Data Quality Verification Methods and Systems in Secure Multi-Party Computation

Abstract

This invention relates to data security technology, and particularly to a data quality verification method and system for multi-party secure computation. The method first sorts the private key shares of data holders, generates polynomials and commitments through polynomial zero-knowledge verifiable distributed private key share distribution, and stores these commitments in an authoritative institution, simultaneously generating private key shares and polynomial zero-knowledge proofs. Next, verifiable random sampling is achieved using polynomial commitments, encoding the statistical characteristics of the dataset into the polynomial commitments. Furthermore, this method supports cross-protocol and multi-framework protocol implementations, enhancing the system's compatibility and flexibility. Finally, a verifiable data quality proof framework is constructed to verify whether the distribution characteristics of data holder shares conform to statistical characteristics. This invention innovatively combines polynomial zero-knowledge proofs and distributed private key share distribution, achieving efficient data quality verification while protecting data privacy, effectively resolving the contradiction between privacy protection and quality verification.

Description

Technical Field

[0001] This invention relates to data security technology, and in particular to a data quality verification method and system for multi-party secure computation, belonging to the fields of cryptography, distributed computing, data quality management and privacy protection. Background Technology

[0002] With the rapid development of big data and artificial intelligence technologies, multi-party collaborative computation and analysis has become an important technological application model. Secure multi-party computation allows multiple data holders to jointly complete specific computational tasks without disclosing the original data, which is of great significance for industries involving sensitive data, such as VIPs, telecommunications, finance, healthcare, and government.

[0003] However, data quality verification faces significant challenges in multi-party secure computation. First, traditional data quality verification methods typically require direct access to the raw data, which contradicts the privacy protection goals of multi-party secure computation. Second, because data is distributed across different participants, centralized quality verification methods are not feasible. Third, malicious participants may provide low-quality or inaccurate data, distorting the computation results, which other participants may find difficult to detect and verify.

[0004] Existing technologies often employ homomorphic encryption and trusted execution environments (TEUs) for data protection. However, these methods have limitations in data quality verification: on the one hand, while homomorphic encryption can protect data privacy, it has high computational complexity and is difficult to support complex data quality verification; on the other hand, TEUs require specific hardware support, lack universality, and still pose security risks.

[0005] Furthermore, existing data quality verification methods in multi-party secure computation mostly focus on integrity and consistency checks, lacking in-depth verification of data distribution characteristics and failing to effectively identify carefully constructed data that does not conform to the true distribution. At the same time, existing methods typically rely on specific computational frameworks, exhibiting poor cross-platform compatibility and difficulty in flexible deployment across various application environments.

[0006] Therefore, how to achieve efficient and reliable data quality verification while protecting data privacy and supporting cross-platform applications has become a pressing technical challenge in the field of multi-party secure computing. Summary of the Invention

[0007] The main objective of this invention is to provide a data quality verification method and system for multi-party secure computation, aiming to solve the technical problems in the prior art, such as the difficulty in balancing data quality verification and privacy protection, low verification efficiency, and poor cross-platform compatibility.

[0008] This invention proposes a data quality verification method in multi-party secure computation, including:

[0009] The method includes the following steps: S1, a distributed private key share distribution method based on polynomial zero-knowledge verifiable methods. For a given access policy and a set of data holders, the private key shares of the data holders are first sorted, then a polynomial is generated, the ciphertext commitment and coefficient commitment of the polynomial are calculated, and these commitments are outsourced and stored to an authoritative institution. Finally, the data holders generate a private key share proof using a secret multiplication sharing scheme and calculate the zero-knowledge proof of the polynomial; S2, a verifiable random sampling method based on polynomial commitments. The statistical characteristics of the dataset distribution are extracted and encoded, and the encoded values ​​are bound to the polynomial commitments; S3, support for cross-protocol and protocol implementations based on multiple frameworks; S4, a verifiable data quality proof framework method based on polynomial commitments. The data distribution characteristics of the data holders' shares are verified, and it is checked whether the distribution conforms to the statistical characteristics.

[0010] Preferably, the S1 method specifically includes: S11, generating a private key share for the data holder; S12, a verifiable representation of the distributed private key share distribution, generating a commitment value; S13, a verifiable proof of data quality; and S14, privacy protection.

