Semantic-Sensitive Location Trajectory Privacy Protection Method, System and Medium for Trajectory-Published Datasets

Abstract

The present invention discloses a semantic-sensitive location trajectory privacy protection method, system and medium for a trajectory publishing data set. The method includes searching for stop points of the user's initial trajectory T 0 to obtain stop points; obtaining POIs by using the category-distance priority method or the Markov matrix method; using the obtained POIs to replace the stop points to obtain a confused trajectory T 1 ; taking the confused trajectory T 1 as input, and obtaining an intermediate trajectory T d by using an intermediate point generation algorithm; taking the initial trajectory T 0 and the intermediate trajectory T d as input, and obtaining an anonymous trajectory T dummy that protects semantic information by using a trajectory correction algorithm. The present invention adopts stop point confusion technology, the category-distance priority method and the Markov matrix, confuses the user's trajectory by using the existing POI information and stop point search technology, and selects the best POI for replacing the stop point according to category and probability judgment, reduces the utility loss of the trajectory after anonymization, and realizes semantic-oriented trajectory privacy protection.

Description

Technical Field

[0001] The present invention belongs to the technical field of trajectory privacy protection, and particularly relates to a semantic-sensitive location trajectory privacy protection method, system and medium for a trajectory publishing dataset. Background Art

[0002] k-anonymity is a widely used privacy protection model. This model ensures that the unique information of data cannot be distinguished from at least other k-1 data information. In trajectory privacy protection, the k-anonymity method synthesizes new anonymous trajectories from several similar trajectories. Such methods usually involve clustering techniques and require a large amount of computing resources, so it is difficult to run on large datasets, and at the same time, the semantic information carried by the trajectory data is ignored. Compared with the k-anonymity method, the virtual-based privacy protection method requires less computing resources and can be applied to sparse or dense trajectory datasets. Through such algorithms, data providers can generate a large number of virtual trajectories to protect users' real data, which can meet the privacy requirements of many application scenarios, but most do not consider the semantic information of the trajectories, so anonymization will result in significant utility loss. Summary of the Invention

[0003] The main purpose of the present invention is to overcome the deficiencies of the prior art and provide a semantic-sensitive location trajectory privacy protection method, system and medium for a trajectory publishing dataset. Users can protect the privacy of their locations according to their own privacy needs, thereby preventing attacks from malicious users. The present invention does not require a large amount of computing resources, and at the same time can retain the semantic information of the location to the greatest extent, thereby minimizing the location utility loss after privacy protection.

[0004] To achieve the above object, the present invention adopts the following technical solutions:

[0005] One aspect of the present invention provides a semantic-sensitive location trajectory privacy protection method for a trajectory publishing dataset, including the following steps:

[0006] Search for stop points for the user's initial trajectory T 0 to obtain stop points;

[0007] Obtain POIs by using the category distance priority method or the Markov matrix method;

[0008] Replace the stop points with the obtained POIs to obtain a confused trajectory T 1 ;

[0009] Taking the confused trajectory T 1 as input, use the midpoint generation algorithm to obtain an intermediate trajectory T d ;

[0010] Taking the initial trajectory T0 and the intermediate trajectory T d As input, an anonymous trajectory T that protects semantic information is obtained using a trajectory correction algorithm dummy .

[0011] As a preferred technical solution, the initial trajectory T of the user 0 is searched for stop points, specifically:

[0012] The initial trajectory T 0 is expressed as: T 0 ={(x 1 , y 1 , t 1 ), (x 2 , y 2 , t 2 ), …, (x n , y n , t n ), U i}; where (x n , y n ) is a positioning point, t n is a timestamp, and U i is a user identifier;

[0013] Set a distance threshold distThreh and a time threshold timeThreh;

[0014] For any two adjacent coordinate points (x i , y i , t i ) and (x j , y j , t j ), if the distance between these two coordinate points is less than the set distance threshold distThreh, or the time interval between the two coordinate points is greater than the set time threshold timeThreh, then (x i , y i , t i ) is marked as a stop point.

