Trajectory privacy protection method based on sensitive semantic position replacement in connected vehicles

Abstract

The present invention discloses a trajectory privacy protection method based on sensitive semantic position replacement in a connected vehicle network. First, the spatiotemporal trajectory of a vehicle is semantically annotated based on points of interest in a map. Then, sensitive semantic positions are extracted, including a starting position, an end position, a user-defined sensitive semantic position, and a stop position. A double semicircle area is set as a replacement area for each sensitive semantic position. Points of interest PoIs with semantic attributes and query probabilities identical or similar to those of the sensitive semantic position are selected from the replacement area to replace the sensitive semantic position, thereby obtaining a reconstructed trajectory. By replacing sensitive semantic positions in the trajectory with other non-sensitive semantic positions, the present invention not only protects the user's trajectory privacy, but also improves the availability of trajectory data.

Description

Technical Field

[0001] The present invention belongs to the technical field of Internet of Vehicles, and more specifically, relates to a trajectory privacy protection method based on sensitive semantic position replacement in Internet of Vehicles. Background Art

[0002] The Internet of Vehicles (IoV) is an emerging technology in the smart city environment that allows vehicles to exchange data with each other. Through data exchange, IoV promotes the development of many applications aimed at improving driving safety and traffic monitoring. While providing safe travel for vehicles, IoV also collects a large amount of trajectory data of vehicles. According to the different information contained in the trajectory data, it can be divided into two types: spatiotemporal trajectory and semantic trajectory. The spatiotemporal trajectory is mainly composed of the position sequence of the vehicle's driving path collected by the GPS positioning device at a certain time interval or distance interval; while the semantic trajectory is the fusion of spatiotemporal trajectory and semantic information. It not only contains the spatiotemporal position sequence of the vehicle, but also contains the semantic attributes of each position in the geographical environment. Compared with the spatiotemporal trajectory, the semantic trajectory will bring more serious privacy leakage problems to users, because attackers can easily reveal the user's access behavior based on the user's semantic trajectory. At present, the privacy protection of semantic trajectories has gradually become a hot issue in the research of privacy protection of IoV trajectories.

[0003] The trajectory privacy protection method in the Internet of Vehicles should not only ensure that the sensitive information contained in the trajectory itself is not leaked, such as display identifiers and quasi-identifiers, but also prevent attackers from inferring other sensitive information of users based on the context of trajectory data, such as location semantics and social relationships. The trajectory privacy protection method in the Internet of Vehicles mainly crawls the trajectory data of users during driving, including user identity, continuous query location and service requests, and then cleans, filters and classifies these trajectory data to achieve user privacy protection. Protecting the user's trajectory privacy can reduce the leakage of users' personal privacy and increase users' stickiness to the Internet of Vehicles on the one hand, and ensure users' travel safety on the other hand, thereby reducing property losses and casualties caused by traffic accidents.

[0004] The current methods for protecting the privacy of semantic trajectories mainly include semantic confusion strategy, differential privacy strategy and semantic trajectory clustering strategy. Among them, the main idea of ​​the semantic confusion strategy is to protect the privacy of the user's semantic trajectory by replacing the sensitive semantic positions in the user's trajectory with other non-sensitive semantic positions. It is not only simple to implement, but also can improve the availability of trajectory data. However, this strategy ignores the user's personalized privacy protection needs and the query probability of semantic positions in different time periods during the replacement process of sensitive semantic positions, resulting in an inability to effectively balance the availability of trajectory data and the degree of privacy protection. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a trajectory privacy protection method based on sensitive semantic position replacement in the Internet of Vehicles. By replacing the sensitive semantic positions in the trajectory with other non-sensitive semantic positions, not only the user's trajectory privacy is protected, but also the availability of trajectory data is improved.