[0011] Preferably, S11 includes: 1A1) selecting an access strategy, and for a given access strategy, distributing private key shares using a Shahal secret sharing scheme to obtain shares for all n data holders; 1B1) generating a key for a ciphertext matrix composed of all shares based on the obtained private key shares of the n holders and according to the specified access structure.

[0012] Preferably, S12 includes: 1A2) a set of data holders; 1B2) committing to outsource storage to an authoritative institution.

[0013] Preferably, S13 includes: 1A3) proof that the holder has obtained a share of the private key; 1B3) zero-knowledge proof for computing the ciphertext.

[0014] Preferably, S14 includes: 1A4) the verifier verifies the proof of the private key share, and issues an abnormal warning if the proof is invalid.

[0015] Preferably, the S2 method specifically includes: S21, encoding the distribution characteristics of the dataset; S22, representing the distribution characteristics of the verifiable dataset; and S23, a verifiable data quality verification method based on multinomial commitment.

[0016] Preferably, S21 includes: 2A1) Selecting features X of the dataset; for a set of N data samples, taking the value of each feature X in each sample as a feature vector x; dividing all feature vectors x into buckets to obtain a bucket set V; 2B1) Generating the feature mean within each bucket from the bucket set V; 2C1) For each bucket set Vi, calculating the number of data samples in each bucket set; and comparing it with the dataset N to obtain the proportion of data samples in each bucket set Vi to the whole, i.e., the bucket distribution ratio pv.

[0017] Preferably, step S22 includes: 2A2) For the generated bucket set V, first sort the bucket set to obtain the order l of each bucket set, encrypt the feature mean of the bucket according to the order l to obtain a hash value h, and use the encrypted bucket feature mean set and h as distributed ciphertext; 2B2) For the data set Di of the data holder, calculate the proportion distribution of the bucket set V of the statistical feature X to obtain the distribution feature set pi of each bucket set Vi; 2C2) For the encrypted bucket set Vi, calculate its private key share di using the secret sharing scheme, and calculate the verification share according to di and pi; 2D2) For the share of each data holder, calculate the ciphertext m formed by the aggregation of multiple shares, and its aggregation coefficient commitment value c, use c and m as the shared dataset ciphertext value, generate a polynomial based on polynomial commitment, and finally the holder sends its polynomial ciphertext commitment s and coefficient commitment ci.

[0018] A data quality verification system for multi-party secure computation includes a system server, a data holder, and a data trainer. The system server is connected to the data holder and the data trainer via network or wireless signals. The system server is used to upload and download private and public data, and provides data quality verification technology to the data holder. The data holder is used to generate a privacy protection share for private data, as well as to detect data quality and anomalies. The data trainer is used to train and verify data quality on public data. The system server includes a data holder dataset module, a data trainer dataset module, a system dataset module, a multi-security protocol module, and an algorithm model module. The data holder dataset module, the data trainer dataset module, and the system dataset module are signal-connected to the multi-security protocol module, and the multi-security protocol module is signal-connected to the algorithm model module.

[0019] The beneficial effects of this invention are mainly reflected in:

[0020] 1. This invention innovatively combines polynomial zero-knowledge proof with distributed private key share distribution, achieving efficient data quality verification while protecting data privacy, thus resolving the contradiction between privacy protection and quality verification;

[0021] 2. The verifiable random sampling method based on multinomial commitment proposed in this invention makes the verification of data distribution characteristics more accurate and efficient, and greatly improves the reliability of the verification.

[0022] 3. This invention supports cross-protocol and multi-framework implementation, significantly enhancing the system's versatility and adaptability, and facilitating application deployment in different computing environments;

[0023] 4. This invention constructs a complete data quality verification framework, integrating verification, anomaly detection and handling mechanisms, thereby improving the robustness and availability of the system. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating a data quality verification method in multi-party secure computation according to the present invention.

[0025] Figure 2 This is a schematic diagram of the distributed private key share distribution method based on polynomial zero-knowledge verifiable method of the present invention.

[0026] Figure 3 This is a schematic diagram of the verifiable random sampling method based on multinomial commitment of the present invention.

[0027] Figure 4 This is a schematic diagram illustrating the cross-protocol and multi-framework-based protocol implementation of the present invention;

[0028] Figure 5 This is a schematic diagram of the verifiable data quality proof framework method based on multinomial commitment of the present invention.

