evaluation index is used as the input for the step-by-step priority evaluation, and the subjective weight vector is obtained by solving. The objective weights are calculated solely based on the transmission delay, computation success rate, and protocol response latency fluctuations in the objective operation log data, resulting in an objective weight vector. The system inputs the two sets of weights into the Uninorm combination operator in the order of their corresponding attributes. Taking attribute C1 as an example, substitute the values ​​into the calculation. After normalization, the attribute comprehensive weight vector is obtained. Both the subjective weight vector and the objective weight vector are arranged in a predetermined order of the three indicators. The Uninorm combination operator performs corresponding fusion on the weights with the same order.

[0051] The system transforms the subjective reputation status and objective operation log data of each participating node into a connection number space composed of identity difference degree, uncertainty difference degree, and opposition difference degree, and the individual value of each subjective evaluation indicator is the intellectual credibility number. Forming a connection number according to the given mapping Objective operation log data is normalized according to the value direction of each objective operation indicator and converted into corresponding components in the same connection number space. The connection numbers formed by the two types of data and the comprehensive weight of attributes are input into the multi-attribute compromise ranking decision matrix. The weighted Euclidean distance of each participating node from the relative positive ideal solution and the relative negative ideal solution is calculated respectively, and the comprehensive credibility utility values ​​of nodes P1, P2, and P3 are obtained as follows: , , .

[0052] After the adaptive secret sharing share is distributed, each participating node performs homomorphic zero-knowledge proof verification on the received adaptive secret sharing share. During the initial distribution phase, the distributing node, based on the finite field elliptic curve discrete logarithmic hard-solvable group, publicly discloses the homomorphic cryptographic commitment values ​​corresponding to the coefficients of the basic secret sharing polynomial. Upon receiving the adaptive secret sharing share, each participating node determines the corresponding comprehensive trusted utility value based on its node identity, and calculates the corresponding mask correction factor by combining the sum of utility values ​​used in the current task and large prime numbers in the finite field. During the verification process, this mask correction factor is temporarily deducted from the adaptive secret sharing share to be verified, yielding the basic secret used for verification. The shared polynomial solution values ​​are used to obtain the benchmark commitment by performing exponential scalar multiplication on the publicly committed value based on their own discrete evaluation points and homomorphically accumulating the results. This generates a non-interactive zero-knowledge proof response with discrete logarithmic knowledge. The verification unit verifies the identity relationship between the group elements mapped to the response value and the benchmark commitment under finite field group multiplication. This determines whether the received adaptive secret sharing share corresponding to the basic secret shared polynomial solution value is located within the algebraic solution space limited by the original polynomial coefficients. If the verification passes, the calculation response delay and calculation success status are written to the objective operation log database for subsequent data asset sharing tasks to update the objective operation log data.

[0053] When data asset sharing computation triggers secret reconstruction, nodes P1 and P2 that have reached the secret reconstruction threshold are selected, and they respectively submit their verified adaptive secret sharing shares. and The system determines the corresponding comprehensive trust utility value based on the node identity identifier of the submitted share, and obtains it according to the established calculation relationship of the mask correction factor. and After deducting the corresponding mask correction factor from the adaptive secret sharing share, the solution value of the basic secret sharing polynomial at the discrete evaluation point is recovered under finite field modular arithmetic. The formula corresponding to node P1 is: The formula corresponding to node P2 is: To restore the obtained and Construct the Lagrange interpolation basis function and calculate the interpolation coefficients of node P1 at the zero point. The interpolation coefficients of node P2 Substituting the numerical solution of the fundamental secret shared polynomial and the interpolation coefficients into the Lagrange reconstruction polynomial, we obtain... This is consistent with the original private value of the data asset to be shared, S=500.

[0054] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.

Claims

1. A multi-party secure computation method for data asset sharing, characterized in that, include: Step S1: Obtain the data asset sharing task instruction, determine the N participating nodes in the data asset sharing task, and collect the subjective knowledge assessment data and objective operation log data of each participating node. The subjective knowledge assessment data includes true membership, false membership, uncertain membership and their credibility components, and the objective operation log data includes transmission delay, calculation success rate and protocol response delay fluctuation. Step S2: Based on the subjective knowledge evaluation data aggregated by the variable weight operator, the subjective reputation status of the data is obtained. Combined with the objective weights determined by the entropy weight method, the comprehensive weight of the attribute is calculated, and the comprehensive credibility utility value of each participating node is obtained. Step S3: Based on the comprehensive reliable utility value, calculate the mask correction factor corresponding to each participating node in the finite field composed of preset large prime numbers; Step S4: Allocate discrete evaluation points, construct a secret sharing polynomial with a reconstruction threshold of a preset threshold, and add the mask correction factor modulo superimposed on the solution value of the secret sharing polynomial at the discrete evaluation points to generate an adaptive secret sharing share. In step S5, each participating node calculates the intermediate ciphertext result based on the adaptive secret sharing share. When secret reconstruction is triggered, the number of participating nodes that have reached the preset threshold is selected. The corresponding mask correction factor is calculated and stripped according to the node identity. The calculation value of the secret sharing polynomial at the discrete evaluation point is restored. The original private data asset is restored by calculating through Lagrange interpolation.

