A method and system for encrypting credit data based on AI large models

Abstract

This invention belongs to the field of data encryption technology, specifically relating to a credit data encryption method and system based on an AI large-scale model. The method includes: acquiring historical credit records and extracting trustworthy records to identify trustworthy credit data associated with a benchmark user; obtaining the influence range of the encrypted data to determine target users; comparing the trustworthy credit data of the benchmark user with the trustworthy credit data of the target user to assess the impact of the encrypted data on the user's credit data usage; comparing a trustworthy reference value with a preset trustworthy reference range, and combining the influence weight of the encrypted data to determine whether the encrypted data possesses trustworthiness and output an evaluation result; and determining and outputting the final encrypted data based on the trustworthy reference value. This invention enhances the reliability and security of the encryption process by establishing an evaluation benchmark based on historical credit records, ensuring that the output encrypted data possesses high trustworthiness while maintaining security.

Description

Technical Field

[0001] This invention belongs to the field of data encryption technology, specifically relating to a credit data encryption method and system based on an AI large model. Background Technology

[0002] With the development of information technology, personal credit data has been deeply applied in many fields such as finance, e-commerce, and leasing, providing a data foundation for risk assessment, identity verification, and personalized services, while also promoting the development of the digital economy. Especially in financial consumption scenarios, credit data such as user income flow and repayment records are integrated to assess the credit risk of borrowers. E-commerce platforms can provide consumers with merchant credit references through historical data such as transaction records, reviews, and return / exchange information, thereby reducing transaction risks. However, due to the privacy sensitivity of credit data itself, in some scenarios, if there is insufficient technical protection and imperfect credit management, there are potential security risks associated with credit data.

[0003] Traditional symmetric or asymmetric encryption methods are mostly only used to encrypt data transmission in communication links. Once credit data is illegally obtained before encryption or after decryption, related privacy information is easily leaked, making it difficult to effectively defend against internal threats or advanced persistent threats. When encrypting large amounts of credit data, existing encryption technologies are prone to problems such as insufficient encryption strength, complex key management, cumbersome encryption process and low efficiency, making it difficult to meet the credit encryption processing needs of large-scale, high-concurrency scenarios.

[0004] During the sharing and transfer of credit data, when credit is frequently accessed for signatures or verifications, users' personal information is kept in a state of long-term surveillance. Users' sensitive information is easily used or abused maliciously, posing a potential threat to personal privacy rights. At the same time, the verification process can easily lead to the leakage of user privacy.

[0005] Therefore, this invention provides a credit data encryption method and system based on a large AI model. Summary of the Invention

[0006] The purpose of this invention is to provide a credit data encryption method and system based on an AI large model.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A credit data encryption method based on an AI large model includes the following steps: Obtain the user keywords corresponding to the data to be encrypted; Obtain user profile information based on user keywords; A user demand graph is constructed based on user profile information to determine whether the user demand graph contains data to be encrypted. Generate encrypted data based on the data to be encrypted; In response to the generation of encrypted data, a validity assessment is performed, and a trusted reference value is generated; Based on trusted reference values, determine and output the final encrypted data; In response to the generation of encrypted data, the validity assessment and generation of trusted reference values ​​include: identifying trusted credit data associated with the benchmark user and the target user respectively; and generating trusted reference values ​​by comparing the trusted credit data of the benchmark user and the target user. The process of identifying credible credit data associated with the baseline user and the target user includes: obtaining historical credit records; compiling credible records according to the chronological order of the historical credit records; and identifying credible credit data through the credible records.

[0008] Preferably, the identification of trusted credit data associated with the benchmark user and the target user respectively further includes: Identify the users corresponding to the trusted credit data as baseline users; obtain the scope of influence of the encrypted data, and identify the users within the scope of influence as target users.

[0009] Preferably, the step of identifying trustworthy credit data through trusted records includes: Extract credit scoring information from trusted records as trusted credit data.

[0010] Preferably, determining and outputting the final encrypted data based on a trusted reference value includes: The system compares the trusted reference value with the specified trusted reference range and combines the influence weight of the encrypted data to determine whether the encrypted data is trustworthy. When the encrypted data is trustworthy, the final encrypted data is output.

[0011] Preferably, generating encrypted data based on the data to be encrypted includes: Based on the preset encrypted data structure and the preset decrypted data structure, the degree of difference in the number of fields is analyzed; using the degree of difference as a reference, the content that needs to be encrypted is selected from the data to be encrypted; key information is generated, and the key information is used to encrypt the content that needs to be encrypted to generate encrypted data.