[0029] Figure 6 This is a schematic diagram of the data quality verification system in multi-party secure computation according to the present invention. Detailed Implementation

[0030] Please refer to the attached document. Figure 1-6 The technical solution of the present invention will be described in detail below with reference to specific embodiments.

[0031] Example 1: A Data Quality Verification Method in Secure Multi-Party Computation

[0032] Reference Figure 1 This invention proposes a data quality verification method in multi-party secure computation, comprising the following steps:

[0033] S1. A distributed private key share distribution method based on polynomial zero-knowledge verifiable method: For a given access policy and set of data holders, the private key shares of the data holders are first sorted, then a polynomial is generated, the ciphertext commitment and coefficient commitment of the polynomial are calculated, and these commitments are outsourced and stored to an authoritative institution. Finally, the data holders generate a private key share proof using a secret multiplication sharing scheme and calculate the zero-knowledge proof of the polynomial.

[0034] S2. A verifiable random sampling method based on multinomial commitments extracts the statistical characteristics of the dataset distribution, encodes the distribution, and binds the encoded values ​​to multinomial commitments.

[0035] S3 supports cross-protocol and protocol implementations based on multiple frameworks;

[0036] S4. A verifiable data quality proof framework based on multinomial commitments verifies the data distribution characteristics of data holder shares and checks whether the distribution conforms to statistical characteristics.

[0037] The following is a detailed explanation of each step.

[0038] S1: A distributed private key share distribution method based on polynomial zero-knowledge verifiable methods

[0039] Reference Figure 2 Step S1 specifically includes the following sub-steps:

[0040] S11, Generate the data holder's private key share;

[0041] S12. A verifiable representation of the distributed private key share distribution, generating a commitment value;

[0042] S13, Verifiable proof of data quality;

[0043] S14. Privacy Protection.

[0044] The specific implementation process of S11 includes:

[0045] 1A1) Select an access strategy. For a given access strategy, use the Shahal secret sharing scheme to distribute private key shares to obtain the shares of all n data holders.

[0046] In a preferred embodiment of the present invention, the Shahal secret-sharing scheme employs... The threshold scheme requires at least t shares to reconstruct the secret, where The specific process is as follows:

[0047] First, choose a large prime number. and polynomial of degree ,in It is a secret to be shared. These are randomly selected coefficients.

[0048] Then, for each data holder Calculate shares mod Thus, only when at least Only when data holders collaborate can the polynomial be reconstructed through Lagrange interpolation. And obtain the secret .

[0049] 1B1) Based on the obtained Each holder's private key share is used to generate a key for a ciphertext matrix consisting of all shares, based on the specified access structure.

[0050] In this embodiment, the access structure defines which subset of data holders have the authority to reconstruct the secret. Based on the obtained... Private key share Constructing a ciphertext matrix ,in , Indicates the encryption function. It is the first in the corresponding access structure The key for each authorized subset. This ensures that only members of the authorized subset can access the corresponding private key share.

[0051] The specific implementation process of S12 includes:

[0052] 1A2), Data Holder Set;

[0053] In this embodiment, the data holder set Indicates participation in secure computation Each entity holds a portion of the data and a corresponding share of the private key.

[0054] 1B2) Commit to outsourcing storage to an authoritative institution.

[0055] To ensure the fairness and reliability of the verification, this invention will generate a polynomial ciphertext commitment. Commitment of coefficients Outsource storage to authoritative institutions. and It is a public generator. These are random numbers. Authoritative institutions only store these commitments and do not access the original data, thus ensuring data privacy.

[0056] The specific implementation process of S13 includes:

[0057] 1A3) Proof that the holder has obtained the share of the private key;

[0058] For each data holder Generate its private key share proof In this invention, the Schnorr protocol is used to generate zero-knowledge proofs:

[0059] 1. Data holders Select random number ,calculate ;

[0060] 2. Computational Challenges ,in It is a safe hash function;

[0061] 3. Calculate the response ;

[0062] 4. Proof of Private Key Share ;

[0063] 1B3) Zero-knowledge proofs for computation of ciphertext.

[0064] Based on private key share and polynomial Zero-knowledge proofs for computational ciphertext Proof of data holder It is confirmed that the correct share of the private key is held, and that the share is polynomial. At point The value of . To improve efficiency, this embodiment adopts the Bulletproofs scheme, which has a smaller proof size and higher verification efficiency.

