A second-hand car cross-domain inquiry data sharing method and system based on differential privacy
By dynamically adjusting the disturbance noise intensity of numerical vehicle condition parameters in cross-domain used car price inquiries, the issues of vehicle privacy leakage and data availability are resolved, achieving reasonable consumption of privacy budget and effective protection of price inquiry data.
Patent Information
- Application Number
- CN202611461721.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-20
- Publication Date
- 2026-11-06
AI Technical Summary
In the process of cross-domain price inquiry for used cars, numerical vehicle condition parameters are easily restored by the recipient after being repeatedly published. Furthermore, existing differential privacy processing fails to effectively distinguish the price sensitivity of vehicle condition parameters, resulting in inappropriate consumption of privacy budget and a decline in data availability.
By associating the target vehicle, receiving platform, and privacy budget cycle, the perturbation noise intensity of each numerical vehicle condition parameter is dynamically adjusted, random noise is generated and superimposed on the vehicle condition parameters, and cross-domain price inquiries are performed in conjunction with categorical vehicle condition parameters to ensure reasonable consumption of the privacy budget and data availability.
It effectively protects vehicle privacy during cross-domain price inquiries, avoids leakage of vehicle identification information, and ensures the availability of price inquiry data and reasonable consumption of privacy budget.
Smart Images
Figure CN123048856A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of used car transaction data processing and privacy protection technology, and in particular to a method and system for cross-domain price inquiry data sharing of used cars based on differential privacy. Background Technology
[0002] In cross-regional used car price inquiries, the primary trading platform needs to share vehicle condition data with the secondary trading platform in the target region to obtain a matching price. Numerical vehicle condition parameters such as vehicle age, mileage, number of ownership transfers, and number of repairs directly affect the vehicle's valuation and are considered sensitive data. If transmitted in plaintext or repeatedly published, the recipient may be able to reconstruct the true values through correlation and comparison, leading to privacy leaks.
[0003] Existing solutions often employ encrypted transmission, desensitization, or differential privacy perturbation, but typically set the same or fixed privacy budget and noise intensity for each numerical vehicle condition parameter. They do not correlate the historical query frequency of the same vehicle, the same receiving platform, or the same field within the privacy budget period, nor do they distinguish the differences in the impact of each parameter on the transaction price.
[0004] When the same parameter is repeatedly published, the receiver can reduce noise and gradually restore the true value by averaging and comparing the results of multiple times; the uniform noise is not differentiated according to price sensitivity, which can easily lead to a decrease in the availability of highly sensitive parameters or insufficient protection of low-sensitive parameters; the privacy budget is not associated with the number of historical publications, receiving platforms and data fields, which can easily lead to overspending or duplicate consumption; cross-domain transmission may also carry vehicle identification numbers, user identity information or overly detailed vehicle models, which may pose a risk of identity association. Summary of the Invention
[0005] This application provides a method and system for sharing cross-domain used car price inquiry data based on differential privacy, which solves the problems that numerical vehicle condition parameters are easily restored by the recipient after being repeatedly published during the cross-domain used car price inquiry process, and the impact on the availability of price inquiry data when different vehicle condition parameters are processed with the same noise.
[0006] Firstly, this application provides a method for cross-domain price inquiry data sharing of used cars based on differential privacy, including: Obtain cross-domain price inquiry requests initiated by users for the target vehicle in the first trading platform, and parse the cross-domain price inquiry requests to obtain the target vehicle's category-type vehicle condition parameters, numerical vehicle condition parameters, and target price inquiry area identifier; The second trading platform for accepting inquiries is determined based on the target inquiry area identifier, and the historical query frequency of the numerical vehicle condition parameters within the first trading platform is extracted. The privacy budget consumption of the target vehicle is determined based on the historical query frequency, and the privacy budget consumption is subtracted from the pre-configured initial privacy budget to obtain the current available privacy budget. Extract the price sensitivity weights corresponding to the numerical vehicle condition parameters, and set the disturbance noise value ranges corresponding to each of the numerical vehicle condition parameters according to the correlation between the current available privacy budget and the price sensitivity weights. Random noise values are generated within their respective disturbance noise value ranges, and each random noise value is superimposed onto each numerical vehicle condition parameter to generate disturbed numerical parameters. The categorical vehicle condition parameters and the disturbed numerical parameters are sent together to the second trading platform for vehicle price matching, to obtain cross-domain price inquiry feedback results, and then the cross-domain price inquiry feedback results are sent to the user through the first trading platform.
[0007] Optionally, the system obtains cross-domain price inquiry requests initiated by users for the target vehicle on the first trading platform, and parses the cross-domain price inquiry requests to obtain the target vehicle's categorical vehicle condition parameters, numerical vehicle condition parameters, and target price inquiry region identifier, including: Receive a cross-domain price inquiry request message sent by a user client. The cross-domain price inquiry request message includes the vehicle identification number, target price inquiry area code, vehicle brand and series, model version, energy type, damage level, vehicle age, mileage, number of ownership transfers, and number of repairs. The vehicle brand and series, the vehicle model version, the energy type, and the damage level are defined as category-type vehicle condition parameters; The vehicle age, mileage, number of ownership transfers, and number of repairs are defined as numerical vehicle condition parameters. The target inquiry area code is determined as the target inquiry area identifier.
