A used car data collaborative query method and system based on privacy computing

Abstract

The application provides a used car data collaborative query method and system based on privacy computing, belonging to the technical field of data processing and information security. The application first acquires vehicle full-life-cycle maintenance data and regional macroeconomic supply and demand data, and constructs a multi-dimensional attribute set through feature vectorization and normalization processing. Then, the distributed Lagrange interpolation secret sharing technology is used, the attribute set is decomposed into logical fragments as polynomial constant terms, and distributed to each computing node. Each node reconstructs the polynomial coefficient based on the logical fragment and the market weight factor, generates attribute association ciphertext. Further, the homomorphic encryption technology is used to perform convolution operation of the attribute association ciphertext and the residual value weight matrix, to generate the estimated value ciphertext. Finally, the ciphertext is desensitized and restored by the multi-party consensus decryption operator to obtain the collaborative query estimated value of the vehicle. The application realizes the precise collaborative evaluation of the value of the used car integrating the regional market fluctuations.

Description

Technical Field

[0001] This application belongs to the field of data processing and information security technology, and in particular relates to a collaborative query method and system for used car data based on privacy computing. Background Technology

[0002] The privacy-preserving computation-based collaborative query method for used car data, through technologies such as multi-party secure computation, can achieve joint verification and value assessment of vehicle condition information without disclosing the original data. This method has broad application prospects in addressing information asymmetry in used car transactions, ensuring the secure flow of data elements, and promoting trust building in cross-regional transactions.

[0003] Existing technologies typically employ privacy set intersection or simple homomorphic encryption techniques to perform ciphertext matching on vehicle maintenance records or insurance data distributed across different institutions to verify whether a vehicle has a history of accidents or traffic violations. These methods primarily focus on Boolean queries or simple statistical counting of static historical data, and the process largely relies on direct comparison of single-dimensional vehicle condition data.

[0004] However, existing methods often overlook the dynamic impact of macroeconomic indicators and supply-demand fluctuations in the transaction location on the vehicle's final value, making it difficult to achieve weighted fusion calculations of multi-dimensional physical vehicle condition and regional market characteristics in encrypted form. This limitation results in the inability to output accurate valuations that conform to local market conditions while protecting data privacy in complex scenarios such as cross-regional registration transfers. Therefore, existing technologies suffer from technical problems such as low valuation accuracy and the inability to effectively integrate dynamic market data during cross-regional collaborative queries, leading to insufficient collaborative query capabilities. Summary of the Invention

[0005] The purpose of this application is to provide a collaborative query method and system for used car data based on privacy computing, so as to solve the problem of insufficient collaborative query in the prior art.

[0006] To address the aforementioned technical problems, in a first aspect, this application provides a collaborative query method for used car data based on privacy computing, comprising: The query node synchronously obtains vehicle maintenance data throughout the entire life cycle of the vehicle, as well as macroeconomic indicators and supply and demand fluctuation data of used cars in the transaction area; The query node performs numerical mapping on vehicle maintenance data through feature vectorization to obtain vehicle condition rating levels. The vehicle condition rating levels are then mapped to a preset evaluation space along with normalized macroeconomic indicators and supply and demand fluctuation data for feature concatenation, resulting in a multi-dimensional attribute set. The query node uses a multidimensional attribute set as the constant term of a secret shared polynomial constructed based on distributed Lagrange interpolation. By selecting multiple non-zero independent variable values ​​in a finite field and substituting them into the secret shared polynomial for evaluation, multiple independent logical partitions are generated and distributed to each computing node of the collaborative query. Each computing node uses the received logical fragments and preset market weight factors to obtain the attribute association ciphertext of the vehicle in the transaction area by distributively solving and reorganizing the coefficients of the secret shared polynomial. Each computing node uses homomorphic encryption to perform matrix convolution operations on the attribute association ciphertext and the preset residual weight matrix to obtain the valuation ciphertext for collaborative query. All nodes use the consensus decryption operator to de-identify and restore the valuation ciphertext to obtain the collaborative query valuation of the evaluated vehicle.

[0007] Optionally, the query node performs numerical mapping on vehicle maintenance data through feature vectorization to obtain a vehicle condition rating. This vehicle condition rating is then mapped to a preset evaluation space along with normalized macroeconomic indicators and supply and demand fluctuation data for feature concatenation, resulting in a multi-dimensional attribute set, including: Extract component wear identifiers from vehicle maintenance data, and based on these component wear identifiers, obtain the corresponding target impact weights according to the preset mapping relationship between wear identifiers and maintenance impact weights. The vehicle condition rating is obtained by weighted summation of the influence weights of all targets; Extract the historical extreme value ranges of macroeconomic indicators and supply and demand fluctuation data, and obtain the normalized macroeconomic indicators and supply and demand fluctuation data by calculating the distribution ratio of the current sampled values ​​of macroeconomic indicators and supply and demand fluctuation data within the historical extreme value ranges; By mapping vehicle condition rating levels, normalized macroeconomic indicators, and supply and demand fluctuation data to a preset multi-dimensional vector coordinate system and then concatenating the vectors, a multi-dimensional attribute set is obtained.

[0008] Optionally, the query node uses a multi-dimensional attribute set as the constant term of a secret shared polynomial constructed based on distributed Lagrange interpolation. By selecting multiple non-zero independent variable values ​​within a finite field and substituting them into the secret shared polynomial for evaluation, multiple independent logical slices are generated, including: The order of the secret shared polynomial is determined according to the preset security threshold, and the multidimensional attribute set is determined as the constant term of the secret shared polynomial. In a finite field, the number of random values ​​equal to the order is generated as the coefficients of the higher-order terms of the secret shared polynomial. Each higher-order term coefficient is combined with a constant term in ascending order of power to obtain the secret shared polynomial. Within a finite field, select multiple non-repeating coordinate values ​​of the independent variable as non-zero independent variable values, substitute these non-zero independent variable values ​​into a secret shared polynomial, perform exponential product and summation, and obtain the corresponding evaluation result. Logical slices are obtained by associating non-zero independent variable values ​​with their corresponding evaluation results as data units.

[0009] Optionally, each computing node utilizes the received logical fragments and preset market weight factors to obtain the attribute association ciphertext of the evaluated vehicle in the trading area by distributively solving and reorganizing the coefficients of the secret shared polynomial, including: Each computing node extracts the evaluation result from the received logical fragment and obtains the weighted fragment value based on the product of the evaluation result and the market weight factor. Each computing node obtains the corresponding interpolation basis operator by exchanging the non-zero independent variable values ​​in each logical segment and calculating the difference ratio between the non-zero independent variable values ​​and other non-zero independent variable values. The interpolation basis operator is then used to map and transform the weighted segment values ​​to obtain the corresponding attribute intermediate components. Each computing node uses a secure communication link to aggregate and reassemble the intermediate components of the attributes in an encrypted state to obtain the attribute-associated ciphertext.

[0010] Optionally, each computing node aggregates and reassembles the intermediate components of the attributes in an encrypted state using a secure communication link to obtain the attribute-associated ciphertext, including: Each computing node sends attribute intermediate components to other computing nodes through a secure communication link and receives attribute intermediate components sent by other computing nodes, thus obtaining an attribute component set that includes all attribute intermediate components. In the encrypted state, each computing node sums up all intermediate components of the attribute components in the attribute component set to obtain the recombined attribute value. Each computing node encapsulates the recombined attribute values ​​to obtain attribute-related ciphertext.