[0065] The specific zero-knowledge proof calculation process is as follows:

[0066] 1. Data holders Calculate commitment ;

[0067] 2. Construct the proof polynomial ;

[0068] 3. Generate the coefficient commitments of the polynomial and construct a zero-knowledge proof. ;

[0069] The specific implementation process of S14 includes:

[0070] 1A4) The verifier verifies the proof of the private key share, and issues an exception warning if the proof is invalid.

[0071] The validator receives the data holder Proof of private key share Then, perform the following verification steps:

[0072] 1. Computational Challenges ;

[0073] 2. Verification ;

[0074] If the verification is successful, the proof is accepted; otherwise, the proof is considered invalid and an anomaly warning is issued. In practice, to improve security, a verification threshold can be set, for example, issuing an anomaly warning only after three consecutive verification failures. This can reduce the false alarm rate caused by temporary factors such as network fluctuations.

[0075] S2: Verifiable random sampling method based on multinomial commitment

[0076] Reference Figure 3 Step S2 specifically includes the following sub-steps:

[0077] S21. Distribution characteristics encoding of the dataset;

[0078] S22, Verifiable representation of dataset distribution characteristics;

[0079] S23. Verifiable data quality verification method based on multinomial commitment.

[0080] The specific implementation process of S21 includes:

[0081] 2A1) Select the feature X of the dataset. For a set of N data samples, take the value of each feature X in each sample as the feature vector x. Divide all feature vectors x into buckets to obtain the bucket set V.

[0082] In a preferred embodiment of the invention, feature X can be a numerical feature (such as age, income) or a categorical feature (such as gender, occupation). The binning strategy varies depending on the feature type:

[0083] For numerical features, equal-width binning (dividing the numerical range into k equal intervals) or equal-frequency binning (making the number of samples in each bin approximately equal) can be used.

[0084] For categorical features, each category corresponds to a bucket.

[0085] For example, for age features, the binning intervals can be set to [0-18, 19-35, 36-50, 51-65, 66+], and the data samples can be assigned to the corresponding bins according to their age values. The number of bins, k, is usually determined based on the data distribution and validation accuracy requirements. Practice shows that a k value between 5 and 10 usually achieves a good balance.

[0086] 2B1) Generate the mean of features within each bucket by dividing the buckets into buckets and then using the bucket set V;

[0087] For each bucket Calculate its characteristic mean:

[0088] ,

[0089] in, Indicates bucket The number of samples in the middle This represents the feature value of the samples within the bucket. For categorical features, the mean can be calculated using the encoded numerical values, or the mode can be used directly instead of the mean.

[0090] 2C1) For each bucket set Vi, calculate the number of data samples in each bucket set, and then compare it with the dataset. By comparing the data samples in each bucket set Vi, we can obtain the proportion of data samples in the whole, i.e., the bucket distribution ratio pv.

[0091] The formula for calculating the bucket distribution ratio is:

[0092] ,

[0093] in, It's a bucket The number of samples in This refers to the total number of samples. These distribution proportions form the distributional fingerprint of the data, which is an important basis for verifying data quality.

[0094] The specific implementation process of S22 includes:

[0095] 2A2) For the generated bucket set V, first sort the bucket set to obtain the order l of each bucket set. Then, encrypt the feature mean of the bucket according to the order l to obtain a hash value h. The encrypted bucket feature mean set and h are used as distributed ciphertext.

[0096] Bucket set sorting can be based on bucket boundary values ​​(for numerical features) or alphabetical order of categories (for categorical features). After sorting, each bucket... Obtain a unique serial number Then, the mean of the features within the bucket. Encrypt:

[0097] ,

[0098] Where Enc is the encryption function. This is the key, and || represents the concatenation operation. Finally, calculate the hash value of all encrypted values:

[0099] ,

[0100] in It is a secure hash function, such as SHA-256. Distributed ciphertext is... composition.

[0101] 2B2) For the data set Di of the data holder, the proportion distribution of the bucket set V of the statistical feature X is used to obtain the distribution feature set pi of each bucket set Vi;

[0102] For each data holder Calculate the distribution ratio of its data in each bucket:

[0103] ,

[0104] in, , Data holder In the bucket The number of samples in yes The total number of samples.