[0006] In order to achieve the above-mentioned object of the invention, the trajectory privacy protection method based on sensitive semantic position replacement in the Internet of Vehicles of the present invention comprises the following steps:

[0007] S1: Let the vehicle space-time trajectory T = {(x1,y1,t1),(x2,y2,t2),…,(x N ,y N ,t N )}, where (x i ,y i ,t i ) represents the sampling position in the vehicle's spatiotemporal trajectory, t i represents the sampling time, (x i ,y i ) represents the sampling time t i The geographic coordinates at the time, 1≤i≤N, N represents the number of sampling times in the vehicle's spatiotemporal trajectory;

[0008] For each sampling position (x i ,y i ,t i ), search for the nearest point of interest (PoI) from the map, and extract the semantic attribute s of the PoI i And assign it to the sampling position (x i ,y i ,t i ), so the sampling position is marked as the semantic position (x i ,y i ,t i ,s i ), vehicle space-time trajectory T i Annotated as semantic track ST = {(x1,y1,t1,s1),(x2,y2,t2,s2),...,(x n ,y n ,t n ,s n )};

[0009] S2: The vehicle user pre-sets a sensitive semantic position set x, and matches each semantic position in the semantic track with the sensitive semantic position set one by one. If the semantic position in the semantic track ST successfully matches the sensitive semantic position in the sensitive semantic position set x, the semantic position is marked as a sensitive semantic position, otherwise no operation is performed; in addition, the starting position and the end position in the semantic track ST are marked as sensitive semantic positions;

[0010] S3: Pre-set the time threshold T according to actual needs threh and the spatial threshold S threh , if there is a sub-trajectory consisting of several consecutive sampling positions in the vehicle's spatiotemporal trajectory T, the sub-trajectory satisfies that the time interval between adjacent sampling positions is greater than T threh , the distance interval is less than S threh , then this sub-trajectory is used as a sensitive stop sub-trajectory, otherwise no operation is performed; for the sensitive stop sub-trajectory, the following formula is used to calculate the corresponding stop position coordinate (x stay ,y stay ):

[0011]

[0012] Among them, i h represents the original serial number of the hth sampling position in the sensitive stay sub-trajectory, represents the geographical coordinates of the hth sampling position in the sensitive stay sub-trajectory, h=1,2,...,H, H represents the number of sampling positions in the sensitive stay sub-trajectory;

[0013] The geographic coordinates (x stay ,y stay ) in the geographic map as the semantic attributes of the stop location;

[0014] In the semantic track ST, the geographical coordinates and semantic attributes of the stay location are used to replace the corresponding sensitive stay sub-track, thereby obtaining a new semantic track ST′, and the stay location is marked as a sensitive semantic location in the semantic track ST′;

[0015] S4: traverse each sensitive semantic position in the semantic track ST′ in turn. If the sensitive semantic position is the starting point of the track, a circular area with the starting point of the track as the center and half the distance between the starting point of the track and the next sampling position as the radius is used as the replacement area;

[0016] If the sensitive semantic position is the end point of the trajectory, a circular area with the end point of the trajectory as the center and half the distance between the end point of the trajectory and the previous sampling position as the radius is used as the replacement area;

[0017] For other sensitive semantic positions, the straight line perpendicular to the vehicle speed at the sensitive semantic position is used as the dividing line of the semicircle. The sensitive semantic position is used as the center of the circle. The radius of the semicircle close to the previous sampling position is half of the distance between the sensitive semantic position and the previous sampling position, and the radius of the semicircle close to the next sampling position is half of the distance between the sensitive semantic position and the next sampling position. The double semicircle area obtained by the above method is used as the replacement area.

[0018] For each replacement area of ​​a sensitive semantic position, all PoIs in the replacement area are used to form a set of candidate interest points;

[0019] S5: The sensitive semantic position is sl k =(x k ,y k ,t k ,s k ), t k Indicates sensitive semantic location sl k The sampling time (x k ,y k ) indicates the sensitive semantic position sl k The geographical coordinates, s k Indicates sensitive semantic location sl k The semantic attributes of k = 1, 2, ..., K, K represents the number of sensitive semantic positions in the trajectory; the sensitive semantic position sl k The set of candidate interest points is P k , where the number of PoIs is D k ;

[0020] Get sensitive semantic location sl from the Internet of Vehicles system k and the set of candidate interest points P k Inside D k PoI at sampling time t k The query times of all positions in the replacement area are obtained, and the total query times of all positions in the replacement area are obtained to calculate the sensitive semantic position sl k Probability of Query k , and the set of candidate interest points P k The query probability of each PoI in Pro PoI,d , d=1,2,...,D k ;