[0012] A credit data encryption system based on an AI large-scale model uses the aforementioned AI large-scale model-based credit data encryption method, and includes the following modules: The data encryption module is used to generate encrypted data based on the data to be encrypted. The trusted data identification module is used to obtain historical credit records; organize trusted records according to the time sequence of historical credit records; identify trusted credit data through trusted records; and determine the user corresponding to the trusted credit data as the benchmark user. The validity assessment module is used to obtain the scope of influence of encrypted data and identify users within the scope of influence as target users; by comparing the trusted credit data of benchmark users and target users, a trusted reference value is generated. The encryption output module is used to determine and output the final encrypted data based on a trusted reference value.

[0013] Preferably, the step of identifying trustworthy credit data through trusted records includes: Extract credit scoring information from trusted records as trusted credit data.

[0014] Preferably, determining and outputting the final encrypted data based on a trusted reference value includes: The system compares the trusted reference value with the specified trusted reference range and combines the influence weight of the encrypted data to determine whether the encrypted data is trustworthy. When the encrypted data is trustworthy, the final encrypted data is output.

[0015] Preferably, generating encrypted data based on the data to be encrypted includes: Based on the preset encrypted data structure and the preset decrypted data structure, the degree of difference in the number of fields is analyzed; using the degree of difference as a reference, the content that needs to be encrypted is selected from the data to be encrypted; key information is generated, and the key information is used to encrypt the content that needs to be encrypted to generate encrypted data.

[0016] Beneficial effects This invention obtains user keywords corresponding to the data to be encrypted, and uses these keywords to obtain user profile information, constructs a user demand graph, and filters the data to be encrypted based on the user demand graph. This ensures the relevance and effectiveness of the encryption process, avoids redundant encryption of non-user demand data, and improves data processing efficiency and resource utilization.

[0017] This invention, based on the degree of difference between a preset encrypted data structure and a preset decrypted data structure, filters out the content to be encrypted from the data to be encrypted, and uses key information to encrypt the content to be encrypted. It can identify and encrypt only the key content with structural differences, reduce the redundancy of encrypted data, improve encryption efficiency and data storage efficiency, and ensure the specificity and security of encrypted content.

[0018] This invention identifies credible credit data associated with a benchmark user by acquiring historical credit records and sorting out credible records from them; it determines the scope of influence of encrypted data to identify target users; it evaluates the impact of encrypted data on user credit data usage by comparing the credible credit data of the benchmark user with the credible credit data of the target user; and it verifies the rationality of encryption by establishing an evaluation benchmark based on historical credit records, thereby enhancing the reliability and security of the encryption process.

[0019] This invention compares a trusted reference value with a preset trusted reference range and combines the influence weight of the encrypted data to determine whether the encrypted data is trustworthy and outputs an evaluation result. Based on the trusted reference value, the final encrypted data is determined and output, and the trustworthiness of the encrypted data is evaluated to ensure the security and trustworthiness of the output encrypted data and avoid data distortion caused by improper encryption. Attached Figure Description

[0020] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system structure diagram of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example 1 Please refer to Figure 1 This embodiment provides a credit data encryption method based on an AI large model, including the following steps: S1. Data Acquisition Steps: The raw credit data to be encrypted can be obtained from one or more data sources, through a credit information database or a real-time data stream interface, and the raw credit data to be encrypted can be used as the data to be encrypted. Another set of credit data with the same or similar structure as the data to be encrypted is extracted from the database as offset data, thereby establishing a benchmark for differential analysis and effectiveness assessment; the offset data is used to provide a quantifiable reference for assessing the uniqueness and sensitivity of the data to be encrypted; the offset data can be selected from standardized and anonymized industry average credit records, or from a sample of credit data from another specific user group. By comparing the values ​​or types of the data to be encrypted and the offset data in the corresponding data fields item by item, and by calculating the difference between the values ​​of the two fields, the similarity or the degree of type mismatch, etc., as a difference measure, reference parameters are generated. The reference parameters serve as an important basis for filtering encrypted content and determining encryption strength.