[0105] 2C2) For an encrypted bucket set Vi, use the secret sharing scheme to calculate its private key share di, and calculate the verification share based on di and pi;

[0106] For each encrypted bucket set Using the aforementioned Shamir secret sharing scheme, a private key share is generated. Then, the verification share is calculated based on the private key share and distribution characteristics:

[0107] ,

[0108] in It is a hash function that maps to points on an elliptic curve. It is the base point on the elliptic curve. This indicates a join operation.

[0109] 2D2) For each data holder's share, calculate the ciphertext m formed by the aggregation of multiple shares, and its aggregation coefficient commitment value c. Use c and m as the shared dataset ciphertext value to generate a polynomial based on the polynomial commitment. Finally, the holder sends its polynomial ciphertext commitment s and coefficient commitment ci.

[0110] The aggregate calculation of multiple verification shares is as follows:

[0111] ,in It is the polymerization coefficient, full The aggregation coefficient commitment value is:

[0112] ,

[0113] Based on aggregated ciphertext Constructing a polynomial ,in These are random coefficients.

[0114] Finally, the encrypted commitment to compute the polynomial is calculated. Commitment of coefficients And send it to the verifier.

[0115] The implementation of S23 is quite complex, involving interactive verification between data holders. Here are its core steps:

[0116] 1. Holder A distributes data to holder B for training;

[0117] 2. Holder B generates multiple data features and sends them to holder A;

[0118] 3. Holder A conducts training and verifies data quality;

[0119] 4. Obtain the feature distribution by sampling from the trained dataset;

[0120] 5. Calculate the distribution characteristics of each bucket set;

[0121] 6. Conduct verifiable data quality testing;

[0122] 7. Generate the commitment value and share private key proof;

[0123] 8. Verify the holder's legitimacy;

[0124] 9. Extract the multinomial commitments and calculate the distribution characteristics of the encoding;

[0125] 10. Compare with the characteristics of storage share distribution;

[0126] 11. Output data quality assurance or abnormal warning;

[0127] S3: Supports cross-protocol and multi-framework protocol implementations.

[0128] Reference Figure 4 This invention supports multiple implementation frameworks, primarily including PyTorch and TensorFlow implementations. Taking the PyTorch implementation as an example, the specific steps include:

[0129] S21. Generate the data holder's private key share;

[0130] S22. Sort the private key shares;

[0131] S23. Generate ciphertext commitments and coefficient commitments;

[0132] S24. Generate share proofs and zero-knowledge proofs;

[0133] S25. Divide the dataset and calculate the feature distribution;

[0134] S26. Encode the features and generate a ciphertext commitment;

[0135] S27. Send encrypted commitment information to the data requester.

[0136] This multi-framework support design enables the invention to be flexibly deployed in different computing environments, meeting the needs of different users and improving the system's versatility and practical value.

[0137] S4: A Verifiable Data Quality Proof Framework Based on Multinomial Commitments

[0138] Reference Figure 5 The data quality verification process of this invention mainly includes:

[0139] S41. Extract the ciphertext and coefficients from the multinomial commitment and calculate the corresponding distribution characteristics;

[0140] S42. Compare the calculated features with the features of the data holder to verify whether they are consistent.

[0141] The verification process employs zero-knowledge proof technology to ensure data privacy is not compromised. If the features are consistent and the zero-knowledge proof is valid, the data quality is considered normal; otherwise, the data quality is considered problematic, and the corresponding exception handling mechanism is triggered.

[0142] In practical applications, a tolerance threshold can be set; for example, data quality is considered acceptable if the feature deviation does not exceed 5%. This design takes into account reasonable fluctuations that may exist in real-world data, improving the system's practicality and robustness.

[0143] Example 2: A Data Quality Verification System for Secure Multi-Party Computation

[0144] Reference Figure 6This invention also proposes a data quality verification system for multi-party secure computation corresponding to the above method. The system comprises a system server 1, a data holder 2, and a data trainer 3. The system server 1 is connected to the data holder 2 and the data trainer 3 via network or wireless signals. The system server 1 is used to upload and download private and public data, and provides data quality verification technology to the data holder 2. The data holder 2 is used to generate a data holder privacy protection share, data quality, and data anomaly detection for private data. The data trainer 3 is used to train and verify data quality for public data. The system server 1 includes a data holder dataset module 11, a data trainer dataset module 12, a system dataset module 13, a multi-security protocol module 14, and an algorithm model module 15. The data holder dataset module 11, the data trainer dataset module 12, and the system dataset module 13 are signal-connected to the multi-security protocol module 14, and the multi-security protocol module 14 is signal-connected to the algorithm model module 15.

