Position privacy protection method based on time and space

Abstract

The invention relates to the technical field of privacy protection, in particular to a space-time-based location privacy protection method, which comprises the following steps: receiving a location service request sent by a requesting user, and constructing an anonymous set; constructing a grid system based on a road network and a user track, and calculating an identifier of a grid unit of a user position point; drawing a semantic tree by using the semantic positions in the grids; calculating the space-time similarity between the request user and the auxiliary user by using the grid, and generating an anonymous set by using the auxiliary user; when an auxiliary user leaves the anonymous set, selecting an auxiliary user position by utilizing an evaluation standard according to cached auxiliary user position information, and constructing an auxiliary user virtual track; and constructing an anonymous set by using the virtual track, and sending the anonymous set to the LBS to obtain service According to the invention, the problem of protecting the location privacy of the user when the user uses the LBS service is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of privacy protection, and in particular to a location privacy protection method based on time and space. Background Art

[0002] With the advent of the information age, wireless positioning technology has developed rapidly, and Internet mobile terminals such as smart phones and tablet computers have been rapidly popularized. Location-Based Service (LBS) has also emerged. Its convenient, fast and efficient services are deeply loved by people and are more and more widely used, and can be seen everywhere in daily life. While LBS is being used, a large amount of user location data has been generated. These data are stored in the cloud server and have great academic, commercial and social value. However, at the same time, it also poses a threat to the personal privacy and security of users.

[0003] The publication number is CN119421142A, which constructs two generative adversarial networks to respectively learn the travel habits of users and the trajectory information of real users, and generates virtual user LBS query services and corresponding trajectory data. However, this method has problems such as large training difficulty, poor stability and difficult evaluation.

[0004] In trajectory privacy protection, the k-anonymity technology is a commonly used privacy protection method. It generalizes or confuses the trajectory data of users with other k-1 pieces of trajectory data on key attributes, making it difficult to directly identify the trajectory data of a single user. However, the k-anonymity technology also has some significant drawbacks, which limit its effectiveness and security in complex privacy protection scenarios. Summary of the Invention

[0005] Aiming at the deficiencies of the existing methods, the present invention solves the problem of protecting the location privacy of users when using LBS services.

[0006] The technical solution adopted by the present invention is: a location privacy protection method based on time and space includes the following steps:

[0007] Step 1: Receive a location service request sent by a user and record the location information of the requesting user;

[0008] The user location information includes an identity credential, longitude and latitude, and time.

[0009] Step 2: Construct a grid system based on the road network and user trajectories, and calculate the identifier of the grid cell where the requesting user's location point is located;

[0010] As a preferred embodiment of the present invention, the identifier formula of the grid cell is:

[0011]

[0012] Among them, (c, r) is used to represent the index of each cell, and (x a , y a ) are the longitude and latitude of the lower left corner and the upper right corner of the selected query area, (x c , y c ) are the longitude and latitude of the position to be found, and m is the number of side indices of the divided grid system.

[0013] Step 3: Draw a semantic tree using the semantic positions in the grid;

[0014] As a preferred embodiment of the present invention, the construction of the semantic tree includes:

[0015] Step 31: Form an intermediate layer using semantic classification and semantic clustering;

[0016] Step 32: Semantically generalize the intermediate layer nodes upwards to generate semantic categories until the root node is generated;

[0017] Step 33: Starting from the intermediate layer, subdivide the nodes into subcategories.

[0018] Step 4: Calculate the spatio-temporal similarity between the requesting user and the assisting user using the grid, and generate an anonymous set using the assisting user;

[0019] As a preferred embodiment of the present invention, Step 4 specifically includes:

[0020] First, calculate the similarity of the movement paths of the requesting user and the assisting user;

[0021] Secondly, calculate the similarity of the movement speeds of the requesting user and the assisting user;

[0022] Finally, calculate the movement similarity between the requesting user and the assisting user using the path and speed similarities. The formula is:

[0023] S(a1, a j ) = β1 × δ(a1, a j ) + β2 × γ(a1, a j ) (4)

[0024] Among them, δ(a1, a j ) refers to the movement path similarity between the requesting user a1 and the assisting a j , and γ(a1, a j ) refers to the movement speed similarity between the requesting user a1 and the assisting user a j , and β1 and β2 are weights.