[0011] Optionally, each computing node uses homomorphic encryption to perform matrix convolution operations on the attribute association ciphertext and a preset residual weight matrix to obtain the estimated ciphertext for the collaborative query, including: Map the intermediate components of the attributes in the attribute association ciphertext to a preset multidimensional ciphertext coordinate system to obtain a ciphertext matrix with the same dimension as the residual weight matrix. Each computing node calculates the product of the values ​​of the intermediate component of each attribute in the ciphertext matrix and the corresponding weight component in the residual weight matrix within the encrypted space to obtain the encrypted weight value. Each computing node sums up all the encrypted weight values ​​to obtain the estimated ciphertext.

[0012] Optionally, all nodes include query nodes and each compute node; All nodes use the consensus decryption operator to anonymize and restore the valuation ciphertext, obtaining the collaborative query valuation of the evaluated vehicle, including: All nodes use the corresponding preset decryption component to perform decryption operations on the estimated ciphertext to obtain the corresponding decryption shared component; All nodes use a secure communication link to aggregate each decrypted shared component, resulting in a shared component set that includes all decrypted shared components; All nodes use the decryption operator to reorganize the shared component set to obtain the collaborative query estimate of the evaluated vehicle.

[0013] Secondly, this application provides a collaborative query system for used car data based on privacy computing, including: The acquisition module is used to query nodes to synchronously acquire vehicle maintenance data throughout the entire life cycle of the vehicle being evaluated, as well as macroeconomic indicators and supply and demand fluctuation data of used cars in the trading area. The mapping module is used to query nodes to perform numerical mapping on vehicle maintenance data through feature vectorization to obtain vehicle condition rating levels. The vehicle condition rating levels are then mapped to a preset evaluation space along with normalized macroeconomic indicators and supply and demand fluctuation data for feature concatenation to obtain a multi-dimensional attribute set. The evaluation module is used to query the constant term of the secret shared polynomial constructed by the node using a multi-dimensional attribute set as a distributed Lagrange interpolation. By selecting multiple non-zero independent variable values ​​in a finite field and substituting them into the secret shared polynomial for evaluation, multiple independent logical partitions are generated and distributed to each computing node of the collaborative query. The generation module is used by each computing node to obtain the attribute association ciphertext of the vehicle in the trading area by distributively solving and reorganizing the coefficients of the secret shared polynomial using the received logical fragments and preset market weight factors. The generation module is also used by each computing node to perform matrix convolution operation on the attribute association ciphertext and the preset residual weight matrix using homomorphic encryption to obtain the valuation ciphertext for collaborative query. All nodes use the consensus decryption operator to desensitize and restore the valuation ciphertext to obtain the collaborative query valuation of the evaluated vehicle.

[0014] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is configured to execute the computer program to implement the steps of the privacy-based computation-based collaborative query method for used car data as described in the first aspect above.

[0015] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the privacy-based computation-based collaborative query method for used car data as described in the first aspect above.

[0016] The privacy-preserving computation-based collaborative query method for used car data provided in this application first constructs a multi-dimensional attribute set that covers both physical vehicle condition and dynamic market characteristics by combining the vehicle's full lifecycle maintenance data with the macroeconomic indicators and supply and demand fluctuation data of the transaction area. This breaks through the limitations of traditional methods that rely solely on static historical data.

[0017] Subsequently, by utilizing distributed Lagrange interpolation secret sharing technology and homomorphic encryption algorithm, under the condition that each computing node only holds logical shards and does not access the original data, the deep integration calculation of multi-dimensional attributes, market weight factors and residual models was completed collaboratively, realizing the logical coupling of physical vehicle condition and market environment in the ciphertext space.

[0018] This approach, while strictly protecting vehicle owner privacy and trade secrets, dynamically adjusts valuation results based on real-time supply and demand in the transaction location, significantly improving the objectivity and accuracy of vehicle valuation in cross-regional transfer scenarios. Therefore, this application effectively solves the technical problem of insufficient valuation accuracy in cross-regional collaborative queries caused by the inability to securely integrate dynamic market data in existing technologies.

[0019] Furthermore, this application first extracts component wear indicators and performs a weighted summation based on maintenance impact weights, thereby transforming discrete maintenance details into quantifiable vehicle condition scores that accurately reflect the degree of physical damage to the vehicle. Simultaneously, it uses historical extreme value intervals to calculate the distribution ratio of current macroeconomic and supply-demand data, normalizing dynamic market indicators and effectively eliminating the dimensional inconsistency caused by differences in economic scale in cross-regional transactions.

[0020] Subsequently, the physical vehicle condition and relativized market features are mapped to a unified multi-dimensional vector coordinate system and then stitched together, constructing a standardized attribute set compatible with multi-source heterogeneous data. This feature processing method provides a unified metric for subsequent privacy-preserving computations, ensuring effective mathematical coupling between vehicle physical state and regional market heat while protecting privacy. Therefore, this application effectively solves the technical problem in existing technologies where the lack of unified data quantification and normalization standards hinders the accurate fusion of multi-dimensional features in cross-regional collaborations, thus affecting valuation accuracy. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart illustrating a collaborative query method for used car data based on privacy computing, provided as an embodiment of this application; Figure 2 A flowchart illustrating a method for generating a multidimensional attribute set provided in an embodiment of this application; Figure 3 A flowchart illustrating a method for generating logical fragments provided in an embodiment of this application; Figure 4 A schematic diagram of the structure of a used car data collaborative query system based on privacy computing provided in this application embodiment; Figure 5 This is a schematic diagram of the hardware structure of an electronic device provided in one embodiment of this application. Detailed Implementation

[0023] Existing collaborative query technologies for used car data based on privacy computing mainly rely on privacy set intersection or basic homomorphic encryption methods, focusing on Boolean verification of static historical data such as vehicle repair or accident records. They are difficult to effectively integrate macroeconomic indicators and supply and demand fluctuation characteristics of the transaction location in an encrypted state.

[0024] This lack of a dynamic market dimension makes it impossible to output accurate valuations that conform to local market conditions while protecting the data privacy of all parties in complex transaction scenarios such as cross-provincial and municipal transfers. There are technical problems such as low reference value of collaborative query results due to the rigid valuation model and lack of regional adaptability.

[0025] To address the aforementioned issues, this application proposes a collaborative query method for used car data based on privacy-preserving computation. Its core lies in constructing a multi-dimensional attribute set encompassing full lifecycle maintenance data and regional supply and demand fluctuation data, and then using distributed Lagrange interpolation secret sharing technology to transform this attribute set into logical shards. Specifically, without accessing the original data, each computing node collaboratively reconstructs attribute-related ciphertext incorporating market weight factors based on the logical shards, and completes the convolution operation between the ciphertext and the residual value model using homomorphic encryption technology.

[0026] This method abandons the single-dimensional static comparison model and ensures accurate alignment of vehicle value across regions by deeply coupling physical vehicle condition with dynamic market characteristics in the encrypted space, while strictly protecting privacy. It solves the problem of secure data flow between different locations and eliminates valuation biases caused by ignoring market fluctuations, significantly improving the accuracy and usability of cross-domain collaborative queries for used cars.