[0145] In the specific implementation, the functions of each module are as follows:

[0146] 1. Data Holder Dataset Module 11: Responsible for verifying and uploading the private dataset provided by Data Holder 2, ensuring that the data format and structure meet system requirements.

[0147] 2. Data Training Module 12: Responsible for training the dataset provided by Data Training Module 3, verifying the quality of the trained dataset, and detecting potential quality issues.

[0148] 3. System Dataset Module 13: Responsible for managing public datasets generated by the system, including data storage, access control, and update maintenance functions.

[0149] 4. Multi-security protocol module 14: Implements the core function of privacy protection for private data provided by data holder 2 and data trainer 3, including key technologies such as distributed private key share distribution, multinomial commitment generation and verification in the above methods.

[0150] 5. Algorithm Model Module 15: Responsible for training the model on the data uploaded by the system server 1, generating the model, and performing quality checks on the trained dataset to evaluate the impact of data quality on model performance.

[0151] The system operation process can be summarized as follows: Data holder 2 and data trainer 3 upload data to system server 1; multi-security protocol module 14 performs privacy protection processing on the data; algorithm model module 15 performs training and quality verification based on the processed data; and finally, the verification results are fed back to data holder 2 and data trainer 3. Throughout the process, the original data will not be leaked, and the system can effectively detect and respond to data quality issues.

[0152] This system supports flexible deployment modes and can run on cloud servers, edge devices, or hybrid environments. In scenarios with particularly high security requirements, blockchain technology can be introduced to record verification results and key operations on the blockchain, further enhancing the system's transparency and immutability.

[0153] To illustrate the implementation process of this invention more intuitively, a specific application scenario example is given below.

[0154] Suppose three institutions, A, B, and C, are data holders who want to jointly train a risk assessment model without sharing the original customer data. Each institution is concerned that the other institutions may provide poor-quality data, affecting the model's accuracy. In this case, the method and system of this invention can be used for data quality verification.

[0155] First, each of the three institutions categorizes its data features (such as customer age, income, credit history, etc.) into buckets, calculating the mean and distribution ratio of features within each bucket. Taking age features as an example, five buckets can be set up: [18-25], [26-35], [36-45], [46-60], and [61+], and the sample ratio and mean age within each bucket can be calculated.

[0156] Each institution then generates a share of its private key and distributes it to other institutions using the Shamir secret-sharing scheme. For example, using a (2,3) threshold scheme, any two institutions can reconstruct the secret by collaborating. Each institution computes a polynomial ciphertext commitment and a coefficient commitment, which are then stored in a mutually recognized authority (such as an industry regulatory body).

[0157] Next, each institution generates a zero-knowledge proof of its private key share and computes a multinomial commitment to the data distribution characteristics. During the verification phase, institution A can verify whether the data provided by institutions B and C conforms to the claimed distribution characteristics without accessing the original data.

[0158] The method of this invention can effectively detect the following data quality problems:

[0159] 1. Distribution skew: The data distribution is significantly different from the declared distribution;

[0160] 2. High proportion of outliers: The distribution ratio of some buckets is abnormal;

[0161] 3. Data falsification: The mean of features does not match the actual situation;

[0162] In this way, the three organizations can protect customer privacy while ensuring the quality of the data used in the computation, thereby improving the reliability and effectiveness of the final model.

[0163] This invention proposes a data quality verification method and system for multi-party secure computation. By innovatively combining multinomial zero-knowledge proofs, distributed private key share distribution, and multinomial commitment techniques, it achieves efficient data quality verification while protecting data privacy. The method supports multiple implementation frameworks, adapts to different computing environments, and possesses good versatility and scalability. Compared to existing technologies, this invention significantly improves privacy protection, verification reliability, and application flexibility, providing crucial technical support for the practical application of multi-party secure computation.