[0008] Optionally, a second trading platform for accepting inquiries is determined based on the target inquiry area identifier, and the historical query frequency of the numerical vehicle condition parameter within the first trading platform is extracted, including: Read the regional service registration form that records the transaction platform identifier, service area code, and current response time; Select the trading platform with the same service area code as the target inquiry area identifier and the shortest current response time from the regional service registration table as the second trading platform; A release tag is generated based on the vehicle identification number, the transaction platform identifier of the second transaction platform, the field name of the numerical vehicle condition parameter, and the predefined privacy budget period; Retrieve from the privacy release log of the first trading platform the release record that has the release tag and has output the disturbed numerical parameter to the second trading platform; The number of published records is counted, and the statistical result is determined as the historical query frequency.
[0009] Optionally, the privacy budget consumption of the target vehicle is determined based on the historical query frequency, and the privacy budget consumption is subtracted from the pre-configured initial privacy budget to obtain the currently available privacy budget, including: Read the initial privacy budget of the target vehicle during the privacy budget period; Based on the historical query frequency, retrieve the corresponding number of publishing records from the privacy publishing log, and read the single publishing budget recorded in each publishing record; The privacy budget consumption of the target vehicle during the privacy budget period is obtained by summing up the read single release budget. The current available privacy budget is obtained by subtracting the privacy budget consumption from the initial privacy budget. When the current available privacy budget is zero, stop generating disturbed numerical parameters and return a prompt message to the user indicating that cross-domain price inquiries will be stopped for this privacy budget period.
[0010] Optionally, the price sensitivity weights corresponding to the numerical vehicle condition parameters are extracted, and based on the correlation between the currently available privacy budget and the price sensitivity weights, disturbance noise value ranges corresponding to each of the numerical vehicle condition parameters are set, including: Read the regional transaction statistics summary provided by the second trading platform and stored in the first trading platform before cross-domain inquiry. The regional transaction statistics summary records the allowed value range of each numerical vehicle condition parameter under different types of vehicle condition parameters and the transaction price change between adjacent value segments. The price sensitivity weight of each numerical vehicle condition parameter is determined based on the proportion of the change in transaction price corresponding to each numerical vehicle condition parameter in the total change in transaction price. The currently available privacy budget is allocated as a field privacy budget according to the price sensitivity weight; The sum of the privacy budgets for all fields is determined as the single release budget, and the single release budget does not exceed the currently available privacy budget; The difference between the two ends of the allowed range of values for each numerical vehicle condition parameter is determined as the field's privacy sensitivity. Based on the ratio of the field privacy sensitivity to the field privacy budget, the noise change scale is determined, and the range of all real numbers centered at zero is determined as the disturbance noise range for the corresponding numerical vehicle condition parameter.
[0011] Optionally, random noise values are generated within their respective disturbance noise value ranges, and each random noise value is superimposed onto each numerical vehicle condition parameter to generate a disturbed numerical parameter, including: For each numerical vehicle condition parameter, a continuous probability distribution that meets preset conditions is determined within the corresponding disturbance noise value range; Random noise values are extracted from each continuous probability distribution, and the random noise values are added to the corresponding numerical vehicle condition parameters. The summation result that exceeds the corresponding allowed value range is limited to the range endpoint closest to the summation result, and the summation result representing the number of times is adjusted to an integer to obtain the perturbed numerical parameter; The vehicle identification number, the transaction platform identifier of the second transaction platform, the field name of the numerical vehicle condition parameter, the privacy budget period, the single release budget, and the release time are written into the privacy release log.
[0012] Optionally, the categorical vehicle condition parameter and the disturbed numerical parameter are jointly sent to the second trading platform for vehicle price matching to obtain cross-domain price inquiry feedback results. These results are then distributed to the user through the first trading platform, including: Establish a secure cross-domain communication link between the first and second trading platforms, and generate a one-time request identifier for this cross-domain inquiry. Delete the vehicle identification number and user identity information from the data to be sent, and replace the vehicle model version with the vehicle model grouping that contains the vehicle model version in the regional transaction statistics summary; The one-time request identifier, the target inquiry area identifier, the vehicle condition parameter of the completed replacement category, and the disturbed numerical parameter are encapsulated into a cross-domain request data packet; The cross-domain request data packet is sent to the second transaction platform through the cross-domain secure communication link, so that the second transaction platform determines the matching price according to the value segment of the vehicle condition parameter and the value segment of the disturbed numerical parameter. The system receives the one-time request identifier and the matching price returned by the second trading platform, determines the user who initiated the cross-domain inquiry request based on the one-time request identifier, and provides the matching price to the user through the first trading platform.
