A vehicle location privacy protection method and system based on local differential privacy

Abstract

The application relates to the field of privacy protection, in particular to a vehicle location privacy protection method and system based on local differential privacy. The method comprises the following steps: dividing a target area where a vehicle runs to obtain at least one large area, and dividing each large area to obtain a plurality of small areas; obtaining a total variance of a vehicle number estimation of each small area in the large area based on a disturbance flip probability of epsilon-local differential privacy, and adding the total variances of the vehicle number estimations of all the small areas in the large area to obtain a comprehensive total variance; obtaining an optimization function based on the comprehensive total variance, and solving the optimization function to obtain an optimal disturbance flip probability; and taking the optimal disturbance flip probability as the disturbance flip probability of epsilon-local differential privacy to disturb the location information of the vehicle, so that the privacy of the vehicle location is protected. The application can realize the optimal balance between privacy protection and data availability, and improves the evaluation accuracy of vehicle distribution.

Description

Technical Field

[0001] This invention relates to the field of privacy protection, specifically to a method and system for protecting vehicle location privacy based on local differential privacy. Background Technology

[0002] With the development of vehicle-road-cloud multi-network convergence vehicle-to-everything (V2X) technology, vehicle location information is shared and exchanged between the vehicle, roadside, and cloud. Hackers are intercepting large amounts of vehicle location data through cyberattacks. On the one hand, the location information of connected vehicles contains critical national infrastructure information, which may fall into the hands of criminals. For example, by analyzing vehicle density and flow, the geographical location of sensitive infrastructure can be inferred, and therefore it may be used for targeted attacks for malicious purposes. On the other hand, the time-series location data of the same vehicle constitutes trajectory information, which can reveal sensitive information such as personal habits, whereabouts, and even home addresses. Once this data is illegally obtained and used, it will seriously infringe on personal privacy.

[0003] Differential privacy technology does not rely on the attacker's background knowledge, providing a strict and quantifiable level of privacy protection. The relationship between privacy and data effects can be balanced by adjusting the privacy budget. Differential privacy offers a robust privacy protection mechanism, resisting various privacy attacks such as differential attacks, background knowledge attacks, and connection attacks. It is particularly suitable for location data privacy protection in data mining and multi-network convergence scenarios in the big data era. Local differential privacy is the application of differential privacy technology in local scenarios. It does not rely on a trusted third party; users can independently perturb their data locally before submitting the perturbed data to the service provider. This local processing mechanism provides excellent protection for user privacy. For example, Chinese patent CN111915891A discloses a real-time traffic detection method based on local differential privacy. It uses ε-local differential privacy protection on the vehicle's real location data—that is, perturbing the vehicle's real location data—so that the server cannot know the vehicle's true location, yet it can still construct a real-time traffic map and provide accurate real-time traffic information. However, this method relies entirely on experience or human intervention to determine the perturbation probability, making it difficult to achieve an optimal balance between privacy protection and data availability, and consequently, to guarantee the accuracy of the final vehicle distribution assessment. Summary of the Invention

[0004] To address the challenge of achieving an optimal balance between privacy protection and data availability in local differential privacy, which in turn makes it difficult to guarantee the accuracy of the final vehicle distribution assessment, this invention proposes a vehicle location privacy protection method based on local differential privacy, comprising: The target area for vehicle operation is divided into at least one large area, and each large area is further divided into multiple smaller areas. Based on the perturbation flip probability of ε-local differential privacy, the total variance of the number of vehicles in each small area in the large area is estimated, and the total variance of the number of vehicles in all small areas in the large area is summed to obtain the comprehensive total variance. The optimization function is obtained based on the comprehensive total variance, and the optimal perturbation flip probability is obtained by solving the optimization function. The optimal perturbation flip probability is used as the perturbation flip probability of ε-local differential privacy to perturb the vehicle's position information, thereby achieving privacy protection of the vehicle's position.

[0005] Optionally, obtaining the optimization function based on the comprehensive total variance includes: The overall variance is integrated, and the constant coefficients are discarded to obtain the primary optimization function; The correlation between perturbation flip probability and privacy budget in large areas of ε-local differential privacy is obtained based on the unary coding mechanism of ε-local differential privacy; The primary optimization function is simplified using the aforementioned correlation to obtain the optimized function.