[0164] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A data quality verification method in multi-party secure computation, characterized in that... It includes the following steps: S1. A distributed private key share distribution method based on polynomial zero-knowledge verifiable methods: Given an access policy and a set of data holders, first determine the private key shares of the n data holders. Sort by number i, where i is the data holder's number, ranging from 1 to n, and n is the total number of data holders. This represents the private key share of the i-th data holder; Then, the Shamir secret-sharing scheme is used to generate a polynomial of degree t-1. Where x is the independent variable and p is a large prime number, For the secret to be shared, The coefficients are randomly selected polynomials, where j is the coefficient index, and the value of j ranges from 1 to t-1, where t is the threshold value. This represents the j-th coefficient of the polynomial; For each data holder i, calculate the value of x by substituting i into the polynomial f(x). This generates the private key share of the data holder. ; Compute the polynomial commitment of polynomial f(x) For each value of j from 0 to t-1, calculate g is the first generator, and h is the second generator. Let j be the coefficient of the polynomial. For the corresponding random number; Multinomial commitment Outsource storage to authoritative institutions; For each data holder i, their private key share is generated using the Schnorr protocol. Zero-knowledge proof The specific process is as follows: Data holder i selects a random number. ,calculate Calculate the challenge value Where H is a hash function, used to calculate the response. Thus, proof is obtained. ; Calculate private key share Zero-knowledge proof of the relationship between the polynomial f(x) and the polynomial f(x) ; Verifier's proof of private key share Verification is performed, specifically, the equation is verified. If the statement is invalid, an error warning will be issued. S2. Verifiable random sampling method based on multinomial commitment: Select data feature X, and divide the N data samples into buckets to obtain a set of K buckets. Where K is the total number of buckets, j is the bucket number, and the value of j ranges from 1 to K; For each bucket Calculate the mean of the features within each bucket, where j ranges from 1 to K. ,in Indicates bucket The number of samples in the bucket, where x represents the feature value of the sample in the bucket; Calculation bucket Distribution ratio , where N is the total number of data samples; Sort the bucket set V according to the bucket boundary values ​​to obtain each bucket. Serial number ; For each bucket j ranges from 1 to K, and the average feature value within the bucket is calculated using the encryption function Enc. and its serial number Encryption is performed to obtain the encrypted value. Where k is the encryption key, Indicates a connection operation; Calculate the hash value of all encrypted values Where H is a hash function; Encrypted bucket feature mean set And the hash value h is used as distributed ciphertext; For each data holder Let i range from 1 to n, where n is the total number of data holders. We calculate the distribution of their data across different buckets to obtain a set of distribution characteristics. ,in Indicates data holder In the bucket The sample proportion in the data, j ranges from 1 to K, and K is the total number of buckets; The calculation formula is: ,in Indicates data holder In the bucket The number of samples in Indicates data holder Total number of samples; For each encrypted bucket set Calculate its private key share using a secret sharing scheme. ; Based on private key share and distribution characteristics Calculate the verification share Where H is the hash function that maps to the points on the elliptic curve. This indicates a connection operation, where G is the base point of the elliptic curve; For the verified share of all data holders, calculate the aggregated ciphertext. The summation symbol This represents the summation of i from 1 to n, where n is the total number of data holders. Let be the polymerization coefficient, satisfying , Verified share for data holder i; Calculate the commitment value of the aggregation coefficient Where H is the hash function, Indicates a connection operation; Construct a verification polynomial based on the aggregated ciphertext m. ,in The coefficients of the polynomial are randomly selected, j ranges from 1 to t, t is the degree of the polynomial, and x is the independent variable; Calculate and verify the encrypted commitment of polynomial P Where Commit is the commitment function; Commitment to calculating polynomial coefficients Where H is the hash function, Indicates a connection operation; The data holder sends a encrypted commitment s and a coefficient commitment. ; S3, Multi-frame adaptation: It supports the implementation of steps S1 and S2 in the PyTorch and TensorFlow frameworks, where step S1 includes cryptographic operations such as polynomial generation, commitment computation, and zero-knowledge proof generation, and step S2 includes cryptographic operations such as data bucketing, encryption, and commitment computation. S4. Verification of distribution characteristics: The verifier receives the ciphertext commitment s and coefficient commitment sent by the data holder in step S2. Extract polynomial coefficient information; The distribution characteristics of the data are calculated based on the extracted coefficient information; Compare the calculated distribution characteristics with the distribution characteristics declared by the data holders. Compare the two to determine if they are consistent; If they match, then the validity of the zero-knowledge proof is further verified, i.e., the equation is verified. Whether it is valid, among which Calculate self-response , and Calculate the self-proof separately In and private key share ; If the equation holds true, then the private key share proof is valid and the data quality verification passes. If the distribution characteristics are inconsistent or the equation does not hold, the data quality verification fails and an anomaly warning is triggered.