[0013] Secondly, this application provides a cross-domain used car price inquiry data sharing system based on differential privacy, including: The acquisition and parsing module is used to acquire cross-domain inquiry requests initiated by users for target vehicles in the first trading platform, and to parse the cross-domain inquiry requests to obtain the target vehicle's categorical vehicle condition parameters, numerical vehicle condition parameters, and target inquiry area identifier. The extraction module is used to determine the second trading platform that accepts the inquiry based on the target inquiry area identifier, and to extract the historical query frequency of the numerical vehicle condition parameter in the first trading platform; The determination module is used to determine the privacy budget consumption of the target vehicle based on the historical query frequency, and subtract the privacy budget consumption from the pre-configured initial privacy budget to obtain the currently available privacy budget; The setting module is used to extract the price sensitivity weights corresponding to the numerical vehicle condition parameters, and set the disturbance noise value ranges corresponding to each of the numerical vehicle condition parameters according to the correlation between the current available privacy budget and the price sensitivity weights. The generation and superposition module is used to generate random noise values in their respective disturbance noise value ranges, and superimpose each random noise value onto each numerical vehicle condition parameter to generate disturbed numerical parameters. The matching module is used to send the categorical vehicle condition parameters and the disturbed numerical parameters to the second trading platform to perform vehicle price matching, obtain cross-domain inquiry feedback results, and send the cross-domain inquiry feedback results to the user through the first trading platform.
[0014] Thirdly, this application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement a method for cross-domain price inquiry data sharing of used cars based on differential privacy as described in the first aspect above.
[0015] Fourthly, this application provides a computer storage medium storing a computer program, which, when executed by a computer, implements a method for cross-domain price inquiry data sharing of used cars based on differential privacy as described in the first aspect.
[0016] This application links the target vehicle, the second trading platform, numerical vehicle condition parameters, and the privacy budget period into a unified data release scope, ensuring that repeated inquiries within the same data release scope jointly consume the privacy budget. The transaction price change provided by the regional transaction statistics summary is used to allocate the field privacy budget, which in turn determines the noise variation scale of the corresponding numerical vehicle condition parameters. After deleting the vehicle identification number and user identity information and completing vehicle model grouping replacement, the first trading platform only sends the disturbed numerical parameters and the replaced category-based vehicle condition parameters, thus creating a continuous processing flow for privacy budget control, numerical disturbance, cross-domain transmission, and vehicle price matching.
[0017] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 The flowchart illustrates a cross-domain price inquiry data sharing method for used cars based on differential privacy, as provided in this application.
[0020] Figure 2 This illustration shows the process of selecting a second trading platform and statistically analyzing historical query frequency provided in this application.
[0021] Figure 3 This diagram illustrates the process of transitioning the publication status of the privacy publication log provided in this application.
[0022] Figure 4 A flowchart of the numerical vehicle condition parameter disturbance processing procedure provided in this application is shown.
[0023] Figure 5 A timing diagram of the cross-domain inquiry interaction process between the first and second trading platforms provided in this application is shown.
[0024] Figure 6 The diagram shows a schematic of the structure of a cross-domain price inquiry data sharing system for used cars based on differential privacy provided in this application. Detailed Implementation
[0025] To enable those skilled in the art to understand the technical solution of this application, the following description, in conjunction with the accompanying drawings, explains the method for cross-domain price inquiry data sharing of used cars based on differential privacy.
[0026] The first trading platform in this application is one that receives cross-domain price inquiry requests from users and stores the original vehicle condition data of the target vehicle. The second trading platform is one that covers the target price inquiry area and determines the matching price based on the cross-domain price inquiry data.
[0027] Categorical vehicle condition parameters are used to represent discrete attributes of a vehicle, including vehicle brand, model, version, energy type, and damage level. Numerical vehicle condition parameters are used to represent the vehicle's usage status, which can be expressed numerically, including vehicle age, mileage, number of ownership transfers, and number of repairs.
[0028] The regional transaction statistics summary is a statistical data set pre-generated by the second trading platform based on historical transaction records within the target inquiry area and provided to the first trading platform. The regional transaction statistics summary does not include transaction records for individual vehicles; it primarily records vehicle type groupings, the allowable range of numerical vehicle condition parameters, value segments, and price changes between adjacent value segments.
[0029] like Figure 1 As shown, the method provided in this application includes steps S101 to S106.
[0030] Step S101: The parsing module obtains the cross-domain inquiry request initiated by the user for the target vehicle in the first trading platform, and parses the cross-domain inquiry request to obtain the target vehicle's category-type vehicle condition parameters, numerical-type vehicle condition parameters, and target inquiry area identifier.
[0031] Specifically, users input or confirm information about the target vehicle through their client and specify the target region for which they wish to inquire about pricing. The client then generates a cross-regional pricing request message based on this information and sends it to the first trading platform.
[0032] The cross-domain price inquiry request message includes the vehicle identification number, target price inquiry area code, vehicle brand and series, model version, energy type, damage level, vehicle age, mileage, number of ownership transfers, and number of repairs.
[0033] The parsing module reads various data based on the field names in the cross-domain inquiry request message, identifies the vehicle brand, model, version, energy type, and damage level as categorical vehicle condition parameters, identifies the vehicle age, mileage, number of ownership transfers, and number of repairs as numerical vehicle condition parameters, and identifies the target inquiry area code as the target inquiry area identifier.
[0034] For example, in a cross-domain price inquiry request, the vehicle brand and series are A series, the model version is A1, the energy type is fuel, the damage level is minor damage, the vehicle age is 6 years, the mileage is 78,000 kilometers, the number of ownership transfers is 1, the number of repairs is 2, and the target price inquiry area code is R02. The parsing module then obtains four categorical vehicle condition parameters, four numerical vehicle condition parameters, and the target price inquiry area identifier R02.