[0015] As a preferred technical solution, the POI is obtained using the category distance priority method, specifically:

[0016] Determine the initial radius parameter r and the POI category level K;

[0017] Set a POI retrieval area with the initial stop point p as the center and r as the radius, and search for all POIs within the circular POI retrieval area from the LBS database;

[0018] Sort the searched POIs according to the category distance. Among them, the POIs belonging to the same category as the stop point p have the highest priority as replacement points. If they are of the same category as the stop point p, the POIs with the smaller distance have higher priority.

[0019] If candidates of the same category level K are searched, use them as the output results.

[0020] If no candidates of the same category level K are found within the POI retrieval area, increase the radius parameter r and search again.

[0021] When r ≥ the set maximum radius r_max and the search result is empty, select the POI with the smallest distance from the stop point p as the output result.

[0022] As a preferred technical solution, the POIs are obtained by using the Markov matrix method, specifically:

[0023] Obtain the POI category conversion matrix of the Geolife dataset. Each row of the POI category conversion matrix is the probability from the category at the current position to other categories at the next position.

[0024] Find the category with the highest transition probability and search for the next POI from the target category. Among them, the probability M of transferring from position i to position j ij is specifically as follows:

[0025]

[0026] where Pr(·) is the Markov probability transfer function, dst is the target position, src is the source position, C i and C j represent two stop points.

[0027] Use the weighted random function (C, P) to determine the category of the next POI, where C is the list of all categories and P is the probability corresponding to each category; run the weighted random function (C, P) a sufficient number of times to obtain a random number sequence that satisfies the input probability distribution, thereby obtaining the required POI.

[0028] As a preferred technical solution, using the confused trajectory T 1 as the input, the intermediate trajectory T d is obtained by using the intermediate point generation algorithm, specifically:

[0029] Traverse each position on the confused trajectory T 1 For each intermediate point, generate a circle with the center p i-1 and the radius d, where

[0030] For each θ degree, select a position and add it to the candidate trajectory set Candidates, as shown in the following formula:

[0031] Candidates = Candidates ∪ dest(p i-1 , d, K·θ);

[0032] where dest(p i-1 , d, K·θ) is a function that determines the target position starting from the origin p i-1 with a distance d and a rotation angle θ;

[0033] The confused trajectory T 1 Select a suitable stopping point from the candidate trajectory set Candidates and output the intermediate trajectory T d .

[0034] As a preferred technical solution, using the initial trajectory T 0 and the intermediate trajectory T d as inputs, an anonymous trajectory T dummy that protects semantic information is obtained by using a trajectory correction algorithm, specifically:

[0035] Check whether the number of position points included in the initial trajectory T 0 and the intermediate trajectory T d is the same;

[0036] If they are the same, add the original timestamp to t and assign them to the corresponding position points, and assign them to the corresponding position points on T d as shown in the following formula:

[0037]

[0038] where, is a point on the intermediate trajectory T d , is a point on the initial trajectory T 0 , and the random shift random_timeshift should be within a certain range suitable for the configured protection level;

[0039] Calculate the slope, that is, the movement trend of the two trajectories. If the difference in slopes is greater than the set threshold, remove the virtual trajectory and regenerate it. The calculation of the slope is as follows:

[0040]

[0041] where, and are the midpoint coordinates of the trajectory.

[0042] Another aspect of the present invention provides a semantic-sensitive location trajectory privacy protection system for trajectory publishing datasets, which is applied to the above-mentioned semantic-sensitive location trajectory privacy protection method for trajectory publishing datasets, and includes a stop point search module, a trajectory obfuscation module, an intermediate trajectory generation module, and a trajectory correction module;

[0043] The stop point search module is used to search for stop points in the user's initial trajectory T 0 to obtain stop points;

[0044] The trajectory obfuscation module is used to obtain POIs by using the category distance priority method or the Markov matrix method, and use the obtained POIs to replace the stop points to obtain the obfuscated trajectory T 1 ;

[0045] The intermediate trajectory generation module is used to use the obfuscated trajectory T 1 as input, and use the intermediate point generation algorithm to obtain the intermediate trajectory T d ;

[0046] The trajectory correction module is used to use the initial trajectory T 0 and the intermediate trajectory T d as input, and use the trajectory correction algorithm to obtain the anonymous trajectory T dummy .