[0025] Step Five: When an assisting user leaves the anonymous set, based on the cached location information of the assisting user, select a new location of the assisting user using the evaluation criteria, and construct a virtual trajectory of the assisting user.

[0026] As a preferred embodiment of the present invention, when an assisting user leaves the anonymous set includes: when the area of the anonymous set of an assisting user exceeds the area threshold at a certain moment, add a new location of the assisting user to replace the original location of the assisting user and enter the anonymous set.

[0027] As a preferred embodiment of the present invention, when selecting the location of the assisting user, ensure the direction similarity and distance similarity between the real trajectory sequence and the false trajectory sequence. The formula is:

[0028]

[0029] where, represents the grid location where the requesting user a1 is located at time t; represents the grid location where the requesting user a1 is located at t i+1 where; represents the grid location where the assisting user a i is located at t; represents the grid location where the new assisting user a' i is located at time t + 1.

[0030] As a preferred embodiment of the present invention, the evaluation criteria are weighted and fused by calculating the location freshness and the Markov chain prediction location probability.

[0031] As a preferred embodiment of the present invention, the formula of the evaluation criteria is:

[0032] Μ = β1 × 1 / Fresh(a i ) + β2 × p(loc i |loc i-1 ) (12)

[0033] where, refers to the novelty of the assisting user a i to the grid location x, and p(loc i |loc i-1 ) refers to the probability that the location point loc i-1 migrates to the location point loc i , and β1 and β2 are weights.

[0034] As a preferred embodiment of the present invention, the formula of the freshness is:

[0035]

[0036] where, Indicates assisting user a i Regarding the novelty of grid position x, Represents the historical access frequency of the assisting user to this grid position, Indicates assisting user a i The time when the grid position was last accessed, Indicates assisting user a i The duration of accessing this grid position, Indicates the degree of difference between this position and the historical positions of assisting user a i

[0037] Step Six: Construct an anonymous set using the virtual trajectory and send it to the LBS to obtain services.

[0038] As a preferred embodiment of the present invention, a spatio-temporal location privacy protection system includes: a memory for storing instructions executable by a processor; a processor for executing the instructions to implement the spatio-temporal location privacy protection method.

[0039] As a preferred embodiment of the present invention, a computer-readable medium storing computer program code, the computer program code implementing the spatio-temporal location privacy protection method when executed by a processor.

[0040] Advantages of the present invention:

[0041] 1. By constructing a virtual trajectory and a k-anonymous set, the present invention can effectively prevent attackers from identifying the real location of a user by analyzing the user's trajectory data; compared with traditional k-anonymous technologies, the present invention not only considers the user's location data, but also combines spatio-temporal similarity and semantic tree analysis to further enhance the strength and accuracy of privacy protection; dynamically adjusts the anonymous set according to the user's moving speed and direction to ensure the continuous effectiveness of privacy protection during the user's movement; when the assisting user leaves the anonymous set, through moving similarity calculation and location semantic coherence, ensures that the generated virtual trajectory is highly similar to the user's real trajectory, thereby reducing the negative impact on service quality;

[0042] 2. Introducing a semantic tree and a Markov chain model can evaluate the semantic and logical rationality of the assisting user's location selection; by analyzing the historical behavior and location semantics of the assisting user, a more reasonable virtual trajectory can be generated, avoiding generating illogical location sequences, and further enhancing the reliability of privacy protection;

[0043] 3. By comprehensively considering multi-dimensional information such as the user's location, time, moving path, and speed, it can adapt to complex privacy protection scenarios, especially in the case where the user moves frequently or the trajectory changes greatly, and still effectively protect the user's privacy. Description of the Drawings​

[0044] Figure 1 This is a flowchart of the time - space - based location privacy protection system technology of the present invention.

[0045] Figure 2 It is a schematic diagram of the hardware of the time - space - based location privacy protection system of the present invention;

[0046] Figure 3 It is a schematic diagram of the trajectory tree of the present invention;

[0047] Figure 4 In, a is a schematic diagram for selecting the next location grid, Figure 4 In, b is a schematic diagram of grid expansion. Specific embodiments

[0048] The following further illustrates the present invention in conjunction with the accompanying drawings and embodiments. This figure is a simplified schematic diagram, which only illustrates the basic structure of the present invention in a schematic way, so it only shows the components related to the present invention.