[0035] The acquisition and parsing module also checks whether the cross-domain quotation request message is missing any necessary fields, and whether the vehicle age, mileage, number of ownership transfers, and number of repairs are non-negative numbers. If any necessary field is missing or the field type is incorrect, the acquisition and parsing module stops the current cross-domain quotation request.
[0036] The above request message parsing and field type checking can be completed using message field reading methods commonly used in this field, without further restrictions on the specific message format.
[0037] In step S102, the extraction module determines the second trading platform that accepts the inquiry based on the target inquiry area identifier, and extracts the historical query frequency of numerical vehicle condition parameters within the first trading platform.
[0038] like Figure 2 As shown, the first trading platform maintains a regional service registration form. This form records each trading platform's trading platform identifier, service area code, and current response time.
[0039] The first trading platform periodically sends probe requests to registered trading platforms and records the time from sending the probe request to receiving the probe response as the response duration. To avoid the influence of a single network fluctuation on platform selection, the current response duration can be the median of the response durations obtained from the most recent successful probes.
[0040] The extraction module first searches the regional service registration table for trading platforms whose service area code matches the target inquiry area identifier, and then selects the trading platform with the shortest current response time from the found trading platforms as the second trading platform.
[0041] For example, if the target inquiry area is identified as R02, and there are two trading platforms covering the R02 area in the area service registration table, with current response times of 82 milliseconds and 105 milliseconds respectively, the extraction module will identify the trading platform with a current response time of 82 milliseconds as the second trading platform.
[0042] When the current response times of two trading platforms are the same, the second trading platform can be determined according to the character order of the trading platform identifiers; when there is no trading platform covering the target inquiry area, the first trading platform returns information to the user that the target inquiry area is temporarily unavailable.
[0043] After identifying the second trading platform, the extraction module generates a release tag based on the vehicle identification number, the trading platform identifier of the second trading platform, the field names of numerical vehicle condition parameters, and the privacy budget period.
[0044] The release tag is used to distinguish data release records from different vehicles, different receiving platforms, and different privacy budget periods. The first transaction platform can generate the release tag using a keyed hash operation. The vehicle identification number (VIN) participates in the generation of the release tag, but is not directly sent to the second transaction platform as part of the release tag content.
[0045] The privacy budget cycle is a pre-defined cumulative budget period stipulated by the first trading platform. For example, the privacy budget cycle may begin at midnight on the first day of each calendar month and end at midnight on the first day of the following calendar month. The key used to generate the release token remains unchanged within the same privacy budget cycle.
[0046] The extraction module uses the publication tag as the search condition to find publication records in the privacy publication log that have output disturbed numerical parameters to the second trading platform, and determines the number of publication records found as the historical query frequency.
[0047] For example, if there are three published records in the privacy publication log that have the same publication tag as the current publication, then the historical query frequency is 3 times.
[0048] The release tag links the vehicle, the receiving platform, the data fields, and the privacy budget period. Therefore, the historical query frequency represents the number of times the same numerical vehicle condition parameters for the target vehicle have been released to the same second trading platform within the current privacy budget period, rather than the total number of queries for all vehicles on the first trading platform.
[0049] Step S103: The determination module determines the privacy budget consumption of the target vehicle based on the historical query frequency, and subtracts the privacy budget consumption from the pre-configured initial privacy budget to obtain the currently available privacy budget.
[0050] A privacy budget is a dimensionless value used to limit the cumulative release of numerical vehicle condition parameters. The first trading platform configures an initial privacy budget for the data release range corresponding to the release marker. Each time a disturbed numerical parameter is released to the second trading platform, the first trading platform records the single release budget used in the corresponding release record.
[0051] The module determines the number of publication records to be read according to the historical query frequency and accumulates the single publication budget in each publication record.
[0052] Let the initial privacy budget be Historical query frequency is , No. The budget for a single release in the release record is The privacy budget consumption is:
[0053] The current available privacy budget is:
[0054] in, Indicates the amount of privacy budget spent. This represents the currently available privacy budget. All the above parameters are dimensionless values and can be directly added or subtracted.
[0055] For example, if the initial privacy budget is 20, the historical query frequency is 3 times, and the single-post budgets for the three post records are 4, 3, and 4 respectively, then the privacy budget consumption is 11, and the current available privacy budget is 9.
[0056] When the current available privacy budget is zero, the module determines whether to stop the current cross-domain pricing and returns a prompt message to the user indicating that cross-domain pricing will be stopped for the current privacy budget period.
[0057] The first trading platform allows users to pre-configure a single release budget cap. The single release budget is the smaller of the currently available privacy budget and the single release budget cap.
[0058] Taking the aforementioned current available privacy budget of 9 as an example, when the single-release budget limit is 8, the single-release budget used in this cross-domain quotation is 8. This avoids using the entire current available privacy budget in a single cross-domain quotation.
[0059] In actual operation, multiple cross-domain quotation requests may simultaneously read the same currently available privacy budget. To avoid duplicate allocation, the determination module locks the corresponding budget record when determining the budget for a single release and records this single release budget as a pre-allocated quota. Other requests cannot reuse the corresponding quota until the pre-allocated quota is released or converted into a released budget. Database locking and transaction commit can be implemented using database transaction methods commonly used in this field.