[0021] The hierarchical structure of the location semantic tree is set in advance according to the division of semantic attributes, and then the sensitive semantic locations and the candidate interest point set P k Inside D k The PoIs are taken as leaf nodes and added to the corresponding nodes in the position semantic tree according to the semantic attributes to obtain the position semantic tree of the sensitive semantic position replacement area;

[0022] For sensitive semantic location sl k , and the set of candidate interest points P is calculated using the following formula k The similarity of each PoI in Sem_Similarity(sl k ,sl PoI,d ):

[0023] Sem_Similarity(sl k ,sl PoI,d )=γ×Sem(sl k ,sl PoI,d )+(1-γ)×Pro(sl k ,sl PoI,d )

[0024]

[0025] Among them, Sem(sl k ,sl PoI,d ) indicates the sensitive semantic position sl k and the set of candidate interest points P k The dth point of interest sl PoI,d The semantic consistency of Represents the sensitive semantic position sl in the position semantic tree k and the set of candidate interest points P k The dth point of interest sl PoI,d The number of hops between max Indicates the maximum number of hops between two leaf nodes in the location semantics tree, Pro(sl k ,sl PoI,d ) indicates the sensitive semantic position sl k and the set of candidate interest points P k The dth point of interest sl PoI,d Probability consistency of query, Pro max Indicates the replacement area t i The maximum value of the difference between the two location query probabilities at the time, γ represents the preset semantic similarity coefficient;

[0026] The sensitive semantic position sl k Replace the D in the area k PoIs are classified according to the sensitive semantic location sl k The similarities are sorted from large to small, and then the feasibility of replacement is judged in turn. Once there is a feasibility of replacement, the PoI is used as the sensitive semantic position sl k The replacement interest point, the feasibility of replacement is determined by: replacing the sensitive semantic position sl in the semantic trajectory ST′ kThe last sampling position of the object is connected to the current PoI to obtain a line, and it is determined whether the line passes through the preset obstacle on the map. If so, there is no replacement feasibility, otherwise there is replacement feasibility;

[0027] S6: Replace each sensitive semantic position in the semantic trajectory ST′ with the corresponding replacement interest point, and then extract the vehicle spatiotemporal trajectory from the generated semantic trajectory as the reconstructed trajectory of the vehicle spatiotemporal trajectory T.

[0028] The trajectory privacy protection method based on sensitive semantic position replacement in the Internet of Vehicles of the present invention first performs trajectory semantic annotation on the vehicle's spatiotemporal trajectory based on points of interest in a map, and then extracts sensitive semantic positions, including a starting position, an end position, a user-defined sensitive semantic position, and a stop position. A double semicircle area is set for each sensitive semantic position as a replacement area, and PoIs with semantic attributes and query probabilities that are the same or similar to those of the sensitive semantic position are screened out from the replacement area to replace the sensitive semantic position, thereby obtaining a reconstructed trajectory.

[0029] The present invention can effectively protect the sensitive information in the vehicle's spatiotemporal trajectory by screening sensitive semantic positions that contain more semantic information than ordinary sampling positions for replacement, thereby effectively protecting the trajectory privacy; when screening and replacing points of interest, semantic consistency, query probability consistency and replacement feasibility are comprehensively considered, thereby improving the data availability of the reconstructed trajectory. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a flowchart of a specific implementation method of a trajectory privacy protection method based on sensitive semantic position replacement in a connected vehicle network of the present invention;

[0031] Figure 2 is an example diagram of the implementation process of the present invention in this embodiment;

[0032] Figure 3 is a trend diagram of the average recognition rate of the present invention and two comparative methods as the number of PoIs changes;

[0033] Figure 4 It is a trend chart showing the average recognition rate of the present invention and two comparative methods as the number of sensitive semantic positions changes. DETAILED DESCRIPTION

[0034] The specific implementation of the present invention is described below in conjunction with the accompanying drawings so that those skilled in the art can better understand the present invention. It should be noted that in the following description, when the detailed description of known functions and designs may dilute the main content of the present invention, these descriptions will be omitted here.