[0049] As Figure 1 shown, a time - space - based location privacy protection method includes the following steps:

[0050] Step 1: The central server receives the location service request sent by the user and records the location information α1=(id, loc, t) of the requesting user;

[0051] Among them, id is the user identity credential, t is the time, t≥0, loc = {x , y}, x is the longitude coordinate, x≥0, y is the latitude coordinate, y≥0;

[0052] Step 2: Define the scope, establish a grid system, and calculate the grid identifier of the user location point;

[0053] Use the road network and user trajectories to divide the map into a grid system; the activity area is defined as the smallest closed rectangle of the trajectory; within an activity area, the grids are divided into as similar sizes as possible; use a coordinate and grid system to achieve a regular spatial distribution; the grid division based on the user activity area is the basis for analyzing user behavior, and the grid division can be based on the relevant theories of urban road planning; if the divided grid pixels are too small, most moving points will be detected as stationary points, affecting the detection quality; the grid area size is between 0.2 and 0.25 km, which helps to improve the quality and accuracy of fixed - point detection;

[0054] The grid division includes: first generate a boundary rectangle containing the activity trajectory of the requesting user, and then divide the rectangular space into grids. Each grid in the map is square and is determined according to the latitude and longitude coordinates associated with the user's activity.

[0055] Request the service provider that provides the service to determine a large query area, assumed to be a square area, with the lower left vertex (x a , y a ) to the upper right vertex (x b , y b ). It means that an irregular space area can be simulated as a square area by using the minimum bounding rectangle; this square needs to contain the request user's activity trajectory at the previous moment and reserve an area for the request user's activity trajectory at the next moment; divide the query area into m×m grid cells of the same size, where m is specified by the request user; each grid cell is represented by (c, r), where c is the column index from left to right and r is the row index from bottom to top, satisfying 0 ≤ c, r < m; given the coordinates (x c , y c ) of a position point in a grid cell, the identification formula for this grid cell is:

[0056]

[0057] where (c, r) is the index used to represent each cell, (x a , y a ), (x b , y b ) are the longitude and latitude of the lower left and upper right corners of the selected query area, (x c , y c ) are the longitude and latitude of the position to be found, and m is the number of side index of the divided grid system.

[0058] Step 3. Use various semantic positions in the grid to draw a semantic tree;

[0059] Semantic position refers to the position in the trajectory data that can reflect the meaning of the user's behavior; the semantic attributes of all semantic positions in an area can be represented as a tree structure, such as the position semantic tree shown in Figure 3 ;

[0060] Branch semantic tree construction process:

[0061] Step 31. Classify semantics according to common granularity levels, such as junior high school, high school, university, gym, etc., and form an intermediate layer through semantic clustering;

[0062] Step 32. The intermediate layer nodes can be further semantically generalized upwards to generate more general semantic categories until the root node is generated;

[0063] Step 33. Starting from the intermediate layer, further divide the nodes into specific subclasses; the subclasses represent more specific semantic categories;

[0064] The leaf nodes in the location semantic tree structure represent the specific semantic attributes of the location in the geospatial environment, while the non-leaf nodes symbolize broader or more general classifications or generalizations of the semantic attributes of their child nodes; when measuring the semantic similarity between two user locations, the distance between these two locations in the location semantic tree can be examined; specifically, if two locations have the same parent node in the location semantic tree, it can be considered that these two locations have similar semantic attributes.

[0065] Step 4: Use the grid to calculate the spatio-temporal similarity between the requesting user and the surrounding assisting users, and select k - 1 assisting users to generate the k-anonymous set U;

[0066] First, calculate the movement similarity, which is the spatio-temporal similarity. The movement similarity refers to the degree of similarity between two users in terms of their continuous query movement paths and movement speeds. The larger the value, the more similar the movement paths and movement speeds are.