[0006] Optionally, the privacy budget for a large area in the ε-local differential privacy is obtained through the following steps: Obtain a score showing the correlation between individual privacy and each sub-region within a larger region; The correlation level scores of each sub-region are summed to obtain the comprehensive correlation level score of the corresponding large region. The privacy budget for a large region is determined based on the comprehensive correlation score.

[0007] Optionally, solving the optimization function to obtain the optimal perturbation flip probability includes: The optimal perturbation flip probability is obtained by solving the optimization function using the simulated annealing algorithm, wherein the state variable of the simulated annealing algorithm is the perturbation flip probability, and the optimization objective of the simulated annealing algorithm is to minimize the optimization function.

[0008] Optionally, the simulated annealing algorithm is based on the state of the nth iteration. It becomes the state of the (n+1)th iteration. The acceptance probability P is:

[0009] in, Let be the value of the optimization function obtained in the nth iteration. Let T be the value of the optimization function obtained in the (n+1)th iteration, and T be the temperature.

[0010] Optionally, the step of perturbing the vehicle's position information using the optimal perturbation flip probability as the perturbation flip probability of ε-local differential privacy includes: The vehicle's location information is unigram encoded to obtain a one-dimensional vector; The optimal perturbation flip probability is used as the perturbation flip probability of ε-local differential privacy to flip the elements in the one-dimensional vector to achieve perturbation of vehicle position information.

[0011] Optionally, the optimal perturbation flip probability includes a first optimal perturbation flip probability and a second optimal perturbation flip probability. After using the optimal perturbation flip probability as the perturbation flip probability of ε-local differential privacy, the perturbation flip probability of ε-local differential privacy becomes:

[0012]

[0013] in, This represents the first optimal perturbation flip probability. This represents the second-optimal perturbation flip probability. For the first vector A, the first vector is... The possible values ​​of each element, For the perturbation of the first The possible values ​​of each element, Let be a probability function, 0 <q*<p*<1, Budget for privacy.

[0014] Optionally, the step of unary encoding the vehicle's location information to obtain a one-dimensional vector includes: The large regions and the smaller regions within them are numbered sequentially. Obtain the vehicle's GPS location information; Determine the large area and corresponding small area where the GPS location is located; If we assign a value of 1 to the small region where the vehicle is located within the large region, and assign a value of 0 to the other small regions, we obtain a one-dimensional vector.

[0015] Optionally, after perturbing the vehicle's position information based on the optimal perturbation rollover probability, the method further includes: The location vectors obtained after disturbance are transmitted to the cloud management center or roadside infrastructure to perform aggregation analysis on the location vectors to obtain the estimated number of vehicles corresponding to each small area; The estimated number of vehicles is corrected by maximum likelihood estimation to obtain an unbiased estimate of the number of vehicles in each small region. Based on the unbiased estimate of the number of vehicles in each small region, the vehicle distribution of the entire large region is obtained.

[0016] Furthermore, this application proposes a vehicle location privacy protection system based on local differential privacy, comprising: The region division module is used to divide the target area where the vehicle operates into at least one large region, and then divide each large region into multiple smaller regions. The variance acquisition module is used to obtain the total variance of the number of vehicles in each small area of ​​the large area based on the perturbation flip probability of ε-local differential privacy, and to sum the total variance of the number of vehicles in all small areas of the large area to obtain the comprehensive total variance. The optimal perturbation flip probability acquisition module is used to obtain an optimization function based on the comprehensive total variance, and to solve the optimization function to obtain the optimal perturbation flip probability. The perturbation module is used to perturb the vehicle's position information by using the optimal perturbation flip probability as the perturbation flip probability of ε-local differential privacy, thereby achieving privacy protection of the vehicle's position.

[0017] Optionally, the optimal perturbation flip probability acquisition module obtains an optimization function based on the comprehensive total variance, including: The overall variance is integrated, and the constant coefficients are discarded to obtain the primary optimization function; The correlation between perturbation flip probability and privacy budget in large areas of ε-local differential privacy is obtained based on the unary coding mechanism of ε-local differential privacy; The primary optimization function is simplified using the aforementioned correlation to obtain the optimized function.

[0018] Optionally, the privacy budget for a large area in the ε-local differential privacy within the variance acquisition module is obtained through the following steps: Obtain a score showing the correlation between individual privacy and each sub-region within a larger region; The correlation level scores of each sub-region are summed to obtain the comprehensive correlation level score of the corresponding large region. The privacy budget for a large region is determined based on the comprehensive correlation score.