[0035] Example

[0036] Figure 1 This is a flowchart of a specific implementation of the trajectory privacy protection method based on sensitive semantic position replacement in the Internet of Vehicles of the present invention. Figure 1 As shown, the specific steps of the trajectory privacy protection method based on sensitive semantic position replacement in the Internet of Vehicles of the present invention include:

[0037] S101: Trajectory semantic annotation:

[0038] In order to achieve semantic annotation of all sampling positions in the trajectory, it is necessary to find the PoI closest to the trajectory sampling position and assign the semantic attribute of the PoI to the semantic position, where the PoI can usually be obtained from a public map through an application program interface. The specific method of trajectory semantic annotation in the present invention is:

[0039] Let the vehicle space-time trajectory T = {(x1,y1,t1),(x2,y2,t2),...,(x N ,y N ,t N )}, where (x i ,y i ,t i ) represents the sampling position in the vehicle's spatiotemporal trajectory, t i represents the sampling time, (x i ,y i ) represents the sampling time t i The geographic coordinates at the time, 1≤i≤N, N represents the number of vehicle spatiotemporal trajectory sampling times.

[0040] For each sampling position (x i ,y i ,t i ), search the map for the PoI closest to it, and extract the semantic attribute s of the PoI i And assign it to the sampling position (x i ,y i ,t i ), so the sampling position is marked as the semantic position (x i ,y i ,t i ,s i ), vehicle space-time trajectory T i Annotated as semantic track ST = {(x1,y1,t1,s1),(x2,y2,t2,s2),...,(x n ,y n ,t n ,s n )}.

[0041] S102: Sensitive semantic location identification:

[0042] The vehicle user pre-sets a sensitive position set χ, and matches each semantic position in the semantic trajectory with the sensitive semantic position set one by one. If the semantic position in the semantic trajectory ST successfully matches the sensitive position in the sensitive semantic position set χ, the semantic position is marked as a sensitive semantic position, otherwise no operation is performed.

[0043] In addition, the starting position and the ending position in the semantic track ST are marked as sensitive semantic positions.

[0044] S103: Stay position identification:

[0045] In addition to the trajectory starting position, trajectory ending position, and sensitive positions set by the vehicle user, there is also a sensitive semantic position in the vehicle's spatiotemporal trajectory, namely the stop position. The stop position needs to be extracted from the high-density sampling points of the vehicle's spatiotemporal trajectory data. This is because when vehicles stop or wander in a certain area, their speed usually approaches 0, and they will generate a higher density of sampling positions. Therefore, the present invention uses a stop position extraction algorithm based on sampling density to extract the stop position in the vehicle's spatiotemporal trajectory. The specific method is as follows:

[0046] Set the time threshold T in advance according to actual needs threh and the spatial threshold S threh , if there are several sub-trajectories consisting of consecutive sampling positions in the vehicle's spatiotemporal trajectory T, the sub-trajectory satisfies that the time interval between adjacent sampling positions is greater than T threh , the distance interval is less than S threh , then this sub-trajectory is used as a sensitive stop sub-trajectory, otherwise no operation is performed. For the sensitive stop sub-trajectory, the following formula is used to calculate the corresponding stop position coordinate (x stay ,y stay ):

[0047]

[0048] Among them, i h represents the original serial number of the hth sampling position in the sensitive stay sub-trajectory, represents the geographical coordinates of the hth sampling position in the sensitive stay sub-trajectory, h=1, 2, ..., H, and H represents the number of sampling positions in the sensitive stay sub-trajectory.

[0049] The geographic coordinates (x stay ,y stay ) in the geographic map as the semantic attributes of the stay location.

[0050] In the semantic track ST, the geographical coordinates and semantic attributes of the stay location are used to replace the corresponding sensitive stay sub-track, thereby obtaining a new semantic track ST′, and the stay location is marked as a sensitive semantic location in the semantic track ST′.