[0023] S2. Data preparation steps: Obtain user keywords related to the data subject to determine the inherent needs and sensitive information boundaries of the individuals or entities associated with the data to be encrypted; by analyzing the historical behavior records of the data subject, specifically the data subject's transaction flow, product browsing history, service query records, etc., count the frequency of various words, and filter out words with a frequency exceeding a preset frequency threshold as user keywords that represent its core focus. The preset frequency threshold can be set to 3 times. Based on user keywords and user profile information such as basic attribute information of the data subject, a structured user demand graph is constructed. The basic attribute information includes user profile information such as gender, age range, occupation category, geographical region, and historical content preferences. In the user demand graph, nodes represent entities such as users, keywords, credit products or services, and edges represent the relationships between these entities such as ownership, browsing, and interest. The user demand graph transforms scattered user behavior data into an interconnected knowledge network. Determine whether the specific information items in the data to be encrypted already exist in the user demand graph as nodes or in a form strongly associated with key nodes, thereby determining whether the data to be encrypted belongs to the user's core needs or sensitive information category. The structural integrity of the user demand graph is judged, and the sufficiency of the user demand graph in characterizing user features is evaluated. Specifically, the topological structure index of the user demand graph can be calculated, and the proportion of isolated nodes in the user demand graph that have no connection with other nodes can be counted to the total number of nodes. Alternatively, the average connection density of the user demand graph can be calculated, which is the average of the ratio of the actual number of edges between all nodes in the graph to the maximum number of possible edges. If the proportion of isolated nodes is too high or the average connection density is lower than the preset integrity threshold, it indicates that the current user demand graph has failed to effectively capture the full picture of user needs; if the proportion of isolated nodes is low or the average connection density is higher than the preset integrity threshold, it is considered that the current user demand graph has met the integrity standard. The preset integrity threshold is a value based on the conventional values ​​of user demand graphs. Specifically, the preset integrity threshold can be 0.15 when it is the proportion of isolated nodes and 0.25 when it is the average connection density. The security sensitivity level of specific data can be assessed based on the local structural features of the graph. Specifically, the density of the number of edges and nodes within a graph sub-region that is directly associated with the data to be encrypted can be calculated as the local density. The local density reflects the closeness of the association between the data point and the user's core interest point. If the local density is greater than the preset density threshold, it indicates that the data point is closely connected to multiple core interest points of the user and has a very strong correlation. Therefore, the security sensitivity level of the data point is determined to be high. If the local density is not greater than the preset density threshold, the security sensitivity level of the data point is determined to be low. The security sensitivity level is used to guide the selection of encryption strategies. The preset density threshold is a common value for the user demand map, which can be 0.4.

[0024] S3. Steps to determine encrypted information: Filter out the sensitive parts of the data to be encrypted and encrypt them; reorganize the data to be encrypted to determine which data fields need to be encrypted. The process of reorganizing encrypted data includes: comparative analysis of preset decryption and encryption data structures; where both preset decryption and encryption data structures are predefined data models. The preset decryption data structure defines the complete set of fields of the original credit data, describing the complete set of fields and organization that the original credit data should have after decryption; the preset encryption data structure defines the set of fields that the encrypted data should contain, describing the set of fields and organization that the encrypted data should contain; the preset encryption data structure is usually a subset of the original data or a transformed form. By comparing the preset decryption data structure and the preset encryption data structure, fields that exist in the decryption data structure but not in the encryption data structure are identified as the content to be encrypted. The degree of difference between the preset decryption data structure and the preset encryption data structure is quantified by calculating the field overlap ratio between the preset decryption data structure and the preset encryption data structure. Specifically, if the decryption structure has 10 fields and the encryption structure has 5 fields, and all 5 fields are in the decryption structure, then the field overlap ratio is 50%, and the degree of difference is the sum of the field overlap ratio and 1. The calculated degree of difference reflects that the other 50% of the fields need to be hidden. Based on this degree of difference, and combined with an adjustable offset coefficient, which is dynamically adjusted according to the security sensitivity level, the screening criteria are tightened or relaxed to select fields from the data to be encrypted that have a confidentiality level higher than the preset confidentiality threshold; where the preset confidentiality threshold can be 0.5. Fields with a confidentiality level higher than the preset confidentiality threshold are selected as the content to be encrypted. Based on the selected content, combined with timestamps or generated random numbers, a unique key is generated using preset deterministic generation rules. The preset deterministic generation rules include: concatenating the field names and values ​​of all content to be encrypted into a long string, concatenating this string with the current timestamp accurate to milliseconds, using standard hash algorithms such as SHA-256 to calculate the concatenated string, generating a fixed-length hash value, and using this fixed-length hash value as the generated key information; The key information is used to encrypt the content that needs to be encrypted. Specifically, symmetric encryption algorithms such as AES-256 can be used. The generated key information is used as the key to encrypt the field value of each content that needs to be encrypted, so as to obtain the encrypted ciphertext and generate the final encrypted data.

