Encrypted data quality declaration verification method based on AI and privacy computing technology

Abstract

The invention discloses an encrypted data quality declaration verification method based on an AI and privacy computing technology, and relates to the technical field of data encryption, privacy computing and artificial intelligence. Performing normalization, denoising and feature selection preprocessing operation on the original data, and encrypting the preprocessed data through a homomorphic encryption technology to obtain encrypted data; and quality declaration generation: performing quality evaluation on the original data according to a predefined data quality evaluation standard, and generating a quality declaration document containing multiple dimensions of accuracy, integrity, consistency, timeliness and uniqueness. According to the encrypted data quality declaration verification method based on the AI and the privacy computing technology, in the whole verification process, the original data is always in an encrypted state, and data privacy leakage is effectively prevented.

Description

Technical Field

[0001] The present invention relates to the technical fields of data encryption, privacy computing, and artificial intelligence, and specifically provides an encrypted data quality statement verification method based on AI and privacy computing technologies. Background Art

[0002] With the development of digitalization, data privacy protection has become increasingly important. Currently, there are the following technical problems between data quality verification and privacy protection:

[0003] 1. The contradiction between privacy protection and quality verification: Although traditional data encryption technologies (such as symmetric encryption and asymmetric encryption) can protect data privacy, when the data is made public, the quality of the encrypted data cannot be directly verified, which may lead to data users' distrust of data quality.

[0004] 2. The limitations of existing privacy computing technologies: Although privacy computing technologies such as homomorphic encryption, secure multi-party computing, and zero-knowledge proof provide a solution for computing while protecting data privacy, these technologies are not sufficiently applied in the scenario of data quality verification, and there is a lack of an effective method for specifically verifying encrypted data quality statements.

[0005] 3. The lack of a privacy computing scheme combined with AI technology: Existing privacy computing technologies lack deep integration with AI technology, resulting in the inability to fully utilize the advantages of AI in data feature extraction and pattern recognition for data quality assessment. Summary of the Invention

[0006] The purpose of the present invention is to provide an encrypted data quality statement verification method based on AI and privacy computing technologies to solve the problems raised in the above background art.

[0007] To achieve the above purpose, the present invention provides the following technical solution: An encrypted data quality statement verification method based on AI and privacy computing technologies, comprising the following steps:

[0008] Data preprocessing and encryption: Perform preprocessing operations such as normalization, denoising, and feature selection on the original data, and encrypt the preprocessed data through homomorphic encryption technology to obtain encrypted data;

[0009] Quality statement generation: According to predefined data quality assessment criteria, evaluate the quality of the original data, and generate a quality statement document including multi-dimensions such as accuracy, integrity, consistency, timeliness, and uniqueness;

[0010] Feature extraction and model training: Based on deep learning technology, extract key features that can reflect data quality from the original data, construct a neural network model, and perform supervised learning with the features as input and the quality assessment results as output. Improve the accuracy and generalization ability of the model through cross-validation and model tuning techniques;

[0011] Model encryption and disclosure: Encrypt the trained quality assessment model through homomorphic encryption technology, develop a model integrity verification mechanism based on zero-knowledge proof, and disclose the encrypted model together with the encrypted data to the public;

[0012] Quality verification: The public uses the encrypted quality assessment model to verify the quality of the encrypted data. Without decrypting, use the encrypted model to calculate the encrypted data through homomorphic encryption technology, and compare the calculation results with the quality statement provided by the encryptor to judge the authenticity of the quality statement.

[0013] Furthermore, data preprocessing: Detect and handle missing values, and select filling or deletion strategies according to data characteristics; Identify and handle outliers, detect them using box plots and Z-score methods, standardize or normalize the data to make features in different dimensions comparable; Use correlation analysis and variance analysis methods to select key features, encode categorical features, process time features, and create composite features that contribute to data quality assessment; Divide the data into training set, validation set and test set to ensure the consistency and representativeness of the data distribution in each set.

[0014] Furthermore, extract the following features: statistical features, structural features, and domain features.

[0015] Furthermore, the statistical features:

[0016] Basic statistics: mean, variance, skewness, kurtosis;

[0017] Distribution features: entropy, Gini coefficient, quantiles;

[0018] Correlation features: correlation coefficient matrix between features.

[0019] Furthermore, the structural features include data pattern consistency features, missing value distribution features, and data type consistency features.

[0020] Furthermore, the domain features include business rule compliance features, domain knowledge constraint features, and semantic consistency features.