[0067] The movement path similarity is measured by the ratio of the number of grids passed by two trajectories to the size of the entire trajectory; given the requesting user a1 and the assisting user a j , whose corresponding movement paths are path1 and path j , calculate the movement path similarity between the two users. The formula is:

[0068]

[0069] Among them, path1.gird refers to the grids passed by the trajectory of the requesting user a1, path1.size refers to the size of the trajectory of the requesting user a1, path j .gird refers to the grids passed by the trajectory of the assisting user a j , path j .size refers to the size of the trajectory of the assisting user a j .

[0070] The movement speed similarity is measured by the ratio of the difference in average speeds of two users to the maximum average speed; given the requesting user a1 and the assisting user a j , whose corresponding average speeds are ν1 and ν j , calculate the movement speed similarity formula between the two users as:

[0071]

[0072] Among them, v1, v j are the average speeds of the requesting user a1 and the assisting user a j respectively, and are calculated through the cache module of the central server.

[0073] The formula for calculating the movement similarity between two users is:

[0074] S(a1,a j ) = β1×δ(a1,a j ) + β2×γ(a1,a j ) (4)

[0075] Wherein, S(a1,a j ) refers to the movement similarity between the requesting user a1 and the assistant a j . δ(a1,a j ) refers to the movement path similarity between the requesting user a1 and the assistant a j . γ(a1,a j ) refers to the movement speed similarity between the requesting user a1 and the assistant user a j . β1 and β2 are weights, β1, β2 ≥ 0, and β1 + β2 = 1;

[0076] According to the movement similarity, i.e., spatio-temporal similarity, select the k - 1 assistant users closest to the requesting user a1 to form the anonymous set U.

[0077] U The expression of U is: U = {a1, a2, a i , ··· a k}, where a1 is the requesting user, and a i are the selected best k - 1 anonymous assistant users, 2 ≤ i ≤ k;

[0078] The present invention adopts a centralized system architecture; based on the assumption that the central server is completely trustworthy; as Figure 2 shown, there is a cache module in this central server; this cache module is responsible for storing the data related to the service requests sent by users in the past period of time, and these data provide important reference bases for user behavior analysis and privacy protection.

[0079] Step Five: When an assistant user leaves the anonymous set U, according to the location information of the assistant user cached in the central server, use the evaluation criteria to select appropriate locations to form the virtual trajectory of the assistant user;

[0080] Due to the limitation of the size of the anonymous set, a too large size will affect the service quality, so an area threshold s θ is set before calculation to ensure the service quality;

[0081] At a certain moment, if the area of the anonymous set U exceeds s θ due to reasons such as speed and direction of an assistant user, new locations need to be added to replace the location of this assistant user and enter the anonymous set U;

[0082] For the first anonymous positioning request, due to location anonymity, trajectory features are not considered; for the i-th (i > 1) anonymous positioning request, the position change between two consecutive positioning requests should be considered; using the grid method, the selected position should satisfy the direction similarity and distance similarity between the trajectory sequence of the requesting user and the trajectory sequence of the assisting user, so that the attacker cannot discover the true trajectory sequence of the requesting user.

[0083] Suppose at time t i+1 , due to reasons such as the speed and direction of the assisting user deviating from the direction of the requesting user, if this requesting user continues to be included, it will cause the anonymous area to exceed s θ , making it vulnerable to attack. At this time, a new assisting user needs to be selected according to the moving speed and direction of the requesting user to replace the position of this assisting user and enter the anonymous set U. The identity credential and time of this original assisting user are retained, and only its position information is replaced with the position information of the new assisting user.

[0084] When selecting the position of the assisting user, the formulas for ensuring the direction similarity and distance similarity between the trajectory sequences are as follows:

[0085]

[0086] Among them, represents the grid position where the requesting user a1 is located at time t; represents the grid position where the requesting user a1 is located at t i+1 ; represents the grid position where the assisting user a i is located at t; represents the new assisting user a' i at time t + 1.

[0087] Using formula (5) ensures that the trajectory position vector of the newly added assisting user is the same as that of the requesting user.

[0088] As Figure 4 in (a), the maximum vector of the requesting user's position is: Therefore, the grid coordinates of the newly added assisting user at time t i+1 are calculated as:

[0089] When the candidate grid does not contain any assisting users, it is expanded; as Figure 4 shown in (b), each time the side length of the candidate grid is increased by 1 until there is an assisting user available for selection in the candidate grid.