[0047] Another aspect of the present invention provides a storage medium storing a program, which when executed by a processor, implements the above-mentioned semantic-sensitive location trajectory privacy protection method for trajectory publishing datasets.

[0048] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0049] (1) The present invention proposes a stop point obfuscation technology, which uses the existing POI information and stop point search technology to obfuscate the user's trajectory, realizing semantic-oriented trajectory privacy protection.

[0050] (2) By adopting the category-distance priority method (CDP) and the Markov matrix (MM), the present invention selects the best POI for replacing the stop points according to category and probability judgment, protects the semantic information of the trajectory after anonymization, and thus reduces the utility loss of the anonymized trajectory.

[0051] (3) The present invention allows users to customize random and θ, adjust the required privacy level, and can determine the anonymized trajectory according to the user's own privacy protection needs. Brief Description of the Drawings

[0052] Figure 1It is a schematic diagram of three entities of the semantic-sensitive location trajectory privacy protection method for the trajectory publishing dataset in the embodiment of the present invention;

[0053] Figure 2 It is the overall flowchart of the semantic-sensitive location trajectory privacy protection method for the trajectory publishing dataset in the embodiment of the present invention;

[0054] Figure 3 It is a schematic diagram of the category transition matrix in the embodiment of the present invention;

[0055] Figure 4 It is a schematic diagram of the intermediate point search in the embodiment of the present invention;

[0056] Figure 5 It is a schematic diagram of the structure of the semantic-sensitive location trajectory privacy protection system for the trajectory publishing dataset in the embodiment of the present invention;

[0057] Figure 6 It is a schematic diagram of the structure of the storage medium in the embodiment of the present invention. Detailed implementation manners

[0058] In order to enable those skilled in the art of the present technology to better understand the solutions of this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts belong to the scope of protection of this application.

[0059] Embodiment

[0060] The present invention obtains the stop points (semantic information) of the trajectory by defining time and distance thresholds, then queries the POI information near the stop points from the LBS database or the LBS open API, and then uses the searched POI to replace the original stop points of the trajectory, so as to meet the privacy protection requirements. At the same time, in order to retain the original semantic information of the user's trajectory, we use the category-distance priority method (CDP) to determine the POI for replacement, so as to minimize the loss of trajectory utility. In addition, in order to prevent more powerful attackers (with rich background knowledge), we also design the Markov matrix (MM) method to strengthen the protection. We select the POI category with the highest probability in the Markov matrix as the next POI category, so as to simulate the more real behavior pattern of the user and achieve the effect of strengthening the protection.

[0061] As Figure 1As shown in the figure, this embodiment provides a semantic-sensitive location trajectory privacy protection method for trajectory publishing datasets, which involves three entities: the data collector, the server, and the user terminal. The data collector collects the user's GPS data, and the server processes the GPS data through the algorithm designed in the present invention and then sends it to the user terminal, thereby achieving the effect of protecting the user's trajectory privacy.

[0062] The specific implementation process is as Figure 2 shown, and it includes the following steps:

[0063] S1. Search for stop points in the user's initial trajectory T 0 to obtain stop points. Specifically:

[0064] S1.1. The initial trajectory T 0 consists of a series of spatio-temporal three-dimensional tuples and user identifiers, and is expressed as: T 0 ={(x 1 , y 1 , t 1 ), (x 2 , y 2 , t 2 ), …, (x n , y n , t n ), U i}; where (x n , y n ) is a positioning point, t n is a timestamp, and U i is the user identifier;

[0065] S1.2. Set the distance threshold distThreh and the time threshold timeThreh;

[0066] S1.3. For any two adjacent coordinate points (x i , y i , t i ) and (x j , y j , t j ), if the distance between these two coordinate points is less than the set distance threshold distThreh, or the time interval between the two coordinate points is greater than the set time threshold timeThreh, then mark (x i , y i , t i ) as a stop point.