[0051] S104: Sensitive semantic position replacement area determination:

[0052] After the sensitive semantic position extraction is completed, it is necessary to determine the replacement area of ​​the sensitive semantic position and the selection strategy of the PoI in the area. In order to maximize the similarity between the reconstructed trajectory and the original trajectory shape, the present invention takes into account the speed of the vehicle in the process of determining the replacement area of ​​the sensitive semantic position. This is because the distance between the continuous sampling points of the vehicle is proportional to the speed of the vehicle (the sampling interval remains unchanged). Considering the influence of the vehicle speed on the sampling distance, the Euclidean distance between the sensitive semantic position and the sampling positions before and after it is used to construct the replacement area of ​​the sensitive semantic position. In addition, in the process of replacing the sensitive semantic position, it is also necessary to avoid the reverse mutation attack caused by the overlap of the replacement area of ​​the sensitive semantic position. Therefore, in the process of determining the replacement area of ​​the sensitive semantic position, the present invention needs to consider both the user speed and the reverse mutation. In the process of determining the replacement area of ​​the sensitive semantic position, the double semicircle area can well avoid the appearance of overlapping areas, so the double semicircle is used in the present invention to constitute the replacement area. In summary, the specific setting method of the replacement area of ​​the sensitive semantic position in the present invention is as follows:

[0053] Each sensitive semantic position in the semantic track ST′ is traversed in turn. If the sensitive semantic position is the starting point of the track, a circular area with the starting point of the track as the center and half the distance between the starting point of the track and the next sampling position as the radius is used as the replacement area.

[0054] If the sensitive semantic position is the end point of the trajectory, a circular area with the end point of the trajectory as the center and half the distance between the end point of the trajectory and the previous sampling position as the radius is used as the replacement area.

[0055] For other sensitive semantic positions, in order to ensure that the replacement areas of adjacent sensitive semantic positions do not overlap, it is first necessary to determine the dividing line of adjacent semicircles, that is, the straight line perpendicular to the vehicle speed at the sensitive semantic position is used as the dividing line of the semicircle, with the sensitive semantic position as the center of the circle, the radius of the semicircle close to the previous sampling position is half of the distance between the sensitive semantic position and the previous sampling position, and the radius of the semicircle close to the next sampling position is half of the distance between the sensitive semantic position and the next sampling position. The double semicircle area obtained by the above method is used as the replacement area.

[0056] For each replacement region of a sensitive semantic position, the PoIs in the replacement region are used to form a set of candidate interest points.

[0057] S105: Sensitive semantic position replacement interest point determination:

[0058] After the replacement areas of all sensitive semantic positions in the vehicle's spatiotemporal trajectory are determined, it is necessary to select a suitable PoI from the set of candidate points of interest for each sensitive semantic position to replace the sensitive semantic position. For the selection of PoI, it is necessary to consider the consistency of the semantic attributes and query probabilities of the sensitive semantic position and the PoI in the replacement area at the same time. This is because the availability of trajectory data can be improved by considering the consistency of the semantic attributes of the sensitive semantic position and the PoI in the replacement area, and the consistency of the query probability of the sensitive semantic position and the PoI in the replacement area can effectively prevent query probability attacks. For example, if the sensitive semantic position in the vehicle trajectory is an oceanarium, and the PoIs with the same or similar semantic attributes in the replacement area of ​​the sensitive semantic position include bars, KTVs, etc. Although these semantic locations are all entertainment venues, the query probability of the oceanarium during the day is significantly higher than that of the other two locations. If a bar is used to replace the oceanarium, the attacker will infer the user's sensitive semantic position based on the query probability of the bar during the day. Therefore, the present invention selects a PoI with the same or similar semantic attributes and query probability as the sensitive semantic position as a replacement point of interest for sensitive semantic position replacement, and the specific method is as follows:

[0059] The sensitive semantic position is sl k =(x k ,y k ,t k ,s k ), t k Indicates sensitive semantic location sl k The sampling time (x k ,y k ) indicates the sensitive semantic position sl k The geographical coordinates, s k Indicates sensitive semantic location sl k The semantic attributes of k = 1, 2, ..., K, K represents the number of sensitive semantic positions. k The set of candidate interest points is P k , where the number of PoIs is D k .

[0060] Get sensitive semantic location sl from the Internet of Vehicles system k and the set of candidate interest points P k Inside D k PoI at sampling time t k The query times of all positions in the replacement area are obtained, and the total query times of all positions in the replacement area are obtained to calculate the sensitive semantic position sl k Probability of Query k , and the set of candidate interest points P k The query probability of each PoI in Pro PoI,d , d=1,2,...,Dk .