[0060] Step S104: Set the price sensitivity weights corresponding to the numerical vehicle condition parameters extracted by the module, and set the disturbance noise value ranges corresponding to each numerical vehicle condition parameter according to the correlation between the current available privacy budget and the price sensitivity weights.
[0061] This step is the core processing procedure of this application. Different numerical vehicle condition parameters have varying degrees of impact on used car transaction prices. Therefore, this application does not assign the same field privacy budget to the four numerical vehicle condition parameters, but rather allocates it based on the transaction price changes in the regional transaction statistics summary.
[0062] When the second trading platform generates a regional transaction statistics summary, it first divides historical transaction records by vehicle brand, model, energy type, and damage level, and then divides vehicle age, mileage, number of transfers, and number of repairs into multiple value segments.
[0063] The second trading platform calculates the median transaction price within adjacent value segments and the difference between adjacent median transaction prices. The regional transaction statistics summary records these differences as the change in transaction price for the corresponding numerical vehicle condition parameter.
[0064] For example, the data recorded in the regional transaction statistics summary is as follows.
[0065]
[0066] The average transaction price change in the table is the arithmetic mean of the transaction price changes between adjacent valid value segments. The second trading platform only uses the corresponding transaction price change when both adjacent value segments have a sufficient number of historical transaction records.
[0067] Let the first The change in transaction price for each numerical vehicle condition parameter is The number of numerical vehicle condition parameters is The minimum allocation coefficient is Price sensitivity weighting is:
[0068] in, Indicates the first Price sensitivity weights for numerical vehicle condition parameters This is used to ensure that numerical vehicle condition parameters with small changes in transaction prices still achieve non-zero field privacy budgets, and to meet... .
[0069] All price changes are expressed in the same currency unit, therefore The ratio to the sum of all price changes is a dimensionless value. The price sensitivity weight is also a dimensionless value.
[0070] When all transaction price changes are zero, the setting module will distribute the single release budget evenly among the numerical vehicle condition parameters.
[0071] Taking the data in the previous table as an example, the number of numerical vehicle condition parameters is 4, the minimum allocation coefficient is 0.05, and the sum of all transaction price changes is 40,000 yuan. The calculated price sensitivity weights for vehicle age, mileage, number of ownership transfers, and number of repairs are 0.41, 0.29, 0.19, and 0.11, respectively, and the sum of the four price sensitivity weights is 1.
[0072] Let the budget for a single release be Then the first The privacy budget for each numerical vehicle condition parameter field is:
[0073] When the budget for a single release is 8, the privacy budgets for the fields of vehicle age, mileage, number of ownership transfers, and number of repairs are 3.28, 2.32, 1.52, and 0.88, respectively, and the sum of the privacy budgets for the four fields is 8.
[0074] Let the lower limit of the allowed value range be . The upper limit of the allowed value range is The privacy sensitivity of the field is:
[0075] The noise variation scale is:
[0076] in, Indicates the first The maximum possible change of a numerical vehicle condition parameter within its permissible range. This indicates the corresponding noise variation scale.
[0077] Since the privacy budget field is a dimensionless value, the noise variation scale uses the same unit as the corresponding numerical vehicle condition parameter. For example, the noise variation scale for vehicle age is in years, and the noise variation scale for mileage is in kilometers.
[0078] Based on the aforementioned data, the noise variation scales of the four numerical vehicle condition parameters are as follows.
[0079]
[0080] As can be seen from the table, numerical vehicle condition parameters with larger changes in transaction price achieve larger field privacy budgets and correspond to smaller relative noise change scales.
[0081] The setting module determines the range of all real numbers centered at zero as the range of disturbance noise values, and provides the noise change scale to step S105 to determine the distribution of random noise values within the range of disturbance noise values.
[0082] like Figure 4 As shown, the regional transaction statistics summary, price sensitivity weight, field privacy budget, field privacy sensitivity, and noise change scale form a continuous processing relationship, with the result of the previous processing serving as the input for the next processing step.
[0083] Step S105: The generation and superposition module generates random noise values in their respective disturbance noise value ranges, and superimposes each random noise value onto each numerical vehicle condition parameter to generate the disturbed numerical parameter.
[0084] The generation and overlay module determines the noise variation scale centered at zero for each numerical vehicle condition parameter. The probability density function of a continuous Laplace distribution is:
[0085] in, This represents the value of random noise. This indicates the noise variation scale of the corresponding numerical vehicle condition parameter. Indicates the extraction of random noise values The probability density.
[0086] Random noise values and noise variation scales have the same unit, therefore the exponential part... It is a dimensionless numerical value.
[0087] When generating random noise values, the generation and superposition module can first generate values located at... Uniform random numbers within Then, the random noise value is obtained according to the following formula:
[0088] in, The sign of a uniformly random number is indicated by its positive or negative sign. Since uniformly random numbers are dimensionless, the input to the logarithmic function is also a dimensionless value.
[0089] The generation and overlay module adds random noise values to the corresponding numerical vehicle condition parameters:
[0090] in, This represents the original numerical vehicle condition parameters. This represents the sum of values after adding random noise.
[0091] When the sum is less than the lower limit of the allowed value range, the generation and overlay module adjusts the sum to the lower limit of the allowed value range; when the sum is greater than the upper limit of the allowed value range, the generation and overlay module adjusts the sum to the upper limit of the allowed value range.