[0027] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0028] To address the problems of existing technologies, embodiments of this application provide a method, apparatus, device, computer storage medium, and computer program product for collaborative querying of used car data based on privacy computing. The method for collaborative querying of used car data based on privacy computing provided in this application embodiment will be described first below.

[0029] Figure 1 This illustration shows a flowchart of a collaborative query method for used car data based on privacy computing, provided in one embodiment of this application. Figure 1 As shown, the method includes: S101, the query node synchronously obtains vehicle maintenance data throughout the entire life cycle of the vehicle being evaluated, as well as macroeconomic indicators and supply and demand fluctuation data of used cars in the trading area.

[0030] It should be noted that the query node is configured to run in an isolated trusted execution environment during this process, and is only responsible for the standardization preprocessing and feature desensitization of the raw data, ensuring that the raw plaintext data is not visible outside the computing node cluster.

[0031] A query node refers to a computing device or server that initiates a value assessment request and has the capability to aggregate and process multi-source data. As the task initiator in the privacy computing network, it is responsible for the collection and standardized preprocessing of raw data. Vehicle maintenance data throughout the entire lifecycle of the vehicle being assessed refers to all repair and maintenance records generated for the target vehicle from its manufacture and sale to the current assessment point in time. This may include entry timestamps, repair item categories, details of replaced parts, and corresponding mileage readings.

[0032] Macroeconomic indicators for a transaction area refer to statistical values ​​that reflect the economic development level and residents' consumption capacity of the location where the target vehicle is to be traded. These can include regional GDP, per capita disposable income, or the retail price index for automobiles. Supply and demand fluctuation data for used cars refer to dynamic indicators that reflect the activity level and inventory status of a specific model in the used car market within a specific region. These can include regional inventory turnover days, average transaction cycle for similar models, or the market supply-demand ratio.

[0033] In the specific implementation process, the query node first initiates a targeted query request through the vehicle network data interface or a third-party maintenance data platform based on the unique identification code of the vehicle to be evaluated, in order to obtain the vehicle's maintenance data. ,in This indicates engine overhaul records, This indicates regular maintenance records. This indicates the sheet metal repair record in the event of an accident.

[0034] Meanwhile, the query node identifies the current transaction intention location of the vehicle through geolocation services, and simultaneously captures the region's macroeconomic indicators through regional government data disclosure interfaces and industry association data ports. and supply and demand fluctuation data .in, This represents the local GDP per capita figure. This indicates the automobile consumption index. Indicates inventory turnover rate. This indicates the level of transaction activity.

[0035] S102. The query node performs numerical mapping on vehicle maintenance data through feature vectorization to obtain vehicle condition rating. The vehicle condition rating is then mapped to a preset evaluation space along with normalized macroeconomic indicators and supply and demand fluctuation data to obtain a multi-dimensional attribute set.

[0036] Optionally, step S102, where the query node performs numerical mapping on vehicle maintenance data through feature vectorization to obtain a vehicle condition rating, and maps the vehicle condition rating with normalized macroeconomic indicators and supply and demand fluctuation data to a preset evaluation space for feature concatenation to obtain a multi-dimensional attribute set, may specifically include: Figure 2 A flowchart illustrating a method for generating a multidimensional attribute set according to an embodiment of this application is shown. Figure 2 As shown, the method includes: S1021. Extract component wear identifiers from vehicle maintenance data, and based on the component wear identifiers, obtain the corresponding target impact weights according to the preset mapping relationship between wear identifiers and maintenance impact weights.

[0037] Component wear identifiers are unique labels parsed from unstructured vehicle repair record text or codes, representing the type of damage or repair level of a specific vehicle component. These can include engine assembly replacement, A, B, and C pillar bodywork repair, or routine oil filter replacement. The pre-defined mapping between wear identifiers and repair impact weights refers to a pre-built database or lookup table that defines the quantified weights of various repair actions' negative impact on the vehicle's residual value.

[0038] This weight value is derived through a multivariate regression feature importance analysis of historical transaction big data in the industry. Specifically, a historical vehicle dataset including complete repair records and final transaction prices is selected, and a regression model such as Ridge regression or Lasso regression is constructed with repair item category as the independent variable and vehicle residual value rate as the dependent variable. The standardized regression coefficients corresponding to each repair item are calculated using the least squares method, and the absolute value of this coefficient is used as a quantitative weight value reflecting the contribution of different repair items to the depreciation of vehicle value.

[0039] The target impact weight refers to the individual deduction value or impact factor used for subsequent comprehensive score calculation, obtained by looking up a table based on the actual maintenance records of a specific vehicle. The preset mapping relationship between wear indicators and maintenance impact weights is shown in Table 1 below: Table 1: Mapping Relationship between Loss Identification and Maintenance Impact Weight

[0040] As shown in Table 1, each row represents a specific repair type, including a unique identifier code for that type, a detailed repair description, and a corresponding quantified impact weight value. This table is used to transform unstructured repair records into calculable numerical weights, where the weight values ​​are typically positively correlated with the degree of irreversible impact of the repair on vehicle performance and lifespan.

[0041] In the specific implementation process, the query node first processes the acquired vehicle maintenance data. Text parsing is performed using a pre-defined Named Entity Recognition (NER) model or a keyword tree-based regular expression matching algorithm. Specifically, a feature library is maintained containing keywords for core components such as engine and transmission, keywords for structural components such as beams and A, B, C, and C pillars, as well as keywords for common consumables. By calculating the similarity score of matching terms in the maintenance description corpus, they are automatically mapped to the corresponding wear and tear identification codes.

[0042] Next, the query node retrieves the internally stored mapping table of preset loss identifiers and maintenance impact weights, as shown in Table 1. The query node matches each extracted loss identifier with Table 1 one by one, and retrieves the corresponding value as the target impact weight.

[0043] For example, regarding the aforementioned The query node identified Engine overhaul. For regular maintenance, For the sheet metal repair of the accident, the target influence weights are mapped according to Table 1. , and The final target influence weight set for the vehicle is obtained. .

[0044] S1022. By weighted summation of the influence weights of all targets, the vehicle condition rating is obtained.

[0045] Vehicle condition rating is a standardized numerical index that reflects the quality of a vehicle's current physical condition, calculated based on its historical maintenance records. It is usually expressed as a percentage or a normalized score, with higher scores indicating better condition and vice versa.

[0046] In the specific implementation process, the query node is based on the results obtained in the previous step. Perform aggregate calculations to obtain the vehicle condition rating. Specifically, a baseline maximum score can be set. For example, 100 points, and introduce a time decay coefficient. Furthermore, the longer the repair time, the smaller the impact on the current value. The calculation formula can be expressed as follows: .in For the first The time decay coefficient of the item record.

[0047] S1023. Extract the historical extreme value range of macroeconomic indicators and supply and demand fluctuation data. By calculating the distribution ratio of the current sampled values ​​of macroeconomic indicators and supply and demand fluctuation data within the historical extreme value range, the normalized macroeconomic indicators and supply and demand fluctuation data are obtained.

[0048] Historical extreme value ranges refer to the range of maximum and minimum values ​​for macroeconomic indicators and supply and demand data determined based on statistical data from the past several years (e.g., 5 years) of the trading area, used to define the boundaries of data fluctuations. Distribution ratio refers to the relative position of the currently collected real-time data point within the aforementioned extreme value range, typically represented by a value between 0 and 1. Normalized macroeconomic indicators and supply and demand fluctuation data refer to dimensionless characteristic values ​​that have had their dimensional differences eliminated and possess a uniform value range after the aforementioned range mapping process.