[0067] S2. In the stop point search algorithm, the coordinate points that meet the time and distance thresholds will be replaced as stop points. In this embodiment, the POI is obtained by using the category distance priority method or the Markov matrix method, and the user can select different strength replacement methods according to their own needs;

[0068] S2.1. The POI is obtained by using the category distance priority method (CDP) as follows:

[0069] Determine the initial radius parameter r and the POI category level k;

[0070] Set the POI retrieval area with the initial stop point p as the center and r as the radius, and search for all POIs within the circular POI retrieval area from the LBS database;

[0071] Sort the searched POIs according to the category distance. Among them, the POIs belonging to the same category as the stop point p have the highest priority as replacement points. If they are of the same category as the stop point p, the POIs with smaller distances have higher priorities;

[0072] If candidates of the same category level K are found, use them as the output results;

[0073] If no candidates of the same category level K are found within the POI retrieval area, increase the radius parameter r and search again;

[0074] When r ≥ the set maximum radius r_max and the search result is empty, select the POI with the smallest distance from the stop point p as the output result.

[0075] S2.2. The POI is obtained by using the Markov matrix method (MM) as follows:

[0076] Obtain the POI category conversion matrix of the Geolife dataset, as Figure 3 shown. Each row of the POI category conversion matrix represents the probability from the category at the current position to other categories at the next position;

[0077] Find the category with the highest transition probability and search for the next POI from the target category. Among them, the probability M of transferring from position i to position j ij is specifically as follows:

[0078]

[0079] where Pr(·) is the Markov probability transfer function, dst is the target position, src is the source position, C i and C j represent two stop points.

[0080] However, from the experimental results, since most of the user's activities are centered around the residence, in the generated Markov matrix, almost every row has the highest probability belonging to the "estate" category. To simulate the behavior pattern of real users, the present invention uses a weighted random function (C, P) to determine the category of the next POI, where C is the list of all categories and P is the probability corresponding to each category, that is, the probability of each row of the matrix. After running a sufficient number of times, the random number sequence generated by this function will satisfy the input probability distribution, thus obtaining the desired POI.

[0081] S3. Replace the stopping point with the obtained POI to get the obfuscated trajectory T 1 ;

[0082] S4. Take the obfuscated trajectory T 1 as the input, and use the midpoint generation algorithm to obtain the intermediate trajectory T d , specifically:

[0083] S4.1. Traverse each position on the obfuscated trajectory T 1 . For each midpoint, generate a circle with the center p i-1 and a radius of d, where as Figure 4 shown;

[0084] S4.2. For each θ degree, select a position and add it to the candidate trajectory set Candidates, as follows:

[0085] Candidates = Candidates ∪ dest(p i-1 , d, K·θ);

[0086] where dest(p i-1 , d, K·θ) is a function that determines the target position starting from the origin p i-1 with a distance of d and a rotation angle of θ;

[0087] S4.3. The obfuscated trajectory T 1 selects a suitable stopping point from the candidate trajectory set Candidates and outputs the intermediate trajectory T d .

[0088] In this algorithm, random in the d formula in S4.1 and θ in the Candidates formula in S4.2 can be dynamically adjusted according to the custom size to configure and adjust the privacy protection level. The larger the value, the smaller the similarity between the virtual trajectory and the original trajectory, and the better the protection performance.

[0089] S5. Take the initial trajectory T 0 and the intermediate trajectory Td As the input, an anonymous trajectory T that protects semantic information is obtained by using a trajectory correction algorithm dummy , specifically as follows:

[0090] S5.1. Check whether the number of position points included in the initial trajectory T 0 and the intermediate trajectory T d is the same;

[0091] S5.2. If they are the same, add the original timestamp to t and assign them to the corresponding position points, and assign them to the corresponding position points on T d as follows:

[0092]

[0093] where is a point on the intermediate trajectory T d , is a point on the initial trajectory T 0 , and the random shift random_timeshift should be within a certain range suitable for the configured protection level;

[0094] S5.3. Calculate the slope, that is, the movement trend of the two trajectories. If the difference in slopes is greater than the set threshold, remove the virtual trajectory and regenerate it; the calculation of the slope is as follows:

[0095]

[0096] where and are the midpoint coordinates of the trajectory.