[0019] Optionally, the optimal perturbation flip probability acquisition module solves the optimization function to obtain the optimal perturbation flip probability, including: The optimal perturbation flip probability is obtained by solving the optimization function using the simulated annealing algorithm, wherein the state variable of the simulated annealing algorithm is the perturbation flip probability, and the optimization objective of the simulated annealing algorithm is to minimize the optimization function.

[0020] Optionally, in the simulated annealing algorithm of the optimal perturbation flip probability acquisition module, the state of the nth iteration is used... It becomes the state of the (n+1)th iteration. The acceptance probability P is:

[0021] in, Let be the value of the optimization function obtained in the nth iteration. Let T be the value of the optimization function obtained in the (n+1)th iteration, and T be the temperature.

[0022] Optionally, the perturbation module uses the optimal perturbation flip probability as the ε-local differential privacy perturbation flip probability to perturb the vehicle's position information, including: The vehicle's location information is unigram encoded to obtain a one-dimensional vector; The optimal perturbation flip probability is used as the perturbation flip probability of ε-local differential privacy to flip the elements in the one-dimensional vector to achieve perturbation of vehicle position information.

[0023] Optionally, the optimal perturbation flip probability in the perturbation module includes a first optimal perturbation flip probability and a second optimal perturbation flip probability. After using the optimal perturbation flip probability as the perturbation flip probability of ε-local differential privacy, the perturbation flip probability of ε-local differential privacy becomes:

[0024]

[0025] in, This represents the first optimal perturbation flip probability. This represents the second-optimal perturbation flip probability. For the first vector A, the first vector is... The possible values ​​of each element, For the perturbation of the first The possible values ​​of each element, Let be a probability function, 0 <q*<p*<1, Budget for privacy.

[0026] Optionally, the disturbance module performs unary encoding on the vehicle's position information to obtain a one-dimensional vector, including: The large regions and the smaller regions within them are numbered sequentially. Obtain the vehicle's GPS location information; Determine the large area and corresponding small area where the GPS location is located; If we assign a value of 1 to the small region where the vehicle is located within the large region, and assign a value of 0 to the other small regions, we obtain a one-dimensional vector.

[0027] Optionally, after the disturbance module disturbs the vehicle's position information based on the optimal disturbance rollover probability, it further includes: The location vectors obtained after disturbance are transmitted to the cloud management center or roadside infrastructure to perform aggregation analysis on the location vectors to obtain the estimated number of vehicles corresponding to each small area; The estimated number of vehicles is corrected by maximum likelihood estimation to obtain an unbiased estimate of the number of vehicles in each small region. Based on the unbiased estimate of the number of vehicles in each small region, the vehicle distribution of the entire large region is obtained.

[0028] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a vehicle location privacy protection method and system based on local differential privacy. The method includes: dividing the target area of ​​vehicle operation into at least one large area, and further dividing each large area into multiple small areas; obtaining the total variance of the estimated number of vehicles in each small area within the large area based on the perturbation and flip probability of ε-local differential privacy, and summing the total variance of the estimated number of vehicles in all small areas within the large area to obtain a comprehensive total variance; obtaining an optimization function based on the comprehensive total variance, and solving the optimization function to obtain the optimal perturbation and flip probability; using the optimal perturbation and flip probability as the perturbation and flip probability of ε-local differential privacy to perturb the vehicle's location information, thereby achieving privacy protection of vehicle location. This application avoids the subjectivity problem caused by manual setting by dividing the target area into large and small areas, constructing an optimization function with flip probability as a parameter based on the small areas, and obtaining the perturbation and flip probability of ε-local differential privacy by solving the optimization function. This achieves an optimal balance between privacy protection and data availability, and improves the accuracy of vehicle distribution assessment. Attached Figure Description

[0029] Figure 1 This is a schematic diagram illustrating the steps of the vehicle location privacy protection method based on local differential privacy proposed in this invention. Figure 2 The present invention proposes Figure 1 A schematic diagram of the step S3 in which the optimization function is obtained based on the comprehensive total variance; Figure 3 The present invention proposes Figure 2 A schematic diagram illustrating the steps involved in obtaining the privacy budget for a large area in ε-local differential privacy within step S32; Figure 4 The present invention proposes Figure 1 A detailed step diagram of step S4; Figure 5 The present invention proposes Figure 4 A detailed step diagram of step S41; Figure 6 This is a further step in the vehicle location privacy protection method based on local differential privacy proposed in this invention; Figure 7This is a schematic diagram illustrating the working principle of the vehicle location privacy protection method based on local differential privacy proposed in this invention for vehicle-road-cloud multi-network convergence. Figure 8 This is a detailed application flowchart of the vehicle location privacy protection method based on local differential privacy proposed in this invention. Figure 9 This is a schematic diagram of the vehicle location privacy protection system based on local differential privacy proposed in this invention. Detailed Implementation