[0061] The hierarchical structure of the location semantic tree is set in advance according to the division of semantic attributes, and then the sensitive semantic locations and the candidate interest point set P k Inside D k A PoI is taken as a leaf node and added to the corresponding node in the position semantic tree according to the semantic attribute to obtain the position semantic tree of the sensitive semantic position replacement area.

[0062] For sensitive semantic location sl k , and the set of candidate interest points P is calculated using the following formula k The similarity of each PoI in Sem_Similarity(sl k ,sl PoI,d ):

[0063] Sem_Similarity(sl k ,sl PoI,d )=γ×Sem(sl k ,sl PoI,d )+(1-γ)×Pro(sl k ,sl PoI,d )

[0064]

[0065] Among them, Sem(sl k ,sl PoI,d ) indicates the sensitive semantic position sl k and the set of candidate interest points P k The dth point of interest sl PoI,d The semantic consistency of Represents the sensitive semantic position sl in the position semantic tree k and the set of candidate interest points P k The dth point of interest sl PoI,d The number of hops between max Indicates the maximum number of hops between two leaf nodes in the location semantics tree, Pro(sl k ,sl PoI,d ) indicates the sensitive semantic position sl k and the set of candidate interest points P k The dth point of interest sl PoI,d Probability consistency of query, Pro max Indicates the replacement area t i The maximum value of the difference between the two location query probabilities at the time instant, γ represents the preset semantic similarity coefficient.

[0066] There are many obstacles in geographic space, such as mountains and rivers. Attackers can prune user trajectories based on these obstacles, leaving more meaningful locations. Therefore, in the process of trajectory reconstruction, it is necessary to judge the feasibility of replacing points of interest in the replacement area. The specific method is: replace the sensitive semantic position sl k Replace the D in the area k PoIs are classified according to the sensitive semantic location sl k The similarities are sorted from large to small, and then the feasibility of replacement is judged in turn. Once there is a feasibility of replacement, the PoI is used as the sensitive semantic position sl k The replacement interest point, the feasibility of replacement is determined by: replacing the sensitive semantic position sl in the semantic trajectory ST′ k The last sampling position of the current PoI is connected to obtain a line, and it is determined whether the line passes through the preset obstacles on the map (usually mountains, rivers, etc.). If so, there is no replacement feasibility, otherwise there is replacement feasibility.

[0067] By judging the feasibility of replacement, it effectively avoids the attacker from inferring the replacement of sensitive semantic positions based on the trajectory fragments passing through obstacles, thereby improving the trajectory privacy protection effect of this method.

[0068] S106: Trajectory reconstruction:

[0069] Each sensitive semantic position in the semantic trajectory ST′ is replaced with the corresponding replacement interest point, and then the vehicle spatiotemporal trajectory is extracted from the generated semantic trajectory as the reconstructed trajectory of the vehicle spatiotemporal trajectory T.

[0070] Figure 2 2 is an example diagram of the implementation process of the present invention in this embodiment. Figure 2 As shown, the vehicle's spatiotemporal trajectory is first semantically annotated, and then the sensitive semantic position is identified. In this embodiment, the vehicle user does not pre-set the sensitive position, so only the starting point and end point of the vehicle's spatiotemporal trajectory are marked as sensitive semantic positions. Then the stop position is identified, the calculated stop position is used to replace the sensitive stop sub-trajectory, and the stop position is marked as a sensitive semantic position. For each sensitive semantic position, its replacement area is determined, and then the replacement interest point is determined within the replacement area. Each sensitive semantic position is replaced with the corresponding replacement interest point, and then the vehicle's spatiotemporal trajectory is extracted from the generated semantic trajectory to obtain a reconstructed trajectory of the vehicle's spatiotemporal trajectory. By comparing the original vehicle spatiotemporal trajectory and the reconstructed trajectory, it can be seen that the reconstructed trajectory generated by the method of the present invention can effectively hide the trajectory information.

[0071] In order to better illustrate the technical effect of the present invention, a specific example is used to perform simulation verification on the present invention.