[0049] In the specific implementation process, the query node first retrieves the historical extreme values ​​of various indicators for city A from the historical database. For example, regarding GDP per capita... Obtain its historical range For inventory turnover rate Obtain its interval Next, the current sampled value is processed using the max-min normalization algorithm. The calculation formula is as follows: To satisfy the subsequent finite field For integer arithmetic requirements, the query node introduces a preset large integer scaling factor. like This maps the normalized floating-point values ​​to fixed-point integers, i.e. .

[0050] Similarly, the normalized and scaled values ​​are calculated. , and For example, if the current GDP per capita At a historical high, If the inventory turnover rate is close to 1; Lower, then Approaching 0. Finally, the normalized macroeconomic vector is obtained. and supply and demand fluctuation vector .

[0051] S1024. By mapping the vehicle condition rating, normalized macroeconomic indicators, and supply and demand fluctuation data to a preset multi-dimensional vector coordinate system and performing vector splicing, a multi-dimensional attribute set is obtained.

[0052] The pre-defined multidimensional vector coordinate system refers to a unified data structure space defined to meet the dimensionality requirements of subsequent privacy-preserving computation algorithms such as Lagrange interpolation. This space specifies the arrangement order and data type of each feature component in the vector. A multidimensional attribute set refers to a single long vector formed by integrating heterogeneous data such as physical vehicle condition, macroeconomics, and market supply and demand. The pre-defined multidimensional vector coordinate system is shown in Table 2 below. Table 2: Reference Table for Preset Multidimensional Vector Coordinate Systems

[0053] As shown in Table 2, this coordinate system specifies the index position of each dimension in the multidimensional attribute set, the corresponding feature name, the data source, the normalization method, and the physical meaning represented by the dimension, ensuring that data from different sources follow a unified structural standard when splicing.

[0054] In the specific implementation process, the query node, according to the preset multi-dimensional vector coordinate system definition shown in Table 2, sequentially concatenates the components calculated in the aforementioned steps. The query node first sets the vehicle condition rating level. Normalization is performed, such as dividing by 100, to obtain Then combine it with and By combining them in the order shown in Table 2, we obtain the multidimensional attribute set. .

[0055] This embodiment accurately reflects the degree of physical damage to vehicles, effectively eliminating dimensional differences in cross-regional indicators and constructing a standardized multi-dimensional attribute set. It achieves deep fusion and alignment of multi-source heterogeneous data, significantly improving the objectivity and accuracy of collaborative assessments.

[0056] S103. The query node uses a multidimensional attribute set as the constant term of the secret shared polynomial constructed based on distributed Lagrange interpolation. By selecting multiple non-zero independent variable values ​​in a finite field and substituting them into the secret shared polynomial for evaluation, multiple independent logical partitions are generated, and the logical partitions are distributed to each computing node of the collaborative query.

[0057] Optionally, in step S103, the process of generating multiple independent logical segments by using a multi-dimensional attribute set as the constant term of a secret shared polynomial constructed based on distributed Lagrange interpolation, and selecting multiple non-zero independent variable values ​​within a finite field to substitute into the secret shared polynomial for evaluation, can specifically include: Figure 3 A flowchart illustrating a method for generating logical fragments according to an embodiment of this application is shown. Figure 3 As shown, the method includes: S1031. Determine the order of the secret shared polynomial according to the preset security threshold, and determine the multidimensional attribute set as the constant term of the secret shared polynomial.

[0058] The preset security threshold refers to the minimum number of fragments required to recover the original secret, as predefined in the secret-sharing scheme, and is usually denoted as . Threshold, among which This represents the total number of nodes participating in the collaborative computation. Minimum number of nodes required to reconstruct the secret and .

[0059] The order of the secret shared polynomial refers to the highest power of the constructed polynomial function, and its value is usually equal to the security threshold value minus one. The constant term refers to the function value when the independent variable in the polynomial is zero. In this application, it specifically refers to the protected original secret, namely the aforementioned generated multidimensional attribute set vector. The correspondence between the preset security threshold and the polynomial order is shown in Table 3 below: Table 3: Correspondence between safety threshold and polynomial order

[0060] As shown in Table 3, this table defines collaborative networks of different sizes, i.e., the number of participating nodes. Recommended security threshold Values ​​and corresponding polynomial order This configuration table is used to dynamically adjust the encryption strength based on the actual number of participants, ensuring that it can prevent a few nodes from colluding to crack the encryption while requiring at least... A maximum of [number] nodes, which can tolerate some nodes being offline. indivual.

[0061] In the actual implementation process, the query node first determines the total number of participating nodes in the current collaborative query network. like Consult the corresponding table shown in Table 3 to determine the safety threshold. This allows us to determine the order of the secret shared polynomial. .

[0062] Next, the query node will generate the multidimensional attribute set from the previous steps. Each element in the equation serves as a constant term in an independent polynomial. This means that for each of the five dimensions of the attribute set, five parallel polynomial structures will be constructed. For example, for the first element of the attribute set... Set it as the first polynomial The constant term, i.e. .

[0063] S1032. Generate a number of random values ​​equal to the order in a finite field as coefficients of higher-order terms in the secret shared polynomial, and combine each higher-order term coefficient with a constant term in ascending order of power to obtain the secret shared polynomial.

[0064] A finite field is an algebraic structure containing a finite number of elements, usually denoted as . ,in It is a large prime number. The results of addition and multiplication operations within this field remain within the field and are used to prevent numerical overflow and enhance the encryption's resistance to analysis. Random numbers refer to unpredictable values ​​generated by a true random number generator within a finite field, used to fill the coefficients of higher-order terms in the polynomial. The secret shared polynomial is a mathematical function constructed based on the Lagrange interpolation principle, with the form: .

[0065] In the specific implementation process, for each dimension, the polynomial is constructed, and the query node is in a finite field. Endogenous generation A random coefficient. For targeting polynomial For example, after ensuring that all input components have been scaled by a scaling factor Generate random numbers after converting to integers and The polynomial obtained by combining the coefficients of the linear and quadratic terms is: .in, It is a constant term. and These are random higher-order coefficients. It is the independent variable. It is a pre-defined large prime number. Similarly, for the attribute set... Build This process continues until a total of 5 polynomials are constructed. This set of polynomials is called the secret shared polynomial set.

[0066] S1033. Select multiple non-repeating coordinate values ​​of independent variables within a finite domain as non-zero independent variable values. Substitute these non-zero independent variable values ​​into a secret shared polynomial for exponential product and summation to obtain the corresponding evaluation results.

[0067] A non-zero independent variable value refers to a unique identification value (ID) assigned to each participating computing node, typically 0. And represent the graph of a polynomial function Coordinates. The summation of power products refers to adding selected coordinates... Substitute the values ​​into the polynomial formula to calculate each term. The process of multiplying and summing the products is called calculation. The function value.

[0068] In the specific implementation process, the query node is... Each participating node is assigned a non-zero independent variable value. Next, each Substitute these values ​​into the five polynomials constructed above for calculation. Taking node 1 as an example... For example, the calculation targets Fragment value: Similarly, calculate the sharding value of this node for other dimensions. Finally, the set of evaluation results obtained for node 1 is... And so on, calculating the set of results for all 5 nodes. to .