2. The data quality verification method in multi-party secure computation according to claim 1, characterized in that... The S1 method specifically includes: S11, Generate the data holder's private key share; S12. A verifiable representation of the distributed private key share distribution, generating a commitment value; S13. Verifiable data quality proof; S14. Privacy Protection.

3. The data quality verification method in multi-party secure computation according to claim 2, characterized in that... S11 includes: 1A1) Select an access strategy. For a given access strategy, use the Shamir secret sharing scheme to distribute private key shares to obtain the shares of all n data holders; 1B1) Based on the private key shares of the n holders obtained, and according to the specified access structure, generate the key of the ciphertext matrix composed of all shares.

4. The data quality verification method in multi-party secure computation according to claim 2, characterized in that... S12 includes: 1A2) Data holder set; 1B2) Commit to outsourcing storage to an authoritative institution.

5. The data quality verification method in multi-party secure computation according to claim 2, characterized in that... S13 includes: 1A3) Proof that the holder has obtained a share of the private key; 1B3) Zero-knowledge proofs for computing ciphertext.

6. The data quality verification method in multi-party secure computation according to claim 2, characterized in that... S14 includes: 1A4) The verifier verifies the proof of the private key share, and issues an exception warning if the proof is invalid.

7. The data quality verification method in multi-party secure computation according to claim 1, characterized in that... The S2 method specifically includes: S21. Distribution characteristics encoding of the dataset; S22. Verifiable dataset distribution characteristics representation; S23. Verifiable data quality verification method based on multinomial commitment.

8. The data quality verification method in multi-party secure computation according to claim 7, characterized in that... S21 includes: 2A1) Select the feature X of the dataset. For a set of N data samples, take the value of each feature X in each sample as the feature vector x. Divide all feature vectors x into buckets to obtain a bucket set V. 2B1) Generate the mean of features within each bucket by dividing the buckets into buckets and then using the bucket set V; 2C1) For each bucket set V i The number of data samples in each bucket set is calculated, and by comparing it with the dataset N, the number of data samples in each bucket set V is obtained. i The proportion of data samples in the middle to the total, i.e., the bucket distribution proportion p. v .

9. The data quality verification method in multi-party secure computation according to claim 7, characterized in that... S22 includes: 2A2) For the generated bucket set V, first sort the bucket set to obtain the order l of each bucket set. Then, encrypt the feature mean of the bucket according to the order l to obtain a hash value h. The encrypted set of feature mean of the bucket and h are used as distributed ciphertext. 2B2) For a data set Di of a data holder, the proportion distribution of the bucket sets V of statistical feature X is obtained, thus yielding the proportion of each bucket set V. i The distribution feature set p i ; 2C2) For the encrypted bucket set V i Calculate its private key share d using a secret sharing scheme. i According to d i and p i The verification share was calculated; 2D2) For each data holder's share, calculate the ciphertext m formed by the aggregation of multiple shares, and its aggregation coefficient commitment value c. Use c and m as the shared dataset ciphertext value to generate a polynomial based on polynomial commitment. Finally, the holder sends its polynomial ciphertext commitment s and coefficient commitment c. i .

10. A data quality verification system for multi-party secure computation, used to implement the data quality verification method for multi-party secure computation as described in any one of claims 1-9, characterized in that... It includes a system server, a data holder, and a data trainer, wherein the system server is connected to the data holder and the data trainer via a network signal or a wireless signal. The system server is used to upload and download private and public data, and provides data quality verification technology to data holders. The data holder is used to generate a data holder privacy protection share, data quality, and data anomaly detection for private data; The data training method is used to train on public data and verify data quality. The system server includes a data holder dataset module, a data trainer dataset module, a system dataset module, a multi-security protocol module, and an algorithm model module. The data holder dataset module, the data trainer dataset module, and the system dataset module are signal-connected to the multi-security protocol module, and the multi-security protocol module is signal-connected to the algorithm model module.

Images