[0072] In the simulation verification process of this embodiment, the SSAR method (reference “Ji, Y, Gui, X, Dai, H, et al. Trajectory Privacy Protection Based on Sensitive Stay Area Replacement in Publishing [J]. Mathematical Problems in Engineering, 2022”) and the PVSP method (reference “Ye A, Zhang Q, Diao Y, et al. A Semantic-Based Approach for Privacy-Preserving in Trajectory Publishing [J]. IEEE Access, 2020, 8: 184965-184975.”) are used as comparison methods, and the average recognition rate is selected as the performance parameter for comparison with the present invention. The average recognition rate refers to the probability that the replacement position in the reconstructed trajectory is re-identified by the attacker, so it can be used to measure the degree of protection of the vehicle semantic trajectory privacy by this method.

[0073] Figure 3 : is a trend chart showing the average recognition rate of the present invention and the two comparison methods as the number of PoIs changes. Figure 3 As shown, when the number of PoIs increases, the average recognition rates of the three methods gradually decrease, and the average recognition rate of the present invention is much lower than that of the other two methods. Figure 4 is a trend chart showing the average recognition rate of the present invention and the two comparison methods as the number of sensitive semantic positions changes. Figure 4 As shown, when the sensitive semantic positions in the trajectory continue to increase, the average recognition rates of the three methods gradually increase, while the average recognition rate of the present invention is much lower than that of the other two methods; it can be seen that the present invention is significantly better than the two comparison methods in terms of the anonymity effect of the semantic trajectory.

[0074] Although the above describes the illustrative specific embodiments of the present invention to facilitate those skilled in the art to understand the present invention, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations using the concept of the present invention are protected.