[0030] This invention proposes a vehicle location privacy protection method based on local differential privacy. It performs regional encoding processing on the target area of ​​vehicle operation and optimizes the differential privacy mechanism through simulated annealing algorithm, which solves the problem of vehicle location data privacy. It is especially suitable for effective privacy protection of location data in vehicle-road-cloud multi-network scenarios.

[0031] Example 1: A vehicle location privacy protection method based on local differential privacy, such as Figure 1 As shown, it includes the following steps S1 to S4.

[0032] S1: Divide the target area for vehicle operation into at least one large area, and divide each large area into multiple smaller areas.

[0033] Based on traffic conditions and natural conditions, the target area for vehicle operation is divided into multiple large areas. Based on the granularity required for statistics, the large areas are further divided into multiple small areas.

[0034] Assign unique numbers to the M large regions, such as 1, 2, ..., M. Also number the H small regions within each large region, such as 1, 2, ..., H.

[0035] The corresponding number for each small area and the GPS location of the corresponding area are recorded in the vehicle's onboard computer, cloud management center, and roadside infrastructure.

[0036] S2: Based on -The perturbation flip probability of local differential privacy obtains the total variance of the vehicle count estimate for each sub-region in the large region, and sums the total variance of the vehicle count estimates for all sub-regions in the large region to obtain the comprehensive total variance.

[0037] - The perturbation probability formula for Local Differential Privacy (LDP) is:

[0038]

[0039] in, This represents the first perturbation flip probability. This represents the probability of the second perturbation flipping. For the first vector A, the first vector is... The possible values ​​of each element, For the perturbation of the first The possible values ​​of each element, Let be a probability function, 0 <q<p<1。

[0040] Therefore, small area Total variance of vehicle count estimate for:

[0041] in, This represents the total number of vehicles in the large area. For small areas i The number of vehicles accounts for a certain percentage of the total number of vehicles in the region. The proportion of the total variance of all sub-regions in the large region is then considered. for:

[0042] S3: Obtain the optimization function based on the comprehensive total variance, and solve the optimization function to obtain the optimal perturbation flip probability.

[0043] In further optimized solutions, such as Figure 2 As shown, the process of obtaining the optimization function based on the comprehensive total variance includes steps S31 to S33: S31: Integrate the total variance and discard the constant coefficients to obtain the primary optimization function; because It represents the total number of vehicles within a large region, assuming the region is defined. It is fixed, therefore the primary optimization function for:

[0044] S32: Obtain the correlation between perturbation flip probability and privacy budget based on the ε-local differential privacy unary coding mechanism.

[0045] Because the unary coding mechanism satisfies - The conditions for local differential privacy are:

[0046] That is, obtain and The required association is:

[0047] in, Budget for privacy.

[0048] S33: The primary optimization function is simplified using the aforementioned correlation to obtain the optimization function.

[0049] The above Substitute into the primary optimization function The optimization function is then obtained. for:

[0050] In further optimized solutions, such as Figure 3 As shown, the privacy budget for a large area in ε-local differential privacy described in step 32 is obtained through the following steps S321~S323: S321: Obtain the correlation level score between each sub-region and personal privacy in the large region.

[0051] If there are many sensitive points (such as schools, hospitals, etc.) in a large area, since hospitals reflect an individual's physical condition and schools reflect an individual's family situation, which are closely related to personal privacy, a higher relevance rating should be assigned to the small areas containing these locations.

[0052] S322: Sum the correlation level scores of each sub-region to obtain the comprehensive correlation level score of the corresponding large region.

[0053] The relevance scores for each sub-region are summed, because the more sensitive points related to personal privacy there are in a larger region, the higher the overall relevance score will be.

[0054] S323: Determine the privacy budget for a large area based on the comprehensive correlation score.

[0055] Regions with higher overall correlation scores require stricter privacy protection and therefore need larger privacy budgets.