[0021] Furthermore, model encryption includes weight encryption and polynomial coefficient encryption.

[0022] Furthermore, the weight encryption uses homomorphic encryption to encrypt the model weights.

[0023] Furthermore, the polynomial coefficient encryption encrypts the coefficients of the approximate polynomial of the activation function.

[0024] The present invention provides an encrypted data quality statement verification method based on AI and privacy computing technologies, having the following beneficial effects: Throughout the verification process, the original data is always in an encrypted state, effectively preventing data privacy leakage; extracting key quality features of the data through AI technologies makes the quality verification more comprehensive and accurate; any third party can verify the quality statement of the encrypted data, enhancing the credibility of data exchange; compared with traditional encrypted verification methods, the quality assessment is performed through an AI model, greatly improving the verification efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a system architecture diagram of an encrypted data quality statement verification method based on AI and privacy computing technologies of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] The following further describes in detail the embodiments of the present invention with reference to the drawings. The following embodiments are used to illustrate the present invention, but cannot be used to limit the scope of the present invention.

[0027] Refer to the attached Figure 1 drawings, and now three encrypted data quality statement verification methods provided by the present invention based on AI and privacy computing technologies are described.

[0028] Embodiment 1

[0029] The first encrypted data quality statement verification method based on AI and privacy computing technologies is used for medical data quality verification:

[0030] Background: Medical institutions hope to share encrypted patient data for medical research, but need to prove to researchers that the data quality meets specific standards;

[0031] The implementation steps are as follows:

[0032] Data preparation and preprocessing: Collect a medical data set containing patient basic information, diagnosis results, examination records, etc., desensitize and standardize the data, and evaluate the quality of the data set in dimensions such as accuracy and integrity;

[0033] Quality statement generation: The medical institution generates the following quality statements: Accuracy: 95.2% (diagnosis accuracy rate); Integrity: 97.8% (completeness of required fields); Consistency: 94.5% (cross-table record consistency); Timeliness: 98.1% (timeliness of data update); Uniqueness: 99.9% (patient identification accuracy rate);

[0034] Model Training and Encryption: Extract medical data features, including numerical distribution, missing patterns, temporal features, etc., train a medical data quality assessment model, achieving an accuracy of 93.5% on the test set, and encrypt the data and model using the BGV homomorphic encryption scheme;

[0035] Public Verification: Researchers obtain the encrypted data and encrypted model, verify the data quality through homomorphic computing, and obtain the verification result: qualified. The verification result shows that the quality of the medical data meets the declared standards and can be used for research;

[0036] Implementation Effect: This implementation plan enables medical institutions to prove to researchers that the data quality is reliable without disclosing patient privacy, promoting the secure sharing and utilization of medical data.

[0037] Example Two

[0038] The Second Encrypted Data Quality Declaration Verification Method Based on AI and Privacy Computing Technology for Financial Risk Control Data Quality Verification:

[0039] Background: Financial institutions need to exchange customer credit assessment data, but need to protect customer privacy and ensure the credibility of data quality.

[0040] Implementation Steps are as follows:

[0041] Data Preparation and Preprocessing: Collect information including customer transaction records, credit scores, repayment history, etc., apply differential privacy technology such as noise addition for preliminary desensitization, and calculate data quality indicators for each dimension;

[0042] Quality Declaration Generation: Accuracy: 96.3% (risk assessment accuracy rate), Completeness: 98.2% (keyword field completeness), Consistency: 95.7% (transaction record consistency), Timeliness: 99.1% (data update timeliness), Uniqueness: 99.8% (customer unique identification rate);

[0043] Feature Engineering and Model Optimization: Extract financial data features, focus on abnormal transaction patterns and credit risk indicators, adopt an ensemble learning method to improve the robustness of the model, and perform pruning optimization on the model to improve the efficiency of homomorphic computing;

[0044] Encryption and Verification: Implement a hierarchical encryption strategy, use different intensities of encryption for data with different sensitivities, design an incremental verification mechanism to support rapid re-verification after data update, adopt secure multi-party computing technology in the verification process, and introduce a trusted third party to enhance the credibility of verification;

[0045] Implementation Effect: This solution promotes the secure sharing of data among financial institutions, ensures the quality credibility of risk control data while protecting customer privacy, and significantly improves the efficiency of joint risk control.