[0025] S4. Steps to generate decryption information: The necessary information from the encryption process is archived to ensure that encrypted data can be accurately restored under authorized conditions; Specifically, the original value of the unencrypted data to be encrypted as a whole or the original value of the encrypted part is used as the core decryption data and stored securely. A secure association record is established, which binds the generated encrypted data, the key information used for encryption, and the corresponding decryption data. The association record is marked with a source mark, which may include the associated user ID, data generation timestamp, and business scenario identifier, etc., to trace the source and context of the data. When decryption is required, the corresponding key information and the original decrypted data can be found based on the identifier of the encrypted data, thereby completing the reliable restoration of the data.

[0026] S5. Steps for identifying trustworthy data: Extract the most representative and credible credit data from the user's historical credit records to provide a stable and reliable reference standard for evaluating the effectiveness of encryption. The system retrieves all historical credit records of a specific user and arranges them chronologically. It then uses preset filtering rules to identify reliable records. Specifically, the preset filtering rules include: selecting a series of records that are consecutive in time, where the time interval between any two adjacent records is no greater than a preset maximum time span, and the fluctuation range of key indicators such as credit scores in the records is less than a preset stability threshold. The preset maximum time span can be 35 days, allowing for a few days of delay within a month, and the preset stability threshold can be 15 points when considering credit scores. Continuous record segments that meet the preset filtering rules are identified as trustworthy records. Credit scoring information or other core credit indicators extracted from trustworthy records together constitute the user's trustworthy credit data. Select one or more users with high-quality, trustworthy credit data and designate them as benchmark users. Use the trustworthy credit data of the benchmark users as the standard for evaluating changes in other users' data and the effectiveness of encryption.

[0027] S6. Effectiveness assessment steps: The effectiveness assessment quantifies the potential impact of encryption operations on the credit assessment of target users; the scope of the impact of encrypted data is determined, which includes all users who may use the encrypted data in their credit assessment process, and these users are defined as target users, who are the objects of the assessment of the impact of encryption operations; The trusted credit data of the selected benchmark user and each target user are compared. The comparison process includes: extracting a set of predefined multi-dimensional credit features from the trusted credit data of the benchmark user and the target user respectively; and forming a credit feature vector corresponding to each credit feature by calculating statistical values ​​such as the average credit score, score variance, and delinquency frequency of the credit features. The predefined multi-dimensional credit features may include the average credit score, score variance, frequency of occurrence of specific negative events such as delinquency, and frequency of credit inquiries. By calculating the quantitative difference between the credit feature vector of the target user and the credit feature vector of the benchmark user, specifically by calculating the Euclidean distance or Mahalanobis distance between the two vectors, an initial numerical comparison result is generated. The comparison result reflects the degree of deviation between the credit profile of the target user and the benchmark profile, which serves as the standard trusted profile, under the background of encrypted operation.

[0028] S7. Steps to return to the reference result: The generated comparison results are converted into standardized evaluation scores, and the comparison results of credible credit data are calculated and obtained. According to the preset conversion rules, the comparison results are mapped to a standardized numerical range, which can be between 0 and 1, to generate the final credible reference value. The preset conversion rules are normalization rules, which specifically calculate the comparison ratio between the comparison results and the preset maximum possible distance. The sum of the comparison ratio and the credible reference value is 1. The preset maximum possible distance is an empirical value, which can be 500. The trusted reference value intuitively represents the matching effect or degree of influence of encrypted data on the target user's credit data. The higher the trusted reference value, the smaller the influence and the better the matching effect.

[0029] S8. Steps to output the encryption result: Before providing the final encrypted data to the user or downstream application, a final verification is performed, comparing the generated trusted reference value with a preset trusted reference range, where the preset trusted reference range can be [0.8, 1.0]. If the trusted reference value falls within the preset trusted reference range, the currently generated encrypted data is deemed trustworthy, and its impact on credit assessment is within an acceptable range. The assessment result is passed, and the encrypted data is marked as the final encrypted data and allowed to be provided to the user. If the trusted reference value is lower than the lower limit of the range, the encrypted data is deemed untrustworthy, the evaluation result is unsuccessful, and a correction instruction is generated. The correction instruction may include: adjusting the offset coefficient to change the selection of the encrypted field, or adjusting the strength of the encryption transformation. Then, the encryption and subsequent evaluation processes are re-executed to form a closed-loop optimization process until the generated encrypted data passes the trustworthiness evaluation.