[0092] For the number of ownership transfers and repairs, the overlay generation module further adjusts the sum to the closest integer. For vehicle age and mileage, the overlay generation module either retains a decimal or adjusts it to an integer according to the record precision of the corresponding field.
[0093] Taking the aforementioned noise variation scale as an example, the original numerical vehicle condition parameters and random noise values are as follows.
[0094]
[0095] The vehicle age in the table is recorded to one decimal place, the mileage is recorded as an integer in kilometers, and the number of ownership transfers and repairs is recorded as integers.
[0096] For a single numerical vehicle condition parameter, the noise variation scale is obtained by dividing the field privacy sensitivity by the field privacy budget, so that the continuous Laplace distribution satisfies the differential privacy condition limited by the corresponding field privacy budget.
[0097] When perturbating the four numerical vehicle condition parameters individually, the sum of the privacy budgets for the four fields equals the single release budget. Therefore, the privacy budget consumption of a single cross-domain query does not exceed the corresponding single release budget.
[0098] After the generation and overlay module completes the generation of disturbed numerical parameters, it writes the vehicle identification number, the transaction platform identifier of the second transaction platform, the field name of the numerical vehicle condition parameter, the privacy budget period, the single release budget, and the release time into the privacy release log.
[0099] like Figure 3 As shown, a publishing record can have three states: reserved, published, and canceled. When data has not yet been sent, the publishing record is in the reserved state; after the second trading platform confirms receipt of the cross-domain request data packet, the publishing record changes to the published state; if a failure occurs before the cross-domain request data packet is sent, the publishing record changes to the canceled state.
[0100] In step S106, the matching module sends the categorical vehicle condition parameters and the disturbed numerical parameters to the second trading platform to perform vehicle price matching, obtain cross-domain price inquiry feedback results, and then sends the cross-domain price inquiry feedback results to the user through the first trading platform.
[0101] The matching module first establishes a cross-domain secure communication link between the first trading platform and the second trading platform, and generates a one-time request identifier for the current cross-domain inquiry.
[0102] Establishing a secure cross-domain communication link can employ commonly used two-way certificate authentication and encrypted transmission methods in this field, without further restrictions on specific communication protocols.
[0103] Before sending data, the matching module removes the vehicle identification number and user identity information, and replaces the vehicle model version with the vehicle model grouping contained in the regional transaction statistics summary.
[0104] Vehicle model groups are pre-determined by the second trading platform based on vehicle brand, model series, vehicle generation, body style, energy type, and power specifications. Multiple vehicle versions that can be included in the transaction price statistics can be grouped into the same vehicle model group.
[0105] For example, if vehicle model version A1 and vehicle model version A2 belong to the same vehicle brand, model series, vehicle generation, body style, and energy type, and their power specifications are within the same range, the second trading platform can group vehicle model version A1 and vehicle model version A2 into vehicle model group G01. When the target vehicle's vehicle model version is A1, the matching module replaces vehicle model version A1 with vehicle model group G01 in the data to be sent.
[0106] The vehicle condition parameters that have been replaced include vehicle brand and series, model group, energy type, and damage level.
[0107] The matching module encapsulates the one-time request identifier, the target inquiry area identifier, the vehicle condition parameter of the completed replacement category, and the disturbed numerical parameter into a cross-domain request data packet, and sends it to the second trading platform through the cross-domain secure communication link.
[0108] The second trading platform searches for the corresponding historical transaction records based on the vehicle condition parameters of the replaced categories, and determines the value segments of the disturbed parameters such as vehicle age, mileage, number of transfers, and number of repairs.
[0109] The second trading platform first selects historical transaction records where all four numerical vehicle condition parameters fall within the same value range. When the number of historical transaction records is less than the preset minimum matching quantity, the second trading platform sequentially adds historical transaction records adjacent to the current value range.
[0110] The specific matching rules are as follows.
[0111]
[0112] After identifying the candidate transaction records, the second trading platform sorts the transaction prices of the candidate transaction records from smallest to largest, and determines the median transaction price as the matching price.
[0113] For example, if the second trading platform obtains 27 candidate transaction records, then the 14th transaction price after sorting will be the matching price. When the 14th transaction price after sorting is 101,000 yuan, the matching price will be 101,000 yuan.
[0114] The second trading platform encapsulates the one-time request identifier and matching price into a cross-domain inquiry feedback result and returns it to the first trading platform.
[0115] The matching module uses the one-time request identifier to locate the user who initiated the cross-domain price inquiry request and sends the matched price to the corresponding user. The one-time request identifier is used to associate the request and the feedback result and does not contain vehicle identification number or user identity information.
[0116] like Figure 5 As shown, the user client first sends a cross-domain price inquiry request to the first trading platform. After the first trading platform completes privacy budget processing and numerical perturbation, it sends a cross-domain request data packet to the second trading platform. After the second trading platform completes vehicle price matching, it returns the cross-domain price inquiry feedback result to the first trading platform. Finally, the first trading platform provides the matched price to the user.
[0117] When the first trading platform has sent a cross-domain request data packet but has not received a response from the second trading platform, the matching module resends the same cross-domain request data packet using the same one-time request identifier. When the second trading platform recognizes the same one-time request identifier, it returns the already saved processing result and does not repeat the vehicle price matching.