[0090] In addition, a reasonable and correct location needs to be selected from the perspectives of both the user and the location to form a virtual trajectory, so as to enhance the intensity of privacy protection. That is, to enhance privacy protection, the system introduces a location evaluation mechanism to ensure that the locations in the virtual trajectory conform to the user's historical behavior patterns, including selecting a reasonable and correct location from the perspectives of both the user and the location to form a virtual trajectory.

[0091] From the user's perspective, try to select locations familiar to the auxiliary user as much as possible to form a virtual trajectory sequence to reduce the possibility of being detected by attackers. The location freshness is used to represent the familiarity of the auxiliary user with the location. The location freshness is mainly determined by four factors, namely, the frequency of the auxiliary user staying at this location, the time when the auxiliary user last visited this location, the length of time the auxiliary user stayed at this location, and the similarity between this location and the previous locations visited by the auxiliary user. The location freshness is calculated using formula (6):

[0092]

[0093] Among them, represents the novelty of the auxiliary user a i to the grid location x, represents the historical access frequency of the auxiliary user a i to this grid location, represents the time when the auxiliary user a i last visited this grid location, represents the duration of the auxiliary user a i visiting this grid location, represents the degree of difference between this location and the historical locations of the auxiliary user a i .

[0094] It can be calculated by means of the historical query probability cached in the central server. The formula is:

[0095]

[0096] Among them, represents the number of queries of the auxiliary user a i to the grid cell x; represents the number of queries of the user a i to all grid cells, and n = 1, 2, 3, 4..., m 2 ;

[0097] The formula of

[0098]

[0099] Among them, represents the auxiliary user a recorded by the central serveri The timestamp of the last visit to this location before the current time t.

[0100] ST(a i ) The formula is:

[0101]

[0102] Where T out refers to the timestamp when the auxiliary user a i left this location recorded by the central server, and T in refers to the timestamp when the auxiliary user a i entered this location recorded by the central server;

[0103] The formula of

[0104]

[0105] Where x i refers to the location point that the auxiliary user a i has ever visited. x is the candidate grid location, and |x i - x| refers to the semantic difference between x i and x, that is, the distance between them in the semantic tree; when x i and x have the same parent node, |x i - x| = 0, otherwise |x i - x| = 1.

[0106] From the perspective of location selection, when determining the next destination for the auxiliary user, two unreasonable situations may be faced: one is that the semantics of the location itself are unreasonable, and the other is that the logic between the front and back locations is unreasonable; in order to evaluate the rationality of these selections, a Markov chain model is introduced; for example, assume that the auxiliary user is currently at a breakfast shop, and the location point of the auxiliary user at the next moment causes the anonymous area to be too large and needs to be replaced with a suitable location point. There are two choices for the next location: a bar and a school; according to the concept of location freshness, the auxiliary user is more familiar with the bar and has a higher access frequency; however, it is semantically inappropriate to go directly from the breakfast shop to the bar because both the breakfast shop and the bar have obvious time attributes, usually corresponding to the morning and night periods of a day respectively; this time conflict makes this location conversion logically untenable.

[0107] Because semantics has a time attribute, when predicting the position probability using the Markov chain, the expression of the candidate position point is loc = {(x, y), type, times}; where (x, y) are the coordinates of the semantic position, type is the type of the semantic position, and the type of the semantic position is the parent node of the leaf node in the semantic tree, and times represents the number of visits to the position point per hour in a day.

[0108] Let p(a) be a transition probability matrix of user a, and the transition probability matrix p(a) is formalized as:

[0109] p(a) = (p(z t = loc1|z t-1 ), p(z t = loc2|z t-1 ),...... p(z t = loc n |z t-1 )) (11)

[0110] Where loc is the position point, n represents the number of semantic positions, z t = loc n means converting the information of the position point to z t for convenient calculation, and p(z t = loc n |z t-1 ) represents the probability that the next position is loc n-1 under the condition that the previous position is loc n .

[0111] The expression of the evaluation criterion is as follows:

[0112]

[0113] Where, refers to the freshness of the position on grid x for assisting user a i , p(loc i |loc i-1 ) refers to the probability that the position point loc i-1 migrates to the position point loc i , β1 and β2 are weights, β1, β2 ≥ 0, and β1 + β2 = 1. The larger M is, the more reasonable the selected position point is.