[0030] Example 2 Please refer to Figure 2 This embodiment provides a credit data encryption system based on an AI large model. In its implementation, this system can be deployed on cloud servers, data centers, back-end servers of financial institutions, or dedicated data processing hardware, and includes the following modules: The data encryption module is used to generate initial encrypted data based on the data to be encrypted. When a piece of data to be encrypted is received, a preliminary relevance judgment is performed to obtain the user keywords of the user associated with the data to be encrypted. The user keywords include the user's identity identifier, recent query behavior, etc. Based on user keywords, the corresponding user profile information is obtained from the user profile database, and a user demand graph reflecting the user's current needs is constructed based on the user profile information; it is then determined whether the data to be encrypted is included in the user demand graph to confirm that the data is related to the user's real needs. After confirming the data correlation, prepare for encryption; based on the preset encrypted data structure and the preset decrypted data structure, analyze the degree of difference in the number of fields between the two; calculate the proportion of non-overlapping fields between the two structures, and then subtract the overlap proportion of the field from 1 to determine the degree of difference; The degree of difference in the number of these fields is used as an important reference to filter out the content that needs to be encrypted from the data to be encrypted; among them, fields with large differences that only exist in the encrypted structure are considered sensitive information, and the sensitive information is identified as the content that needs to be encrypted. Generate key information, which can be a symmetric encryption key or an asymmetric encryption public and private key pair. Then, call AI large-scale model encryption algorithms such as AES and RSA to use the key information to perform encryption transformation on the selected content that needs to be encrypted, thereby generating the initial encrypted data.

[0031] The trusted data identification module is used to identify trusted credit data that can be used as an evaluation benchmark and the corresponding benchmark users based on historical data; it retrieves the historical credit records of specific users from the historical credit database, including loan history, repayment records, credit card usage, utility payment records, etc. Historical credit records are organized and analyzed in chronological order to identify credible records; by comparing users' credit behavior at multiple consecutive or key time points, consistent and stable record patterns are sought; specifically, credit behavior may include records of users consistently making timely payments and having no overdue payments over the past few years, and these continuous and positive records will be integrated and identified as credible records. After identifying credible records, key credit scoring information, such as credit scores assessed by authoritative institutions and credit ratings internally evaluated, is extracted from these records. Among them, the credit scoring information is established as credible credit data because it originates from stable and reliable historical behavior. Users associated with high-level credible credit data are identified as benchmark users, whose credit data represents a reference standard for good and stable credit status.

[0032] The validity assessment module is activated after the initial encrypted data is generated. It is responsible for the core assessment of the validity and credibility of the encrypted data. It identifies the target users as the assessment objects, obtains the generated initial encrypted data, and analyzes the potential scope of impact once the data takes effect. When the initial encrypted data is new loan approval data, its scope of impact includes the user, related guarantors, and relevant financial institutions. All users within this scope of impact are identified as target users. Entering the comparison phase, the trusted credit data of the baseline user determined by the trusted data identification module is obtained, and the trusted credit data of the target user within the scope of influence is also obtained. For the target user, a method similar to that of the trusted data identification module is used to identify the corresponding trusted credit data based on their own historical credit records. After acquiring data from both sides, a reliable reference value is generated by comparing the reliable credit data of the benchmark user with that of the target user. The comparison process is a quantitative calculation process, which can calculate the difference or ratio between the two credit scores, or calculate a comprehensive similarity score based on multi-dimensional credit features. The final generated reliable reference value intuitively reflects the gap or correlation between the target user's credit status and the recognized reliable benchmark.