[0056] In a further preferred embodiment, step S3, which involves solving the optimization function to obtain the optimal perturbation flip probability, includes: The optimal perturbation flip probability is obtained by solving the optimization function using the simulated annealing algorithm, wherein the state variable of the simulated annealing algorithm is the perturbation flip probability, and the optimization objective of the simulated annealing algorithm is to minimize the optimization function.

[0057] The simulated annealing algorithm consists of two parts: the Metropolis algorithm and the annealing process, corresponding to the inner loop and the outer loop, respectively.

[0058] The external circulation is the annealing process, which raises the solid to a high temperature (initial temperature T0), and then lowers the temperature proportionally according to a cooling coefficient α until the final temperature T is reached. f When the cooling process ends, i.e., the annealing process is complete, the initial temperature T0 is 100, the cooling coefficient α is 0.99, and the termination temperature T f It is 0.01.

[0059] The Metropolis algorithm is an inner loop, meaning it iterates L times (e.g., 100 times) at each temperature to find the minimum energy (optimization function) at that temperature (i.e., the corresponding optimal solution). In this embodiment, the perturbation flip probability is used as the state variable, and minimizing the optimization function is the optimization objective. The corresponding simulated annealing algorithm uses the state from the nth iteration... It becomes the state of the (n+1)th iteration. The acceptance probability P is:

[0060] in, Let be the value of the optimization function obtained in the nth iteration. Let T be the value of the optimization function obtained in the (n+1)th iteration, and T be the temperature.

[0061] The optimal perturbation flip probability is obtained based on the above simulated annealing algorithm. and .

[0062] S4: The optimal perturbation flip probability is used as the perturbation flip probability of ε-local differential privacy to perturb the vehicle's position information, thereby achieving privacy protection of the vehicle's position.

[0063] In a further preferred embodiment, the perturbation of the vehicle's position information based on the optimal perturbation rollover probability is, for example... Figure 4 As shown, the process includes the following steps S41 and S42: S41: The vehicle's location information is unigram encoded to obtain a one-dimensional vector.

[0064] In further optimized solutions, such as Figure 5 As shown, step S41 includes the following steps S411~S414, wherein: S411: Number the large area and the smaller areas within the large area sequentially. Number the large area and the smaller areas using the method described in step S1.

[0065] S412: Obtain the vehicle's GPS location information; use a GPS positioning system to track the vehicle's location in real time.

[0066] S413: Determine the large area and the corresponding small area where the GPS location is located.

[0067] The vehicle's onboard computing unit can determine which large area the vehicle's GPS location belongs to and mark the vehicle. This data is made public, thus exposing the vehicle's large area. Then, it can be determined which small area of ​​the large area m the vehicle's location belongs to.

[0068] S414: Assign a value of 1 to the small area where the vehicle is located in the large area, and assign a value of 0 to the other small areas, and you will get a one-dimensional vector.

[0069] If the vehicle is in a small area i The inner part is encoded as a one-dimensional vector A (0, 0, 0, ..., 1, 0), where the first element in vector A is the first element. i One bit is 1, and the others are 0.

[0070] S42: The optimal perturbation flip probability is used as the perturbation flip probability of ε-local differential privacy to flip the elements in the one-dimensional vector to achieve perturbation of the vehicle position information.

[0071] The optimal perturbation flip probability includes a first optimal perturbation flip probability and a second optimal perturbation flip probability. After using the optimal perturbation flip probability as the perturbation flip probability of ε-local differential privacy, the perturbation flip probability of ε-local differential privacy becomes:

[0072]

[0073] in, .

[0074] In further optimized solutions, such as Figure 6 As shown, step S4 is followed by steps S5 and S6: S5: The location vector obtained after disturbance is transmitted to the cloud management center or roadside infrastructure equipment to perform aggregation analysis on the location vector to obtain the estimated number of vehicles corresponding to each small area.

[0075] By counting the total number of vectors in each large region, we can obtain the total number of vehicles Q in the large region, and the th vector of all vectors in each large region. i Number of bits with 1 That is, the corresponding small area i The estimated number of vehicles.

[0076] S6: The estimated number of vehicles is corrected by the maximum likelihood estimation to obtain the unbiased estimate of the number of vehicles in each small area, and the vehicle distribution of the entire large area is obtained based on the unbiased estimate of the number of vehicles in each small area.

[0077] Since the estimated number of vehicles is based on perturbed data, maximum likelihood estimation is used to correct the estimated number of vehicles, thus obtaining an unbiased estimate of the number of vehicles. :

[0078] This allows us to obtain accurate vehicle counts for a large area. and the unbiased estimator for each small region .