[0114] Step 6: Use the virtual trajectory to construct an anonymous set and send it to the LBS to obtain services to resist attacks.

[0115] Inspired by the above-described ideal embodiments of the present invention, through the above description, relevant staff can completely make various changes and modifications without departing from the technical idea of this invention. The technical scope of this invention is not limited to the content in the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A method for protecting location privacy based on space-time, characterized in that, It includes the following steps: Step 1: Receive the location service request sent by the user and record the location information of the requesting user; Step 2: Construct a grid system based on the road network and the user's trajectory, and calculate the identifier of the grid cell where the requesting user's location point is located; Step 3: Draw a semantic tree using the semantic locations in the grid; Step 4: Calculate the spatio-temporal similarity between the requesting user and the assisting users using the grid, and generate an anonymous set using the assisting users; Step 5: When an assisting user leaves the anonymous set, based on the cached location information of the assisting user, select a new location of the assisting user using the evaluation criteria and construct a virtual trajectory of the assisting user; Step 6: Construct an anonymous set using the virtual trajectory and send it to the LBS to obtain services.

2. The location privacy protection method based on time and space according to claim 1, wherein Step 4 specifically includes: First, calculate the similarity of the moving paths between the requesting user and the assisting users; Second, calculate the similarity of the moving speeds between the requesting user and the assisting users; Finally, calculate the moving similarity between the requesting user and the assisting users, and the formula is: S(a1,a j ) = β1×δ(a1,a j ) + β2×γ(a1,a j ) Among them, δ(a1, a j ) refers to the similarity of the moving paths of users a1 and a j , γ(a1, a j ) refers to the similarity of the moving speeds of users a1 and a j , and β1 and β2 are weights.

3. The location privacy protection method based on space-time according to claim 1, characterized in that When an assisting user leaves the anonymous set includes: when the area of the anonymous set of an assisting user exceeds the area threshold at a certain moment, add a new location of the assisting user to replace the original location of the assisting user and enter the anonymous set.

4. The method for protecting location privacy based on space-time according to claim 1, wherein When selecting the location of the assisting user, ensure the direction similarity and distance similarity between the real trajectory sequence and the false trajectory sequence, and the formula is: Among them, represents the grid position where the requesting user a1 is located at time t; represents the grid position where the requesting user a1 is located at time t i+1 ; represents the grid position where the assisting user a i is located at t; represents the new assisting user a' i is located at time t + 1.

5. The method for protecting location privacy based on spatio-temporal according to claim 1, wherein The evaluation criteria are weighted and fused by calculating the location freshness and the Markov chain predicted location probability.

6. The method for protecting location privacy based on spatio-temporal as claimed in claim 5, wherein, The formula of the evaluation criteria is: Among them, refers to assisting user a i with the freshness of the position on grid x, p(loc i |loc i-1 ) refers to the probability that the position point loc i-1 migrates to the position point loc i , and β1 and β2 are weights.

7. The method for protecting location privacy based on space-time according to claim 5, wherein The formula of the freshness is: Among them, represents assisting user a i for the novelty of grid position x, represents the historical access frequency of the assisting user to this grid position, represents assisting user a i the time when the grid position was last accessed, represents assisting user a i the duration of accessing this grid position, represents the degree of difference between this position and the i historical positions of assisting user a.

8. The location privacy protection method based on time and space according to claim 1, characterized in that The formula of the identifier of the grid cell is: Among them, (c, r) is used to represent the index of each cell, and (x a , y a ) are the longitude and latitude of the lower left corner and the upper right corner of the selected query area, (x c , y c ) are the longitude and latitude of the position to be found, and m is the number of side indices of the divided grid system.

9. A spatio-temporal location privacy protection system, characterized in that, It includes: A memory for storing instructions executable by the processor; A processor for executing the instructions to implement the spatio-temporal based location privacy protection method as described in any one of claims 1-8.

10. A computer-readable medium storing computer program code, characterized in that, The computer program code implements the spatio-temporal based location privacy protection method as described in any one of claims 1-8 when executed by the processor.

Images