[0069] S1034. Logical slices are obtained by associating non-zero independent variable values ​​with their corresponding evaluation results as data units.

[0070] Data association refers to linking the identity identifiers of nodes. Coordinates and the calculated secret share Value sets are bound together to form a complete key-value pair structure. A data unit is the smallest transmission packet that includes index information and data payload. A logical fragment is a secret shared data packet that is finally generated and ready to be distributed to various computing nodes. Each fragment includes all the local information required by that node to reconstruct the polynomial, and no single fragment can reveal information about the original multidimensional attribute set.

[0071] In the actual implementation process, the query node will retrieve the data from each node. and the corresponding set of evaluation results Package the data. For example, generate a logical shard to send to node 1. Logical shards sent to node 2 And so on. These logical partitions to This is the final product.

[0072] Subsequently, the query node distributes the logical shards to the compute nodes through an encrypted channel, at which point the original multidimensional attribute set... The data has been disassembled and distributed, and the query nodes immediately physically destroy the original plaintext data and the spliced ​​attribute set in their memory after completing the distribution, thus cutting off the data leakage link at the physical level.

[0073] This embodiment transforms a sensitive multidimensional attribute set into a mathematically irreversible distributed logical partition by introducing a security threshold mechanism and randomized polynomial coefficients. This approach physically severs the plaintext data link between the data holder and the computing node, and eliminates the risk of single-point data leakage by utilizing the discrete characteristics of finite fields.

[0074] S104. Each computing node uses the received logical fragments and preset market weight factors to obtain the attribute association ciphertext of the vehicle in the transaction area by distributively solving and reorganizing the coefficients of the secret shared polynomial.

[0075] Optionally, step S104, in which each computing node uses the received logical fragments and preset market weight factors to obtain the attribute association ciphertext for evaluating the vehicle in the transaction area by distributively solving and reorganizing the coefficients of the secret shared polynomial, may specifically include: S1041. Each computing node extracts the evaluation result from the received logical fragment and obtains the weighted fragment value based on the product of the evaluation result and the market weight factor.

[0076] The market weighting factor is a coefficient vector used to adjust for regional sensitivity at the secret sharing level. Its value is determined based on the Pearson correlation coefficient between various attribute dimensions within the trading area and historical transaction price fluctuations. Specifically, it collects the macroeconomic indicator change series and the average used car transaction price series within a specific time window (e.g., 36 months) of the region, calculates the ratio of their covariance to standard deviation, and uses the obtained normalized correlation value as a weight value reflecting the local market's sensitivity to differences in vehicle condition, brand, or economic environment.

[0077] It needs to be clarified that the market weighting factor The focus is on adjusting the different weightings of vehicle condition and economic indicators in different transaction regions, such as first-tier cities and third-tier cities, while the subsequent residual value weighting matrix... This focuses on characterizing the general value loss pattern of a vehicle as its age and attributes change. Through multiplicative coupling, the two achieve dual dynamic correction of industry benchmarks and regional characteristics.

[0078] The weighted shard value is the intermediate value obtained by multiplying the original secret shard held by the computing node by its corresponding market weight. This value retains the mathematical characteristics of secret shards while incorporating a market adjustment mechanism. The preset market weight factors are shown in Table 4 below: Table 4: Preset Market Weight Factor Comparison Table

[0079] As shown in Table 4, this table defines the weight adjustment coefficients for different attribute dimensions. For example, in some economically developed regions, buyers may place more emphasis on the physical condition of the vehicle and be less sensitive to price fluctuations; while in some price-sensitive regions, the influence weight of supply and demand data will be adjusted accordingly.

[0080] In the specific implementation process, taking node 1 as an example, the computing node first parses the received logical fragments. Extract the evaluation result set for the five dimensions. Next, node 1 calls the preset market weight factor table shown in Table 4 to obtain the corresponding weight vector. .

[0081] Then, node 1 performs element-wise multiplication to calculate the weighted slice value. The specific formula is as follows: For example, for the first dimension, Ultimately, node 1 obtains the weighted sharding value set. The other nodes are calculated similarly. .

[0082] S1042. Each computing node exchanges the non-zero independent variable values ​​in each logical segment with each other and calculates the ratio of the difference between the non-zero independent variable values ​​and other non-zero independent variable values ​​to obtain the corresponding interpolation basis operator. The interpolation basis operator is used to map and transform the weighted segment values ​​to obtain the corresponding attribute intermediate component.

[0083] Interpolation basis operators are those used in the Lagrange interpolation formula to interpolate based on the values ​​of each node. The basis function coefficients of the constant term of the original function are reconstructed using coordinate reconstruction and are usually denoted as... Its computation depends on all nodes participating in the reconstruction. Coordinate values. The intermediate component of the attribute refers to the value obtained by multiplying the weighted piecewise value with the interpolation basis operator. This value is an additive component of the final secret of reconstruction, namely the constant term.

[0084] In the specific implementation process, assume that the set of nodes participating in the reconstruction is as follows: That is, the threshold is met The smallest set. Each node broadcasts its non-zero independent variable values ​​to each other. That is, the node ID. Taking node 1 as an example, it obtains... Next, node 1 calculates its own interpolation basis operator according to the Lagrange interpolation formula. The specific calculation process is shown in the following formula (1): (1) Finally, node 1 will obtain the operator Compared with the weighted sharding value set obtained in the previous step Perform the calculations to obtain the set of intermediate attribute components. That is, for each dimension ,calculate .

[0085] S1043. Each computing node uses a secure communication link to aggregate and reassemble the intermediate components of the attributes in an encrypted state to obtain the attribute-associated ciphertext.

[0086] Attribute-related ciphertext refers to the final result obtained by summing the intermediate components of the attributes of all participating nodes, and this result is mathematically equivalent to the product of the original multidimensional attribute set and the market weight factor. However, the original attribute set was never fully restored during the calculation process, thus achieving fusion computing under privacy protection.

[0087] In this step, all computing nodes elect a temporary leader node through a distributed consensus protocol. Each follower node sends its computed encrypted components to the leader node, which then performs aggregation and reassembly in the ciphertext space. This division of roles and communication mechanism among nodes ensures the orderly execution of computing tasks within a decentralized architecture.

[0088] This embodiment achieves dense fusion of multidimensional attribute features and market weights, which not only ensures the privacy and security of vehicle data, but also ensures that the final valuation model can flexibly and dynamically adapt to the market environment of different trading areas.

[0089] Optionally, step S1043, in which each computing node aggregates the intermediate components of the attributes and performs encrypted summation and recombination using a secure communication link to obtain the attribute-associated ciphertext, may specifically include: S10431. Each computing node sends attribute intermediate components to other computing nodes through a secure communication link and receives attribute intermediate components sent by other computing nodes, thus obtaining an attribute component set that includes all attribute intermediate components.

[0090] A secure communication link refers to a point-to-point data transmission channel established using encryption protocols such as SSL / TLS to ensure that data is not eavesdropped on or tampered with during transmission. An attribute component set refers to the set of attributes collected by a master node or aggregate node from all... A list of intermediate components of the attributes of each participating node.

[0091] In the specific implementation process, assuming node 1 acts as a temporary aggregator, then nodes 2 and 3 will transmit their respective calculated values ​​through a secure communication link. and Send to node 1. After receiving it, node 1 constructs a set of attribute components. Each element is a vector with 5 dimensions.