[0079] like Figure 7 The diagram illustrates the working principle of the method in this application for vehicle-road-cloud multi-network fusion. First, the GPS location information (x, y) of the vehicle is obtained based on the GPS system. In the vehicle-side computing unit, the GPS location information (x, y) is converted into a one-dimensional vector. Differential privacy processing is performed on this one-dimensional vector to obtain a perturbed vector. The perturbed vector is then transmitted to the cloud management center and roadside infrastructure. Aggregation analysis is performed in the cloud management center and roadside infrastructure to obtain estimated data of the vehicle in each small area, thereby obtaining the vehicle distribution in the entire large area. Information is shared between the cloud management center and roadside infrastructure, and vehicle location data is shared in the vehicle-road-cloud system.

[0080] Specific applications are as follows: Figure 8 As shown, the target area is first divided into large and small regions based on traffic conditions and natural conditions, and these regions are numbered. The privacy budget for each large region is determined based on the number of sensitive locations within that large region. Vehicle location information is unigram-encoded, converting vehicle GPS locations into a data format suitable for differential privacy processing, generating a one-dimensional vector A. The total number of vehicles in all small regions is estimated using a method to determine the optimized function f(q)*. Simulated annealing is used to optimize the parameters of the optimized function, finding the parameter q that minimizes the total error, i.e., the optimization function, which is also the local differential privacy flip probability. A unigram encoding mechanism for local differential privacy is used to perturb each data point in vector A to generate vector B. Vector B, after differential privacy processing, is transmitted to the cloud management center and roadside infrastructure. The cloud management center and roadside infrastructure perform aggregation analysis to obtain more accurate small region data. Information is shared between the cloud center and roadside infrastructure, and vehicle location data is shared within the vehicle-road-cloud system.

[0081] Example 2: Based on the same inventive concept, this invention also provides a vehicle location privacy protection system based on local differential privacy, such as... Figure 9 As shown, it includes: The region division module is used to divide the target area where the vehicle operates into at least one large region, and then divide each large region into multiple smaller regions. The variance acquisition module is used to obtain the total variance of the number of vehicles in each small area of ​​the large area based on the perturbation flip probability of ε-local differential privacy, and to sum the total variance of the number of vehicles in all small areas of the large area to obtain the comprehensive total variance. The optimal perturbation flip probability acquisition module is used to obtain an optimization function based on the comprehensive total variance, and to solve the optimization function to obtain the optimal perturbation flip probability. The perturbation module is used to perturb the vehicle's position information by using the optimal perturbation flip probability as the perturbation flip probability of ε-local differential privacy, thereby achieving privacy protection of the vehicle's position.

[0082] In a further preferred embodiment, the optimal perturbation flip probability acquisition module obtains an optimization function based on the comprehensive total variance, including: The overall variance is integrated, and the constant coefficients are discarded to obtain the primary optimization function; The correlation between perturbation flip probability and privacy budget in large areas of ε-local differential privacy is obtained based on the unary coding mechanism of ε-local differential privacy; The primary optimization function is simplified using the aforementioned correlation to obtain the optimized function.

[0083] In a further preferred embodiment, the privacy budget for the large region in the ε-local differential privacy of the variance acquisition module is obtained through the following steps: Obtain a score showing the correlation between individual privacy and each sub-region within a larger region; The correlation level scores of each sub-region are summed to obtain the comprehensive correlation level score of the corresponding large region. The privacy budget for a large region is determined based on the comprehensive correlation score.

[0084] In a further preferred embodiment, the optimal perturbation flip probability acquisition module solves the optimization function to obtain the optimal perturbation flip probability, including: The optimal perturbation flip probability is obtained by solving the optimization function using the simulated annealing algorithm, wherein the state variable of the simulated annealing algorithm is the perturbation flip probability, and the optimization objective of the simulated annealing algorithm is to minimize the optimization function.

[0085] In a further preferred embodiment, the simulated annealing algorithm in the optimal perturbation flip probability acquisition module is based on the state from the nth iteration. It becomes the state of the (n+1)th iteration. The acceptance probability P is:

[0086] in, Let be the value of the optimization function obtained in the nth iteration. Let T be the value of the optimization function obtained in the (n+1)th iteration, and T be the temperature.