[0046] Example 3

[0047] The third method for verifying the quality statement of encrypted data based on AI and privacy computing technology is used for verifying the quality of government data opening:

[0048] Background: Government departments hope to open some data for public use, but need to protect sensitive information and prove the quality of the open data;

[0049] The implementation steps are as follows:

[0050] Data classification and processing: Classify the sensitivity of government data, aggregate and anonymize high-sensitivity data, and establish data quality assessment criteria and processes;

[0051] Multi-level quality statements: Generate detailed quality statements for different types of data (demographics, economic data, public services, etc.), and use blockchain technology to record historical quality statements to ensure that the statements cannot be tampered with;

[0052] Distributed verification mechanism: Adopt the federated learning method to train the quality assessment model distributively, implement a public verification mechanism based on zero-knowledge proof, and design an incentive mechanism for crowdsourcing verification of data quality;

[0053] Transparency and accountability: Establish a platform for publicizing the quality verification results, implement a full-process log record of the verification process, and provide an automatic reporting mechanism for verification exceptions.

[0054] Implementation effect: This solution realizes the secure opening and quality transparency of government data, improves the public's trust and utilization rate of government data, and promotes data-driven public service innovation.

[0055] The embodiments of the present invention are given for purposes of illustration and description, and are not exhaustive or limit the present invention to the disclosed form. Many modifications and variations are obvious to those of ordinary skill in the art. The embodiments are chosen and described in order to best explain the principles of the present invention and its practical application, and to enable those of ordinary skill in the art to understand the present invention and design various embodiments with various modifications suitable for specific purposes.

Claims

1. A method for verifying the quality statement of encrypted data based on AI and privacy computing technologies, characterized in that, It includes the following steps: Data preprocessing and encryption: Perform normalization, denoising, and feature selection preprocessing operations on the original data, and encrypt the preprocessed data through homomorphic encryption technology to obtain encrypted data; Quality statement generation: According to predefined data quality evaluation criteria, evaluate the quality of the original data, and generate a quality statement document including multi-dimensions such as accuracy, integrity, consistency, timeliness, and uniqueness; Feature extraction and model training: Based on deep learning technology, extract key features that can reflect data quality from the original data, construct a neural network model, perform supervised learning with the features as the input and the quality evaluation results as the output, and improve the accuracy and generalization ability of the model through cross-validation and model tuning techniques; Model encryption and disclosure: Encrypt the trained quality evaluation model through homomorphic encryption technology, develop a model integrity verification mechanism based on zero-knowledge proof, and disclose the encrypted model and the encrypted data to the public together; Quality verification: The public uses the encrypted quality evaluation model to perform quality verification on the encrypted data. Without decrypting, use the encrypted model to calculate the encrypted data through homomorphic encryption technology, compare the calculation results with the quality statement provided by the encryptor, and judge the authenticity of the quality statement.

2. The encrypted data quality statement verification method based on AI and privacy computing technology according to claim 1, wherein Data preprocessing: Detect and handle missing values, and select filling or deletion strategies according to data characteristics; Identify and handle outliers, detect them using box plots and Z-score methods, standardize or normalize the data to make features in different dimensions comparable; Use correlation analysis and variance analysis methods to select key features, encode categorical features, process time features, and create composite features that are helpful for data quality evaluation; Divide the data into training set, validation set, and test set to ensure the consistency and representativeness of the data distribution in each set.

3. The encrypted data quality statement verification method based on AI and privacy computing technology according to claim 1, wherein Extract the following features: Statistical features, structural features, domain features.

4. The encryption data quality statement verification method based on AI and privacy computing technology according to claim 3, characterized in that The described statistical features: Basic statistics: mean, variance, skewness, kurtosis; Distribution features: entropy, Gini coefficient, quantile; Correlation features: correlation coefficient matrix between features.

5. The method for verifying the quality statement of encrypted data based on AI and privacy computing technology according to claim 4, characterized in that, The described structural features include data pattern consistency features, missing value distribution features, and data type consistency features.

6. The method for verifying the quality statement of encrypted data based on AI and privacy computing technology according to claim 4, characterized in that, The described domain features include business rule compliance features, domain knowledge constraint features, and semantic consistency features.

7. The encryption data quality statement verification method based on AI and privacy computing technology according to claim 1, wherein, Model encryption includes weight encryption and polynomial coefficient encryption.

8. The method for verifying the quality statement of encrypted data based on AI and privacy computing technology according to claim 7, wherein The described weight encryption uses homomorphic encryption to encrypt the model weights.

9. A method for verifying the quality statement of encrypted data based on AI and privacy computing technology according to claim 7, characterized in that, The described polynomial coefficient encryption encrypts the coefficients of the activation function approximation polynomial.

Images