[0092] S10432. In the encrypted state, each computing node sums up all intermediate components of the attribute components in the attribute component set to obtain the recombined attribute value.

[0093] The recombined attribute value refers to the secret value, or constant term, recovered through vector addition. In this scenario, due to the introduction of a market weight factor, the recovered value is no longer the original attribute. Instead, it is the weighted attribute. Summation in the encrypted state means that the operation takes place within a finite field. The modular addition operation is performed internally and does not involve the direct exposure of plaintext values.

[0094] In the specific implementation process, node 1, acting as the aggregator, utilizes the additive homomorphic property to aggregate the set without knowing the plaintext of each component. The vectors in the array are subjected to homomorphic modular addition operations on their corresponding dimensions. For the first... Each dimension, calculation .

[0095] In the specific implementation process, node 1, acting as the aggregator, utilizes the additive homomorphic property to aggregate the set without knowing the plaintext of each component. Perform homomorphic modular addition on the vectors in the corresponding dimension. 。 For the Each dimension, calculation The square brackets denote ciphertext summation within the additive homomorphic space, ultimately yielding the recombined attribute value vector. .

[0096] S10433. Each computing node encapsulates the recombined attribute values ​​to obtain attribute-related ciphertext.

[0097] Data encapsulation refers to adding metadata headers, checksums, and homomorphic encryption context information required for subsequent processing to the calculated raw data vector to form a standardized ciphertext object. In the specific implementation process, Node 1 will calculate the... Perform formatted encapsulation to generate attribute-associative ciphertext. The encrypted text essentially includes a comprehensive profile of the vehicle's characteristics in the current market environment, such as: physical score × 1.5, GDP score × 1.0, etc., but it exists in the form of finite field values, and its physical meaning cannot be directly interpreted externally. .

[0098] This embodiment achieves secure aggregation of distributed computing results without exposing the original data fragments. By reorganizing the data encapsulation of attribute values, a standardized attribute-related ciphertext, including market weight features, is constructed.

[0099] S105. Each computing node uses homomorphic encryption to perform matrix convolution operation on the attribute association ciphertext and the preset residual weight matrix to obtain the valuation ciphertext for collaborative query. All nodes use the consensus decryption operator to desensitize and restore the valuation ciphertext to obtain the collaborative query valuation of the evaluated vehicle.

[0100] Optionally, the process in step S105 where each computing node performs matrix convolution operations on the attribute association ciphertext and the preset residual weight matrix using homomorphic encryption to obtain the estimated ciphertext of the collaborative query can specifically include: S1051. Map the intermediate components of the attributes in the attribute association ciphertext to the preset multidimensional ciphertext coordinate system to obtain a ciphertext matrix with the same dimension as the residual weight matrix.

[0101] A multidimensional ciphertext coordinate system refers to a data structure mapping rule defined to accommodate matrix operations. It transforms linear attribute-associated ciphertext vectors into a two-dimensional matrix with a specific row and column structure to align with industry-standard residual prediction models. A ciphertext matrix is ​​a two-dimensional array whose elements are all encrypted numerical values, generated after coordinate mapping, and its structure is typically... or In the form of.

[0102] The residual weight matrix refers to a pre-defined coefficient matrix, which is constructed by collecting a sample set that includes historical vehicle attribute features, such as vehicle condition ratings, macroeconomic indicators, supply and demand data, and actual transaction prices. Construct the mean squared error loss function ,in The regularization parameter is used to iteratively optimize the loss function using the stochastic gradient descent (SGD) algorithm until the loss value converges, thereby obtaining the optimal coefficient matrix that embodies the valuation logic of the used car industry, such as the vehicle depreciation curve and the configuration premium rate. The preset multidimensional encrypted coordinate system is shown in Table 5 below: Table 5: Reference Table of Preset Multidimensional Ciphertext Coordinate Systems

[0103] As shown in Table 5, this table specifies the attribute-associated ciphertext vector. To ciphertext matrix The mapping logic involves assigning feature values ​​from different dimensions, such as physical vehicle condition and macroeconomics, to specific rows and columns of a matrix to correspond to modules like the basic vehicle condition factor and market adjustment factor in the residual value model.

[0104] In the specific implementation process, taking the master node as an example, the compute node first parses the attribute association ciphertext generated in the previous step. Extract the recombined attribute value vector. Next, based on the mapping rules in Table 5, a ciphertext matrix is ​​constructed. .because Data that has been processed within a homomorphic encryption or secret sharing domain is considered here as a... Row vector matrix: Simultaneously, a preset residual weight matrix is ​​loaded. The dimension of this matrix is ​​also 1. Its elements This represents the contribution coefficient of each feature to the final valuation.

[0105] S1052. Each computing node calculates the product of the values ​​of the intermediate component of each attribute in the ciphertext matrix and the corresponding weight component in the residual weight matrix within the encrypted space to obtain the encrypted weight value.

[0106] The encrypted space refers to a mathematical environment that supports homomorphic operations such as addition homomorphism or fully homomorphism, where algebraic operations performed on ciphertext within this environment are equivalent to corresponding operations performed on plaintext. The encrypted weight value is the intermediate ciphertext result obtained by multiplying the elements of the ciphertext matrix by the corresponding coefficients in the residual weight matrix, representing the specific monetary contribution of each feature dimension to the final valuation. The matrix convolution operation in this application refers to the linear weighted aggregation of multidimensional features and the residual model within the encrypted space through a sliding window weight mapping mechanism.

[0107] In practical implementation, the computing nodes utilize the scalar multiplication property of homomorphic encryption algorithms to... and Perform element-wise multiplication. Specifically, for the ... At each location, the computing node uses a homomorphic encryption public key to access the intermediate component of the attribute. Encryption is performed, and the scalar multiplication property of homomorphic encryption is used to combine it with the corresponding residual weight. Perform the calculation. The calculation formula is: Note that the finite field form must be encrypted before encryption. This is mapped to a plaintext space supported by homomorphic encryption algorithms, such as an integer ring. The final result is a vector of encryption weight values. .

[0108] S1053. Each computing node sums up all encrypted weight values ​​to obtain the estimated ciphertext.

[0109] The valuation ciphertext refers to the single ciphertext value obtained by aggregating the encryption weight values ​​of all dimensions, and its decrypted plaintext is the final valuation of the vehicle. In the specific implementation process, the computation node pairs vectors... Perform homomorphic addition on all elements to obtain the ciphertext of the collaborative query's valuation. The specific calculation process is shown in formula (2) below: (2) This embodiment utilizes a multidimensional encrypted coordinate system to standardize and map dispersed attribute features into an encrypted matrix, achieving structural alignment between heterogeneous data and industry valuation models. This ensures both the absolute confidentiality of vehicle physical and market privacy information and the mathematical accuracy and immutability of the valuation results.

[0110] Optionally, in step S105, all nodes include query nodes and computation nodes. The process by which all nodes in step S105 de-identify and restore the valuation ciphertext using a consensus decryption operator to obtain the collaborative query valuation of the evaluated vehicle can specifically include: S1054. All nodes use the corresponding preset decryption components to perform decryption operations on the estimated ciphertext to obtain the corresponding decryption shared components.