[0087] In a further preferred embodiment, the perturbation module uses the optimal perturbation flip probability as the ε-local differential privacy perturbation flip probability to perturb the vehicle's position information, including: The vehicle's location information is unigram encoded to obtain a one-dimensional vector; The optimal perturbation flip probability is used as the perturbation flip probability of ε-local differential privacy to flip the elements in the one-dimensional vector to achieve perturbation of vehicle position information.

[0088] In a further preferred embodiment, the optimal perturbation flip probability in the perturbation module includes a first optimal perturbation flip probability and a second optimal perturbation flip probability. After using the optimal perturbation flip probability as the perturbation flip probability of ε-local differential privacy, the perturbation flip probability of ε-local differential privacy becomes:

[0089]

[0090] in, This represents the first optimal perturbation flip probability. This represents the second-optimal perturbation flip probability. For the first vector A, the first vector is... The possible values ​​of each element, For the perturbation of the first The possible values ​​of each element, Let be a probability function, 0 <q*<p*<1, Budget for privacy.

[0091] In a further preferred embodiment, the disturbance module performs unary encoding on the vehicle's position information to obtain a one-dimensional vector, including: The large regions and the smaller regions within them are numbered sequentially. Obtain the vehicle's GPS location information; Determine the large area and corresponding small area where the GPS location is located; If we assign a value of 1 to the small region where the vehicle is located within the large region, and assign a value of 0 to the other small regions, we obtain a one-dimensional vector.

[0092] In a further preferred embodiment, after the disturbance module disturbs the vehicle's position information based on the optimal disturbance rollover probability, it further includes: The location vectors obtained after disturbance are transmitted to the cloud management center or roadside infrastructure to perform aggregation analysis on the location vectors to obtain the estimated number of vehicles corresponding to each small area; The estimated number of vehicles is corrected by maximum likelihood estimation to obtain an unbiased estimate of the number of vehicles in each small region. Based on the unbiased estimate of the number of vehicles in each small region, the vehicle distribution of the entire large region is obtained.

[0093] The system takes as input the real-time GPS location of each vehicle within the city area, the city's large-area division, the sub-area division within each large area, and the privacy protection level of each large area. The output is the spatial distribution of the number of vehicles within the city area at various granularities that satisfy local privacy requirements, thus providing the vehicle distribution within the defined sub-areas. In the context of vehicle-road-cloud multi-network convergence, the transmitted data is perturbed to satisfy local differential privacy, resisting various privacy attacks. This achieves both privacy protection for vehicle location data at the vehicle end and enables real-time traffic status perception and monitoring using vehicle location data at the cloud and roadside terminals.

[0094] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A vehicle location privacy protection method based on local differential privacy, characterized in that, include: The target area for vehicle operation is divided into at least one large area, and each large area is further divided into multiple smaller areas. Based on the perturbation flip probability of ε-local differential privacy, the total variance of the number of vehicles in each small area in the large area is estimated, and the total variance of the number of vehicles in all small areas in the large area is summed to obtain the comprehensive total variance. The optimization function is obtained based on the comprehensive total variance, and the optimal perturbation flip probability is obtained by solving the optimization function. The optimal perturbation flip probability is used as the perturbation flip probability of ε-local differential privacy to perturb the vehicle's position information, thereby achieving privacy protection of the vehicle's position. The process of obtaining the optimization function based on the comprehensive total variance includes: The overall variance is integrated, and the constant coefficients are discarded to obtain the primary optimization function; The correlation between perturbation flip probability and privacy budget in large areas of ε-local differential privacy is obtained based on the unary coding mechanism of ε-local differential privacy; The primary optimization function is simplified using the aforementioned correlation to obtain the optimization function; Among them, small areas Total variance of vehicle count estimate for: in, This represents the total number of vehicles in the large area. For small areas i The number of vehicles accounts for a certain percentage of the total number of vehicles in the region. proportion, This represents the first perturbation flip probability. This represents the probability of the second perturbation flipping. The overall variance of all subregions in each large region for: H Number the total number of sub-regions within each large region.

2. The vehicle location privacy protection method based on local differential privacy according to claim 1, characterized in that, The privacy budget for large areas in the ε-local differential privacy is obtained through the following steps: Obtain a score showing the correlation between individual privacy and each sub-region within a larger region; The correlation level scores of each sub-region are summed to obtain the comprehensive correlation level score of the corresponding large region. The privacy budget for a large region is determined based on the comprehensive correlation score.