2. The multi-party secure computation method for data asset sharing according to claim 1, characterized in that, In step S2, the aggregation of subjective knowledge assessment data to obtain the subjective reputation status of the aggregated computing includes: Step S21, using the Einstein variable weight extended power geometric operator, calculating the similarity and support between the intelligent and trustworthy numbers in each individual value, calculating the variable weight coefficient based on the support, wherein the variable weight coefficient corresponding to the subjective knowledge assessment data that deviates from the group central tendency is reduced; using the Einstein multiplicative conormative operator to calculate the intersection of true membership degrees, using the Einstein additive conormative operator to calculate the union of false membership degrees and uncertain membership degrees, simultaneously aggregating the trustworthy components, and outputting the aggregated computing subjective reputation status of each participating node.

3. The multi-party secure computation method for data asset sharing according to claim 1, characterized in that, In step S2, calculating the comprehensive weight of attributes includes: step S22, calculating the subjective weight of each subjective evaluation index based on the single-valued intelligence credibility number through a step-by-step priority evaluation method; The objective weights of each objective operation indicator are calculated using the entropy weight method based on objective operation log data. The Uninorm combination operator is used to calculate the ratio of the sum of the cross product and dual product of subjective and objective weights, and the normalization is used to derive the comprehensive weight of the attribute for each attribute dimension.

4. A multi-party secure computation method for data asset sharing according to claim 1, characterized in that, In step S2, the comprehensive credibility utility value of each participating node is calculated as follows: Step S23, using the set pair analysis method, the subjective reputation status and objective operation log data of the aggregate computing are transformed into a connection number space constructed by identity difference degree, uncertainty difference degree and opposition difference degree; a multi-attribute compromise ranking decision matrix is ​​constructed, the weighted Euclidean distance of each participating node to the relative positive ideal solution and the relative negative ideal solution in the connection number space is calculated, and the comprehensive credibility utility value corresponding to each participating node is derived.

5. A multi-party secure computation method for data asset sharing according to claim 1, characterized in that, In step S4, constructing the secret sharing polynomial and generating the adaptive secret sharing share includes: step S41, constructing the basic secret sharing polynomial based on the original private data assets, a preset large prime number p-order finite field, random coefficients, and a reconstruction threshold as a preset threshold; adding the mask correction factor to the solution value of the basic secret sharing polynomial at the discrete evaluation point, and deriving the adaptive secret sharing share corresponding to the participating node under the modular calculation of the finite field.

6. A multi-party secure computation method for data asset sharing according to claim 1, characterized in that, In step S5, stripping the mask correction factor and restoring the original private data assets includes: step S51, when a secret reconstruction request is triggered, selecting a set of participating nodes whose number reaches a preset threshold; each selected participating node submits its own adaptive secret sharing share; based on the identity of the participating node that submitted the share, calculating the mask correction factor of the participating node, deducting the mask correction factor from the adaptive secret sharing share, and restoring the solution value of the basic secret sharing polynomial at the discrete evaluation point under modular calculation in a finite field; constructing a Lagrange interpolation basis function based on the restored solution value, solving at the zero point position, and restoring the original private data assets.

7. A multi-party secure computation method for data asset sharing according to claim 6, characterized in that, When a participating node loses communication or its overall trust utility value degrades to zero, the mask correction factor corresponding to the participating node is automatically reset to zero, and the adaptive secret sharing share of the participating node degrades to the solution value of the basic secret sharing polynomial. The remaining participating nodes that meet the preset threshold complete the share stripping and interpolation calculation based on their respective mask correction factors, and the solution restores the original private data assets that maintain the algebraic closed loop.

8. A multi-party secure computation method for data asset sharing according to claim 1, characterized in that, In step S5, the intermediate ciphertext result decryption by each participating node includes: step S52, each participating node takes its corresponding adaptive secret sharing share as input, calculates the homomorphic addition and homomorphic multiplication ciphertext overlap result based on the preset multiplication triple; without decrypting the adaptive secret sharing share of each participating node, the intermediate ciphertext result representing the objective function of the data asset sharing task is interactively generated, and the status log of each participating node is updated.

9. A multi-party secure computation method for data asset sharing according to claim 1, characterized in that, After generating the adaptive secret sharing share in step S4, step S42 is also included, in which each participating node performs homomorphic zero-knowledge proof verification on the received adaptive secret sharing share; when the verification passes, the calculation response delay and calculation success status are written into the objective operation log database, which is used to update the objective operation log data in subsequent data asset sharing tasks.

Images