[0118] If the first trading platform cannot confirm whether the second trading platform has received the cross-domain request data packet, it will process the corresponding published record as published to avoid the output of disturbed numerical parameters not being included in the privacy budget consumption.
[0119] like Figure 6 As shown, the cross-domain price inquiry data sharing system for used cars based on differential privacy includes an acquisition and parsing module 201, an extraction module 202, a determination module 203, a setting module 204, a generation and overlay module 205, and a matching module 206.
[0120] The parsing module 201 is used to execute step S101 to obtain categorical vehicle condition parameters, numerical vehicle condition parameters, and target inquiry area identifiers.
[0121] The extraction module 202 is used to execute step S102, determine the second trading platform and obtain the historical query frequency.
[0122] The determination module 203 is used to perform step S103 to obtain the privacy budget consumption and the currently available privacy budget.
[0123] The setting module 204 is used to execute step S104 to obtain the price sensitivity weight, field privacy budget, noise change scale, and disturbance noise value range.
[0124] The generation and superposition module 205 is used to execute step S105 to generate random noise values and disturbed numerical parameters.
[0125] The matching module 206 is used to execute step S106, sending a cross-domain request data packet to the second trading platform, receiving the cross-domain inquiry feedback result, and providing the matching price to the user.
[0126] The modules described above represent different functions executed by a computer program, and it is not required that each module be set up with a separate physical device. Multiple modules can be executed by the same processing component in the first trading platform, or they can be deployed on different servers of the first trading platform.
[0127] The computing device includes a processing component, a storage component, and a communication component. The storage component stores the computer program, regional service registration forms, regional transaction statistics summaries, and privacy release logs. When the processing component executes the computer program, it completes steps S101 to S106. The communication component enables data communication between the first trading platform and the second trading platform.
[0128] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
Claims
1. A method for cross-domain price inquiry data sharing of used cars based on differential privacy, characterized in that, include: Obtain cross-domain price inquiry requests initiated by users for the target vehicle in the first trading platform, and parse the cross-domain price inquiry requests to obtain the target vehicle's category-type vehicle condition parameters, numerical vehicle condition parameters, and target price inquiry area identifier; The second trading platform for accepting inquiries is determined based on the target inquiry area identifier, and the historical query frequency of the numerical vehicle condition parameters within the first trading platform is extracted. The privacy budget consumption of the target vehicle is determined based on the historical query frequency, and the privacy budget consumption is subtracted from the pre-configured initial privacy budget to obtain the current available privacy budget. Extract the price sensitivity weights corresponding to the numerical vehicle condition parameters, and set the disturbance noise value ranges corresponding to each of the numerical vehicle condition parameters according to the correlation between the current available privacy budget and the price sensitivity weights. Random noise values are generated within their respective disturbance noise value ranges, and each random noise value is superimposed onto each numerical vehicle condition parameter to generate disturbed numerical parameters. The categorical vehicle condition parameters and the disturbed numerical parameters are sent together to the second trading platform for vehicle price matching, to obtain cross-domain price inquiry feedback results, and then the cross-domain price inquiry feedback results are sent to the user through the first trading platform.
2. The method according to claim 1, characterized in that, The system obtains cross-domain price inquiry requests initiated by users for target vehicles on the first trading platform, and parses these requests to obtain the target vehicle's categorical vehicle condition parameters, numerical vehicle condition parameters, and target price inquiry region identifier, including: Receive a cross-domain price inquiry request message sent by a user client. The cross-domain price inquiry request message includes the vehicle identification number, target price inquiry area code, vehicle brand and series, model version, energy type, damage level, vehicle age, mileage, number of ownership transfers, and number of repairs. The vehicle brand and series, the vehicle model version, the energy type, and the damage level are defined as category-type vehicle condition parameters; The vehicle age, mileage, number of ownership transfers, and number of repairs are defined as numerical vehicle condition parameters. The target inquiry area code is determined as the target inquiry area identifier.
3. The method according to claim 1, characterized in that, The second trading platform accepting the inquiry is determined based on the target inquiry area identifier, and the historical query frequency of the numerical vehicle condition parameter within the first trading platform is extracted, including: Read the regional service registration form that records the transaction platform identifier, service area code, and current response time; Select the trading platform with the same service area code as the target inquiry area identifier and the shortest current response time from the regional service registration table as the second trading platform; A release tag is generated based on the vehicle identification number, the transaction platform identifier of the second transaction platform, the field name of the numerical vehicle condition parameter, and the predefined privacy budget period; Retrieve from the privacy release log of the first trading platform the release record that has the release tag and has output the disturbed numerical parameter to the second trading platform; The number of published records is counted, and the statistical result is determined as the historical query frequency.
4. The method according to claim 1, characterized in that, The privacy budget consumption of the target vehicle is determined based on the historical query frequency. This privacy budget consumption is then subtracted from the pre-configured initial privacy budget to obtain the currently available privacy budget, which includes: Read the initial privacy budget of the target vehicle during the privacy budget period; Based on the historical query frequency, retrieve the corresponding number of publishing records from the privacy publishing log, and read the single publishing budget recorded in each publishing record; The privacy budget consumption of the target vehicle during the privacy budget period is obtained by summing up the read single release budget. The current available privacy budget is obtained by subtracting the privacy budget consumption from the initial privacy budget. When the current available privacy budget is zero, stop generating disturbed numerical parameters and return a prompt message to the user indicating that cross-domain price inquiries will be stopped for this privacy budget period.