[0111] The preset decryption component refers to the private key fragment distributed to each participating node in a threshold decryption system. No single node can decrypt the ciphertext using its own private key fragment; only by accumulating a sufficient number of decryption results can the plaintext be restored. The shared decryption component refers to the intermediate data fragment generated after performing a partial decryption operation on the estimated ciphertext using the decryption component of a single node. The preset decryption components are shown in Table 6 below: Table 6: Preset Decryption Components Comparison Table

[0112] As shown in Table 6, this table records the participating node IDs and the private key shards they hold. The correspondence and weighting coefficients during the decryption process are defined. In practice, to ensure consistency within the collaborative system, this embodiment employs the same security threshold as the secret sharing phase. Assuming we adopt Threshold decryption scheme. Each participating node Use your own private key to shard. The final valuation cipher Perform partial decryption calculations. Taking the Paillier threshold as an example, the calculation formula is usually as follows: Calculated This is the decryption shared component of that node.

[0113] S1055. All nodes use the secure communication link to aggregate each decrypted shared component, resulting in a shared component set that includes all decrypted shared components.

[0114] A shared component set refers to a list of intermediate decryption results collected from different nodes. In practice, each node transmits its calculated results via a secure link. Broadcast to the query node or the specified synthesizer node. The query node collects a sufficient number of data points, i.e., at least... After determining the individual components, construct the set. .

[0115] S1056. All nodes use the decryption operator to reorganize the shared component set to obtain the collaborative query estimate of the evaluated vehicle.

[0116] A decryption operator is a mathematical algorithm used to combine multiple decryption results to reconstruct the original plaintext, typically involving operations on Lagrange interpolation coefficients in the exponent domain. Collaborative query valuation refers to the final reconstructed vehicle price in plaintext form. In practice, the query node uses the Lagrange interpolation formula to perform operations on the set in the exponent domain. The components are synthesized, that is... Obtain plaintext values This value represents the collaborative query estimate for evaluating the vehicle.

[0117] This embodiment ensures that no single node can independently restore the ciphertext, eliminating the risk of single-point data leakage and achieving plaintext restoration under multi-party consensus. It guarantees the authority and immutability of the final valuation result while also establishing a robust access control defense during the decryption process.

[0118] Figure 4 This is a schematic diagram illustrating a specific implementation of a privacy-based collaborative query system for used car data, as provided in this application. (Refer to...) Figure 4 The system may include: The acquisition module 410 is used to query the node to synchronously acquire vehicle maintenance data throughout the entire life cycle of the vehicle being evaluated, as well as macroeconomic indicators and supply and demand fluctuation data of used cars in the trading area. The mapping module 420 is used to query nodes to perform numerical mapping on vehicle maintenance data through feature vectorization to obtain vehicle condition rating levels. The vehicle condition rating levels are then mapped to a preset evaluation space with normalized macroeconomic indicators and supply and demand fluctuation data for feature splicing to obtain a multi-dimensional attribute set. The evaluation module 430 is used to query the constant term of the secret shared polynomial constructed by the node using a multidimensional attribute set as a distributed Lagrange interpolation. By selecting multiple non-zero independent variable values ​​in a finite field and substituting them into the secret shared polynomial for evaluation, multiple independent logical fragments are generated and distributed to each computing node of the collaborative query. The generation module 440 is used by each computing node to obtain the attribute association ciphertext of the vehicle in the transaction area by distributively solving and reorganizing the coefficients of the secret shared polynomial using the received logical fragments and preset market weight factors. The generation module 440 is also used by each computing node to perform matrix convolution operation on the attribute association ciphertext and the preset residual weight matrix using homomorphic encryption to obtain the valuation ciphertext for collaborative query. All nodes use the consensus decryption operator to desensitize and restore the valuation ciphertext to obtain the collaborative query valuation of the evaluated vehicle.

[0119] The privacy-based computation-based collaborative query system for used car data in this application is used to implement the aforementioned privacy-based computation-based collaborative query method for used car data. Therefore, the specific implementation of the privacy-based computation-based collaborative query system for used car data can be found in the embodiment section of the privacy-based computation-based collaborative query method for used car data above. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0120] Figure 5 A schematic diagram of the hardware structure of an electronic device provided in one embodiment of this application is shown.

[0121] The electronic device may include a processor 510 and a memory 520 storing computer program instructions.

[0122] Specifically, the processor 510 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0123] Memory 520 may include mass storage for data or instructions. For example, and not limitingly, memory 520 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 520 may include removable or non-removable (or fixed) media. Where appropriate, memory 520 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 520 is non-volatile solid-state memory.

[0124] Memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to the first aspect of this disclosure.

[0125] The processor 510 reads and executes computer program instructions stored in the memory 520 to implement any of the privacy-based computing collaborative query methods for used car data in the above embodiments.

[0126] In one example, the electronic device may also include a communication interface 530 and a bus 540. Wherein, such as Figure 5 As shown, the processor 510, memory 520, and communication interface 530 are connected through bus 540 and complete communication with each other.

[0127] The communication interface 530 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0128] Bus 540 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 540 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0129] The electronic device can execute the privacy-based computation-based collaborative query method for used car data in the embodiments of this application, thereby realizing the privacy-based computation-based collaborative query method for used car data described in conjunction with the accompanying drawings.

[0130] Furthermore, in conjunction with the privacy-based computation-based collaborative query method for used car data in the above embodiments, this application embodiment can provide a computer-readable storage medium for implementation. This computer-readable storage medium stores computer program instructions; when executed by a processor, these computer program instructions implement any of the privacy-based computation-based collaborative query methods for used car data in the above embodiments.

[0131] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0132] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0133] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0134] The aspects of this application have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by dedicated hardware performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0135] The foregoing has provided a detailed description of a collaborative query method and system for used car data based on privacy computing, as provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for privacy computing-based used car data collaborative query, characterized in that, include: The query node synchronously obtains vehicle maintenance data throughout the entire life cycle of the vehicle, as well as macroeconomic indicators and supply and demand fluctuation data of used cars in the transaction area; The query node performs numerical mapping on the vehicle maintenance data through feature vectorization to obtain a vehicle condition rating. The vehicle condition rating is then mapped to a preset evaluation space along with the normalized macroeconomic indicators and the supply and demand fluctuation data to obtain a multi-dimensional attribute set. The query node uses the multidimensional attribute set as the constant term of the secret shared polynomial constructed based on distributed Lagrange interpolation. By selecting multiple non-zero independent variable values ​​in a finite field and substituting them into the secret shared polynomial for evaluation, multiple independent logical partitions are generated, and the logical partitions are distributed to each computing node of the collaborative query. Each computing node uses the received logical fragments and preset market weight factors to obtain the attribute association ciphertext of the vehicle in the transaction area by distributively solving and reorganizing the coefficients of the secret shared polynomial. Each computing node uses homomorphic encryption to perform matrix convolution operation on the attribute association ciphertext and the preset residual weight matrix to obtain the valuation ciphertext for collaborative query. All nodes use a consensus decryption operator to de-identify and restore the valuation ciphertext to obtain the collaborative query valuation of the evaluated vehicle. Each computing node, using the received logical fragments and preset market weight factors, performs distributed solving and recombination of the coefficients of the secret shared polynomial to obtain the attribute association ciphertext for evaluating the vehicle's position in the trading area, including: Each computing node extracts the evaluation result from the received logical shard and obtains the weighted shard value based on the product of the evaluation result and the market weight factor. The evaluation result is obtained by substituting multiple non-repeating independent variable coordinate values ​​selected in the finite domain into the secret shared polynomial for exponential product and summation. Each computing node exchanges the non-zero independent variable values ​​in each logical segment with each other and calculates the difference ratio between the non-zero independent variable values ​​and other non-zero independent variable values ​​to obtain the corresponding interpolation basis operator. The interpolation basis operator is then used to map and transform the weighted segment values ​​to obtain the corresponding attribute intermediate component. Each computing node uses a secure communication link to aggregate and reassemble the intermediate components of the attribute in an encrypted state to obtain the attribute-associated ciphertext. Each computing node uses homomorphic encryption to perform matrix convolution operations on the attribute association ciphertext and a preset residual weight matrix to obtain the estimated ciphertext for collaborative queries, including: The intermediate component of the attribute in the attribute-associated ciphertext is mapped to a preset multidimensional ciphertext coordinate system to obtain a ciphertext matrix with the same dimension as the residual weight matrix; Each computing node calculates the product of the values ​​of the intermediate component of each attribute in the ciphertext matrix and the corresponding weight component in the residual weight matrix within the encrypted space to obtain the encrypted weight value. Each computing node sums up all the encryption weight values ​​to obtain the estimated ciphertext.