Claims

1. A trajectory privacy protection method based on sensitive semantic position replacement in an Internet of Vehicles, characterized in that: The following steps are involved: S1: Let the vehicle space-time trajectory T = {(x1,y1,t1),(x2,y2,t2),…,(x N ,y N ,t N )}, where (x i ,y i ,t i ) represents the sampling position in the vehicle's spatiotemporal trajectory, t i represents the sampling time, (x i ,y i ) represents the sampling time t i The geographic coordinates at the time, 1≤i≤N, N represents the number of sampling times in the vehicle's spatiotemporal trajectory; For each sampling position (x i ,y i ,t i ), search the nearest point of interest (PoI) from the map, and extract the semantic attribute s of the PoI i And assign it to the sampling position (x i ,y i ,t i ), so the sampling position is marked as the semantic position (x i ,y i ,t i ,s i ), vehicle space-time trajectory T i Annotated as semantic track ST = {(x1, y1, t1, s1), (x2, y2, t2, s2), …, (x n ,y n ,t n ,s n )}; S2: The vehicle user pre-sets a sensitive semantic position set x, and matches each semantic position in the semantic track with the sensitive semantic position set one by one. If the semantic position in the semantic track ST successfully matches the sensitive semantic position in the sensitive semantic position set x, the semantic position is marked as a sensitive semantic position, otherwise no operation is performed; in addition, the starting position and the end position in the semantic track ST are marked as sensitive semantic positions; S3: Pre-set the time threshold T according to actual needs threh and the spatial threshold S threh , if there is a sub-trajectory consisting of several consecutive sampling positions in the vehicle's spatiotemporal trajectory T, the sub-trajectory satisfies that the time interval between adjacent sampling positions is greater than T threh , the distance interval is less than S threh , then this sub-trajectory is used as a sensitive stop sub-trajectory, otherwise no operation is performed; for the sensitive stop sub-trajectory, the following formula is used to calculate the corresponding stop position coordinate (x stay ,y stay ): Among them, i h represents the original serial number of the hth sampling position in the sensitive stay sub-trajectory, represents the geographical coordinates of the hth sampling position in the sensitive stay sub-trajectory, h=1,2,…,H, H represents the number of sampling positions in the sensitive stay sub-trajectory; The geographic coordinates (x stay ,y stay ) in the geographic map as the semantic attributes of the stop location; In the semantic track ST, the geographical coordinates and semantic attributes of the stay location are used to replace the corresponding sensitive stay sub-track, thereby obtaining a new semantic track ST′, and the stay location is marked as a sensitive semantic location in the semantic track ST′; S4: traverse each sensitive semantic position in the semantic track ST′ in turn. If the sensitive semantic position is the starting point of the track, a circular area with the starting point of the track as the center and half the distance between the starting point of the track and the next sampling position as the radius is used as the replacement area; If the sensitive semantic position is the end point of the trajectory, a circular area with the end point of the trajectory as the center and half the distance between the end point of the trajectory and the previous sampling position as the radius is used as the replacement area; For other sensitive semantic positions, the straight line perpendicular to the vehicle speed at the sensitive semantic position is used as the dividing line of the semicircle. The sensitive semantic position is used as the center of the circle. The radius of the semicircle close to the previous sampling position is half of the distance between the sensitive semantic position and the previous sampling position, and the radius of the semicircle close to the next sampling position is half of the distance between the sensitive semantic position and the next sampling position. The double semicircle area obtained by the above method is used as the replacement area. For each replacement area of ​​a sensitive semantic position, all PoIs in the replacement area are used to form a set of candidate interest points; S5: The sensitive semantic position is sl k =(x k ,y k ,t k ,s k ), t k Indicates sensitive semantic location sl k The sampling time (x k ,y k ) indicates the sensitive semantic position sl k The geographical coordinates, s k Indicates sensitive semantic location sl k The semantic attributes of k = 1, 2, ..., K, K represents the number of sensitive semantic positions in the trajectory; the sensitive semantic position sl k The set of candidate interest points is P k , where the number of PoIs is D k ; Get sensitive semantic location sl from the Internet of Vehicles system k and the set of candidate interest points P k Inside D k PoI at sampling time t k The query times of all positions in the replacement area are obtained, and the total query times of all positions in the replacement area are obtained to calculate the sensitive semantic position sl k Probability of Query k , and the set of candidate interest points P k The query probability of each PoI in Pro PoI,d , d=1,2,...,D k ; The hierarchical structure of the location semantic tree is set in advance according to the division of semantic attributes, and then the sensitive semantic locations and the candidate interest point set P k Inside D k The PoIs are taken as leaf nodes and added to the corresponding nodes in the position semantic tree according to the semantic attributes to obtain the position semantic tree of the sensitive semantic position replacement area; For sensitive semantic location sl k , and the set of candidate interest points P is calculated using the following formula k The similarity of each PoI in Sem_Similarity(sl k ,sl PoI,d ): Sem_Similarity(sl k ,sl PoI,d )=γ×Sem(sl k ,sl PoI,d )+(1-γ)×Pro(sl k ,sl PoI,d ) Among them, Sem(sl k ,sl PoI,d ) indicates the sensitive semantic position sl k and the set of candidate interest points P k The dth point of interest sl PoI,d The semantic consistency of Represents the sensitive semantic position sl in the position semantic tree k and the set of candidate interest points P k The dth point of interest sl PoI,d The number of hops between max Indicates the maximum number of hops between two leaf nodes in the location semantics tree, Pro(sl k ,sl PoI,d ) indicates the sensitive semantic position sl k and the set of candidate interest points P k The dth point of interest sl PoI,d Probability consistency of query, Pro max Indicates the replacement area t i The maximum value of the difference between the two location query probabilities at the time, γ represents the preset semantic similarity coefficient; The sensitive semantic position sl k Replace the D in the area k PoIs are classified according to the sensitive semantic location sl k The similarities are sorted from large to small, and then the feasibility of replacement is judged in turn. Once there is a feasibility of replacement, the PoI is used as the sensitive semantic position sl k The replacement interest point, the feasibility of replacement is determined by: replacing the sensitive semantic position sl in the semantic trajectory ST′ k The last sampling position of the object is connected to the current PoI to obtain a line, and it is determined whether the line passes through the preset obstacle on the map. If so, there is no replacement feasibility, otherwise there is replacement feasibility; S6: Replace each sensitive semantic position in the semantic trajectory ST′ with the corresponding replacement interest point, and then extract the vehicle spatiotemporal trajectory from the generated semantic trajectory as the reconstructed trajectory of the vehicle spatiotemporal trajectory T.

Images