5. The method according to claim 1, characterized in that, Extract the price sensitivity weights corresponding to the numerical vehicle condition parameters, and based on the correlation between the currently available privacy budget and the price sensitivity weights, set the disturbance noise value ranges for each of the numerical vehicle condition parameters, including: Read the regional transaction statistics summary provided by the second trading platform and stored in the first trading platform before cross-domain inquiry. The regional transaction statistics summary records the allowed value range of each numerical vehicle condition parameter under different types of vehicle condition parameters and the transaction price change between adjacent value segments. The price sensitivity weight of each numerical vehicle condition parameter is determined based on the proportion of the change in transaction price corresponding to each numerical vehicle condition parameter in the total change in transaction price. The currently available privacy budget is allocated as a field privacy budget according to the price sensitivity weight; The sum of the privacy budgets for all fields is determined as the single release budget, and the single release budget does not exceed the currently available privacy budget; The difference between the two ends of the allowed range of values for each numerical vehicle condition parameter is determined as the field's privacy sensitivity. Based on the ratio of the field privacy sensitivity to the field privacy budget, the noise change scale is determined, and the range of all real numbers centered at zero is determined as the disturbance noise range for the corresponding numerical vehicle condition parameter.
6. The method according to claim 1, characterized in that, Random noise values are generated within their respective disturbance noise value ranges. These random noise values are then superimposed onto the corresponding numerical vehicle condition parameters to generate disturbed numerical parameters, including: For each numerical vehicle condition parameter, a continuous probability distribution that meets preset conditions is determined within the corresponding disturbance noise value range; Random noise values are extracted from each continuous probability distribution, and the random noise values are added to the corresponding numerical vehicle condition parameters. The summation result that exceeds the corresponding allowed value range is limited to the range endpoint closest to the summation result, and the summation result representing the number of times is adjusted to an integer to obtain the perturbed numerical parameter; The vehicle identification number, the transaction platform identifier of the second transaction platform, the field name of the numerical vehicle condition parameter, the privacy budget period, the single release budget, and the release time are written into the privacy release log.
7. The method according to claim 1, characterized in that, The categorical vehicle condition parameters and the disturbed numerical parameters are jointly sent to the second trading platform for vehicle price matching to obtain cross-domain price inquiry feedback results. These results are then distributed to the user through the first trading platform, including: Establish a secure cross-domain communication link between the first and second trading platforms, and generate a one-time request identifier for this cross-domain inquiry. Delete the vehicle identification number and user identity information from the data to be sent, and replace the vehicle model version with the vehicle model grouping that contains the vehicle model version in the regional transaction statistics summary; The one-time request identifier, the target inquiry area identifier, the vehicle condition parameter of the completed replacement category, and the disturbed numerical parameter are encapsulated into a cross-domain request data packet; The cross-domain request data packet is sent to the second transaction platform through the cross-domain secure communication link, so that the second transaction platform determines the matching price according to the value segment of the vehicle condition parameter and the value segment of the disturbed numerical parameter. The system receives the one-time request identifier and the matching price returned by the second trading platform, determines the user who initiated the cross-domain inquiry request based on the one-time request identifier, and provides the matching price to the user through the first trading platform.
8. A cross-domain used car price inquiry data sharing system based on differential privacy, characterized in that, The system is used to implement the differential privacy-based cross-domain price inquiry data sharing method for used cars as described in any one of claims 1-7; the system includes: The acquisition and parsing module is used to acquire cross-domain inquiry requests initiated by users for target vehicles in the first trading platform, and to parse the cross-domain inquiry requests to obtain the target vehicle's categorical vehicle condition parameters, numerical vehicle condition parameters, and target inquiry area identifier. The extraction module is used to determine the second trading platform that accepts the inquiry based on the target inquiry area identifier, and to extract the historical query frequency of the numerical vehicle condition parameter in the first trading platform; The determination module is used to determine the privacy budget consumption of the target vehicle based on the historical query frequency, and subtract the privacy budget consumption from the pre-configured initial privacy budget to obtain the currently available privacy budget; The setting module is used to extract the price sensitivity weights corresponding to the numerical vehicle condition parameters, and set the disturbance noise value ranges corresponding to each of the numerical vehicle condition parameters according to the correlation between the current available privacy budget and the price sensitivity weights. The generation and superposition module is used to generate random noise values in their respective disturbance noise value ranges, and superimpose each random noise value onto each numerical vehicle condition parameter to generate disturbed numerical parameters. The matching module is used to send the categorical vehicle condition parameters and the disturbed numerical parameters to the second trading platform to perform vehicle price matching, obtain cross-domain inquiry feedback results, and send the cross-domain inquiry feedback results to the user through the first trading platform.
9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement a method for cross-domain price inquiry data sharing of used cars based on differential privacy as described in any one of claims 1-7.
10. A computer storage medium, characterized in that, The system contains a computer program that, when executed by a computer, implements a method for cross-domain price inquiry data sharing of used cars based on differential privacy, as described in any one of claims 1-7.