3. The vehicle location privacy protection method based on local differential privacy according to claim 1, characterized in that, The step of solving the optimization function to obtain the optimal perturbation flip probability includes: The optimal perturbation flip probability is obtained by solving the optimization function using the simulated annealing algorithm, wherein the state variable of the simulated annealing algorithm is the perturbation flip probability, and the optimization objective of the simulated annealing algorithm is to minimize the optimization function.

4. The vehicle location privacy protection method based on local differential privacy according to claim 3, characterized in that, The simulated annealing algorithm is based on the state of the nth iteration. It becomes the state of the (n+1)th iteration. The acceptance probability P is: in, Let be the value of the optimization function obtained in the nth iteration. Let T be the value of the optimization function obtained in the (n+1)th iteration, and let T be the temperature.

5. The vehicle location privacy protection method based on local differential privacy according to claim 1, characterized in that, The step of perturbing the vehicle's position information using the optimal perturbation flip probability as the ε-local differential privacy perturbation flip probability includes: The vehicle's location information is unigram encoded to obtain a one-dimensional vector; The optimal perturbation flip probability is used as the perturbation flip probability of ε-local differential privacy to flip the elements in the one-dimensional vector to achieve perturbation of vehicle position information.

6. The vehicle location privacy protection method based on local differential privacy according to claim 5, characterized in that, The optimal perturbation flip probability includes a first optimal perturbation flip probability and a second optimal perturbation flip probability. After using the optimal perturbation flip probability as the perturbation flip probability of ε-local differential privacy, the perturbation flip probability of ε-local differential privacy becomes: in, This represents the first optimal perturbation flip probability. This represents the second optimal perturbation flip probability. For the first vector A, the first vector is... The possible values ​​of each element, For the perturbation of the first The possible values ​​of each element, Let be a probability function, 0 <q*<p*<1, Budget for privacy.

7. The vehicle location privacy protection method based on local differential privacy according to claim 5, characterized in that, The step of unary encoding the vehicle's location information to obtain a one-dimensional vector includes: The large regions and the smaller regions within them are numbered sequentially. Obtain the vehicle's GPS location information; Determine the large area and corresponding small area where the GPS location is located; If we assign a value of 1 to the small region where the vehicle is located within the large region, and assign a value of 0 to the other small regions, we obtain a one-dimensional vector.

8. The vehicle location privacy protection method based on local differential privacy according to claim 1, characterized in that, The process of perturbing the vehicle's position information based on the optimal perturbation rollover probability further includes: The location vectors obtained after disturbance are transmitted to the cloud management center or roadside infrastructure to perform aggregation analysis on the location vectors to obtain the estimated number of vehicles corresponding to each small area; The estimated number of vehicles is corrected by maximum likelihood estimation to obtain an unbiased estimate of the number of vehicles in each small region. Based on the unbiased estimate of the number of vehicles in each small region, the vehicle distribution of the entire large region is obtained.

9. A vehicle location privacy protection system based on local differential privacy, characterized in that, include: The region division module is used to divide the target area where the vehicle operates into at least one large region, and then divide each large region into multiple smaller regions. The variance acquisition module is used to obtain the total variance of the number of vehicles in each small area of ​​the large area based on the perturbation flip probability of ε-local differential privacy, and to sum the total variance of the number of vehicles in all small areas of the large area to obtain the comprehensive total variance. The optimal perturbation flip probability acquisition module is used to obtain an optimization function based on the comprehensive total variance, and to solve the optimization function to obtain the optimal perturbation flip probability. The perturbation module is used to perturb the vehicle's position information by using the optimal perturbation flip probability as the perturbation flip probability of ε-local differential privacy, thereby achieving privacy protection of the vehicle's position. The process of obtaining the optimization function based on the comprehensive total variance includes: The overall variance is integrated, and the constant coefficients are discarded to obtain the primary optimization function; The correlation between perturbation flip probability and privacy budget in large areas of ε-local differential privacy is obtained based on the unary coding mechanism of ε-local differential privacy; The primary optimization function is simplified using the aforementioned correlation to obtain the optimization function; Among them, small areas Total variance of vehicle count estimate for: in, This represents the total number of vehicles in the large area. For small areas i The number of vehicles accounts for a certain percentage of the total number of vehicles in the region. proportion, This represents the first perturbation flip probability. This represents the probability of the second perturbation flipping. The overall variance of all subregions in each large region for: H Number the total number of sub-regions within each large region.

Images