[0097] As Figure 5 shown, in another embodiment of the present application, a semantic-sensitive location trajectory privacy protection system for a trajectory publishing data set is provided. The system includes a stop point search module, a trajectory obfuscation module, an intermediate trajectory generation module, and a trajectory correction module;

[0098] The stop point search module is used to search for stop points in the user's initial trajectory T 0 to obtain stop points;

[0099] The trajectory obfuscation module is used to obtain POIs by using the category distance priority method or the Markov matrix method, and use the obtained POIs to replace the stop points to obtain an obfuscated trajectory T 1 ;

[0100] The intermediate trajectory generation module is used to use the obfuscated trajectory T 1 as the input and obtain an intermediate trajectory T d by using an intermediate point generation algorithm;

[0101] The trajectory correction module is used to use the initial trajectory T 0 and the intermediate trajectory T d as inputs, and adopt a trajectory correction algorithm to obtain an anonymous trajectory T that protects semantic information dummy .

[0102] It should be noted here that the system provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure is divided into different functional modules to complete all or part of the functions described above. This system is a semantic-sensitive location trajectory privacy protection method for trajectory publishing data sets in the above embodiment.

[0103] As Figure 6 shown, in another embodiment of the present application, a storage medium is further provided, storing a program, which, when executed by a processor, implements a semantic-sensitive location trajectory privacy protection method for a trajectory publishing data set, specifically:

[0104] Search for stop points in the user's initial trajectory T 0 to obtain stop points;

[0105] Obtain POIs by using the category distance priority method or the Markov matrix method;

[0106] Use the obtained POIs to replace the stop points to obtain a confused trajectory T 1 ;

[0107] Use the confused trajectory T 1 as an input, and adopt an intermediate point generation algorithm to obtain an intermediate trajectory T d ;

[0108] Use the initial trajectory T 0 and the intermediate trajectory T d as inputs, and adopt a trajectory correction algorithm to obtain an anonymous trajectory T that protects semantic information dummy .

[0109] It should be understood that each part of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiment, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following well-known technologies in the art can be used: discrete logic circuits with logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits with appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0110] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited by the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principle of the present invention shall be equivalent replacement methods and are all included in the protection scope of the present invention.

Claims

1. A semantic-sensitive location trajectory privacy protection method for trajectory publishing datasets, characterized in that, it includes the following steps: The initial trajectory of the user T 0 Perform a stop point search on it to obtain the stop point; Obtain POIs using the category distance priority method or the Markov matrix method; Replace the stop points with the obtained POIs to get the obfuscated trajectory T 1 ; With a confused trajectory T 1 As the input, use the midpoint generation algorithm to obtain the intermediate trajectory T d ; With the initial trajectory T 0 and the intermediate trajectory T d as inputs, an anonymous trajectory that protects semantic information is obtained by using a trajectory correction algorithm T dummy , specifically: Check the initial trajectory T 0 and the intermediate trajectory T d to see if the number of included position points is consistent; If they are consistent, add the original timestamp to t and assign them to the corresponding position points, and distribute them to T d the corresponding position points above, as shown in the following formula: ； Among them, P i d is the point on the intermediate trajectory T d on the point, P i 0 is the point on the initial trajectory T 0 on the point, randomly shifted random _ timeshift should be within a certain range suitable for the configured protection level; Calculate the slope, that is, the movement trend of two trajectories. If the difference in slopes is greater than the set threshold, remove the virtual trajectory and regenerate it. The calculation of the slope is as follows: ； Among them, and are the midpoint coordinates of the trajectory.