[0033] The encryption output module is used to determine whether to finally confirm and output encrypted data based on the evaluation results generated by the validity evaluation module; it receives the generated trusted reference value and compares the trusted reference value with the preset trusted reference range; wherein, the trusted reference range defines the acceptable credit deviation threshold range. To make the assessment more comprehensive, the impact weight of the initial encrypted data is also taken into account. The impact weight is a quantitative indicator used to represent the potential impact of the encrypted data on the credit system. Specifically, the weight of a large corporate loan will be much higher than that of a small personal consumer loan. The credibility of the initially generated encrypted data is determined based on whether the trusted reference value falls within the trusted reference range and the magnitude of the influence weight of the encrypted data. Specifically, the encrypted data is deemed to be credible only if the trusted reference value is within an acceptable range and the influence weight of the encrypted data does not exceed a certain security threshold. When the judgment result is deemed trustworthy, the initial encrypted data is output as the final encrypted data, and the final encrypted data is allowed to enter the subsequent storage, transmission or application stages. If the judgment result is that it is not trustworthy, the initial encrypted data will be intercepted and not output, thereby preventing potential low-quality or high-risk data from causing data pollution.

[0034] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from it. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A credit data encryption method based on an AI large-scale model, characterized in that, Includes the following steps: Obtain the user keywords corresponding to the data to be encrypted; Obtain user profile information based on user keywords; A user demand graph is constructed based on user profile information to determine whether the user demand graph contains data to be encrypted. Generate encrypted data based on the data to be encrypted; In response to the generation of encrypted data, a validity assessment is performed, and a trusted reference value is generated; Based on trusted reference values, determine and output the final encrypted data; In response to the generation of encrypted data, the validity assessment and generation of trusted reference values ​​include: identifying trusted credit data associated with the benchmark user and the target user respectively; and generating trusted reference values ​​by comparing the trusted credit data of the benchmark user and the target user. The process of identifying credible credit data associated with the baseline user and the target user includes: obtaining historical credit records; compiling credible records according to the chronological order of the historical credit records; and identifying credible credit data through the credible records.

2. The credit data encryption method based on an AI large model according to claim 1, characterized in that, The identification of trusted credit data associated with the baseline user and the target user respectively also includes: Identify the users corresponding to the trusted credit data as baseline users; obtain the scope of influence of the encrypted data, and identify the users within the scope of influence as target users.

3. The credit data encryption method based on an AI large model according to claim 1, characterized in that, The identification of credible credit data through trusted records includes: Extract credit scoring information from trusted records as trusted credit data.

4. The credit data encryption method based on an AI large model according to claim 1, characterized in that, The process of determining and outputting the final encrypted data based on a trusted reference value includes: The system compares the trusted reference value with the specified trusted reference range and combines the influence weight of the encrypted data to determine whether the encrypted data is trustworthy. When the encrypted data is trustworthy, the final encrypted data is output.

5. The credit data encryption method based on an AI large model according to claim 1, characterized in that, The step of generating encrypted data based on the data to be encrypted includes: Based on the preset encrypted data structure and the preset decrypted data structure, the degree of difference in the number of fields is analyzed; using the degree of difference as a reference, the content that needs to be encrypted is selected from the data to be encrypted; key information is generated, and the key information is used to encrypt the content that needs to be encrypted to generate encrypted data.

6. A credit data encryption system based on an AI large-scale model, using a credit data encryption method based on an AI large-scale model as described in any one of claims 1-5, characterized in that, Includes the following modules: The data encryption module is used to generate encrypted data based on the data to be encrypted. The trusted data identification module is used to obtain historical credit records and organize trusted records according to the time sequence of historical credit records. Trustworthy credit data is identified through trusted records; Identify users whose credit data corresponds to benchmark users; The validity assessment module is used to obtain the scope of influence of encrypted data and identify users within the scope of influence as target users; by comparing the trusted credit data of benchmark users and target users, a trusted reference value is generated. The encryption output module is used to determine and output the final encrypted data based on a trusted reference value.

7. A credit data encryption system based on an AI large model according to claim 6, characterized in that, The identification of credible credit data through trusted records includes: Extract credit scoring information from trusted records as trusted credit data.

8. A credit data encryption system based on an AI large model according to claim 6, characterized in that, The process of determining and outputting the final encrypted data based on a trusted reference value includes: The system compares the trusted reference value with the specified trusted reference range and combines the influence weight of the encrypted data to determine whether the encrypted data is trustworthy. When the encrypted data is trustworthy, the final encrypted data is output.

9. A credit data encryption system based on an AI large model according to claim 6, characterized in that, The step of generating encrypted data based on the data to be encrypted includes: Based on the preset encrypted data structure and the preset decrypted data structure, the degree of difference in the number of fields is analyzed; using the degree of difference as a reference, the content that needs to be encrypted is selected from the data to be encrypted; key information is generated, and the key information is used to encrypt the content that needs to be encrypted to generate encrypted data.

Images