2. The method of claim 1, wherein, The query node performs numerical mapping on the vehicle maintenance data through feature vectorization to obtain a vehicle condition rating. The vehicle condition rating is then mapped to a preset evaluation space along with normalized macroeconomic indicators and supply and demand fluctuation data for feature concatenation, resulting in a multi-dimensional attribute set, including: Extract component wear identifiers from the vehicle maintenance data, and based on the component wear identifiers, obtain the corresponding target impact weight according to the preset mapping relationship between wear identifiers and maintenance impact weights; The vehicle condition rating is obtained by weighted summation of the influence weights of all the aforementioned targets; Extract the historical extreme value ranges of the macroeconomic indicators and the supply and demand fluctuation data, and obtain the normalized macroeconomic indicators and supply and demand fluctuation data by calculating the distribution ratio of the current sampled values ​​of the macroeconomic indicators and the supply and demand fluctuation data within the historical extreme value ranges; The multidimensional attribute set is obtained by mapping the vehicle condition rating, the normalized macroeconomic indicators, and the supply and demand fluctuation data to a preset multidimensional vector coordinate system and then concatenating the vectors.

3. The method of claim 1, wherein, The query node uses the multidimensional attribute set as the constant term of a secret shared polynomial constructed based on distributed Lagrange interpolation. By selecting multiple non-zero independent variable values ​​within a finite field and substituting them into the secret shared polynomial for evaluation, multiple independent logical slices are generated, including: The order of the secret shared polynomial is determined according to a preset security threshold, and the multidimensional attribute set is determined as the constant term of the secret shared polynomial. Within the finite field, a number of random values ​​equal to the order are generated as coefficients of higher-order terms in the secret shared polynomial. Each coefficient of higher-order terms is then combined with the constant term in ascending order of power to obtain the secret shared polynomial. Within the finite domain, select multiple non-repeating coordinate values ​​of the independent variables as the non-zero independent variable values, substitute the non-zero independent variable values ​​into the secret shared polynomial, perform exponential product and summation to obtain the corresponding evaluation result; The logical slice is obtained by associating the non-zero independent variable values ​​with the corresponding evaluation results as data units.

4. The method of claim 1, wherein, Each computing node aggregates and reassembles the intermediate components of the attribute using a secure communication link, and then performs encrypted summation and recombination to obtain the attribute-associated ciphertext, including: Each computing node sends the attribute intermediate component to other computing nodes through the secure communication link and receives the attribute intermediate component sent by other computing nodes, thereby obtaining an attribute component set including all attribute intermediate components; Each computing node, in the encrypted state, sums up all intermediate components of the attribute components in the attribute component set to obtain the recombined attribute value; Each computing node encapsulates the recombined attribute value to obtain the attribute-associated ciphertext.

5. The method of claim 1, wherein, All nodes include the query node and each of the computing nodes; All nodes use the consensus decryption operator to de-identify and restore the ciphertext of the valuation, obtaining the collaborative query valuation of the evaluated vehicle, including: All nodes use the corresponding preset decryption component to perform decryption operation on the estimated ciphertext to obtain the corresponding decryption shared component; All nodes use a secure communication link to aggregate each of the decrypted shared components, resulting in a shared component set that includes all the decrypted shared components; All nodes use the decryption operator to reassemble the shared component set to obtain the collaborative query estimate of the evaluated vehicle.

6. A collaborative query system for used car data based on privacy computing, characterized in that, include: The acquisition module is used to query nodes to synchronously acquire vehicle maintenance data throughout the entire life cycle of the vehicle being evaluated, as well as macroeconomic indicators and supply and demand fluctuation data of used cars in the trading area. The mapping module is used by the query node to perform numerical mapping on the vehicle maintenance data through feature vectorization to obtain the vehicle condition rating level, and to map the vehicle condition rating level with the normalized macroeconomic indicators and the supply and demand fluctuation data to a preset evaluation space for feature splicing to obtain a multi-dimensional attribute set. The evaluation module is used by the query node to take the multidimensional attribute set as the constant term of the secret shared polynomial constructed based on distributed Lagrange interpolation, and to perform evaluation operations by selecting multiple non-zero independent variable values ​​in a finite field and substituting them into the secret shared polynomial to generate multiple independent logical partitions, and then distributing the logical partitions to each computing node of the collaborative query. The generation module is used by each computing node to obtain the attribute association ciphertext of the vehicle in the transaction area by distributively solving and reorganizing the coefficients of the secret shared polynomial using the received logical fragments and preset market weight factors. The generation module is specifically used by each computing node to extract the evaluation results from the received logical slices, and obtain a weighted slice value based on the product of the evaluation results and the market weight factor. The evaluation results are obtained by substituting multiple non-repeating independent variable coordinate values ​​selected in the finite domain into the secret shared polynomial for power product and summation. Each computing node exchanges the non-zero independent variable values ​​in each logical slice and calculates the difference ratio between the non-zero independent variable values ​​and other non-zero independent variable values ​​to obtain the corresponding interpolation basis operator. The interpolation basis operator is used to map and transform the weighted slice value to obtain the corresponding attribute intermediate component. Each computing node uses a secure communication link to aggregate and reassemble the attribute intermediate components in an encrypted state to obtain the attribute associated ciphertext. The generation module is also used to perform matrix convolution operation on the attribute association ciphertext and the preset residual weight matrix by each computing node using homomorphic encryption to obtain the valuation ciphertext of the collaborative query. All nodes desensitize and restore the valuation ciphertext through the consensus decryption operator to obtain the collaborative query valuation of the evaluated vehicle. The generation module is also specifically used to map the intermediate component of the attribute in the attribute-associated ciphertext to a preset multidimensional ciphertext coordinate system to obtain a ciphertext matrix with the same dimension as the residual weight matrix; Each computing node calculates the product of the values ​​of the intermediate component of each attribute in the ciphertext matrix and the corresponding weight component in the residual weight matrix within the encrypted space to obtain the encrypted weight value. Each computing node sums up all the encryption weight values ​​to obtain the estimated ciphertext.

7. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the collaborative query method for used car data based on privacy computing as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the collaborative query method for used car data based on privacy computing as described in any one of claims 1 to 5.

Images