2. The semantic-sensitive location trajectory privacy protection method for trajectory publishing datasets according to claim 1, characterized in that, The initial trajectory of the user T 0 Perform a stop point search, specifically: Initial trajectory T 0 Expressed as: T 0 ={( x 1 , y 1 , t 1 ),( x 2 , y 2 , t 2 ),⋯,( x n , y n , t n ), U i};\ where\ ( x n , y n )\ is\ a\ positioning\ point, t n is\ a\ timestamp, U i is\ the\ user\ identifier; Set distance threshold distThreh and time threshold timeThreh ; For any two adjacent coordinate points ( x i , y i , t i ) and ( x j , y j , t j ), if the distance between these two coordinate points is less than the set distance threshold distThreh , or the time interval between the two coordinate points is greater than the set time threshold timeThreh , then mark ( x i , y i , t i ) as a stop point.

3. The semantic-sensitive location trajectory privacy protection method for trajectory publishing datasets according to claim 1, characterized in that, The method of obtaining POIs using the category distance priority method is specifically: Determine the initial radius parameter r and the POI category level K ; With the initial stop point p as the center, r set a POI retrieval area with this as the radius, and search for all POIs within the circular POI retrieval area from the LBS database; Sort the searched POIs according to the category distance, where the POIs belonging to the same category as the stop point p have the highest priority as replacement points. If they belong to the same category as the stop point p the POIs with smaller distances have higher priorities; If candidates of the same category level are found K they will be used as the output result; If no candidate of the same category level is found within the POI search area K then increase the radius parameter r and conduct the search again; When r is greater than or equal to the set maximum half r_max , and the search result is empty, select the POI with the smallest distance from the stop point p as the output result.

4. The semantic-sensitive location trajectory privacy protection method for trajectory publishing datasets according to claim 1, characterized in that, The method of obtaining POIs using the Markov matrix method is specifically: Obtain the POI category conversion matrix of the Geolife dataset. Each row of the POI category conversion matrix is the probability from the category of the current position to other categories at the next position; Find the category with the highest transition probability and search for the next POI from the target category. Among them, the probability of transitioning from position i to position j is M ij Specifically, it is as follows: ； Among them, is the Markov probability transition function, dst is the target position, src is the source position, C i and C j represent two stopping points; Use a weighted random function (C, P) to determine the category of the next POI, where C is the list of all categories and P is the probability corresponding to each category; Run the weighted random function (C, P) a sufficient number of times to obtain a random number sequence that satisfies the input probability distribution, thereby obtaining the required POIs.

5. The semantic-sensitive location trajectory privacy protection method for trajectory publishing datasets according to claim 1, characterized in that, The obfuscated trajectory T 1 is used as the input, and an intermediate point generation algorithm is adopted to obtain an intermediate trajectory T d , specifically: Traverse the obfuscation trajectory T 1 For each position on, for each intermediate point, generate a circle with a center at p i-1 and a radius of d , where ; For each θ degree, select a position to add to the candidate trajectory set Candidates , as follows: ； Among them starts from the origin p i-1 and is a function that determines the target position starting from the origin, with a distance d and a rotation angle θ ; Confused trajectory T 1 Select a suitable stopping point from the candidate trajectory set Candidates and output the intermediate trajectory T d .

6. A semantic-sensitive location trajectory privacy protection system for trajectory publishing datasets, characterized in that, It is applied to the semantic-sensitive location trajectory privacy protection method for trajectory publishing datasets described in any one of claims 1-5, and includes a stop point search module, a trajectory confusion module, an intermediate trajectory generation module, and a trajectory correction module; The stop point search module is used to search for stop points in the user's initial trajectory T 0 and obtain stop points; The trajectory confusion module is used to obtain POIs by using the category distance priority method or the Markov matrix method, and replace the stop points with the obtained POIs to obtain the confused trajectory T 1 ; The intermediate trajectory generation module is used to take the obfuscated trajectory T 1 as input and obtain the intermediate trajectory by using the intermediate point generation algorithm T d ; The trajectory correction module is used to take the initial trajectory T 0 and the intermediate trajectory T d as inputs, and use a trajectory correction algorithm to obtain an anonymous trajectory that protects semantic information T dummy .

7. A storage medium stores a program, characterized in that: When the program is executed by a processor, it implements the semantic-sensitive location trajectory privacy protection method for trajectory publishing datasets described in any one of claims 1-5.

Images