A privacy-preserving close trajectory mining method based on grid encoding

Abstract

This invention belongs to the field of data mining and discloses a privacy-preserving close contact trajectory mining method based on grid coding. It includes two stages: object trajectory security preprocessing and privacy-preserving close contact trajectory query. In the object trajectory security preprocessing stage, the object trajectory undergoes grid coding to obtain a secure object trajectory code, and the object trajectory is then encrypted. In the privacy-preserving close contact trajectory query stage, the query trajectory undergoes the same grid coding process to obtain a secure query trajectory code. Based on close contact determination criteria, trajectory mining is performed using the object trajectory security code, ultimately yielding a close contact trajectory mining result set. This method, while ensuring privacy and security, is compatible with various scenarios and needs, achieving efficient and reliable close contact trajectory mining.

Description

Technical Field

[0001] This invention belongs to the field of data mining, specifically relating to a privacy-preserving close contact trajectory mining method based on grid coding. Background Technology

[0002] With the rapid development of big data and artificial intelligence technologies, application scenarios based on data processing and analysis are constantly expanding, especially in the fields of trajectory data analysis, privacy protection, and public safety. How to efficiently and securely mine the value of data has become an important research direction and an urgent practical need. In today's society, with the widespread use of mobile devices, IoT sensors, and positioning technologies, massive amounts of trajectory data are being collected, playing a crucial role in areas such as traffic planning and logistics management. However, trajectory data has characteristics such as high dimensionality, temporal sequence, and spatiality, and its analysis process faces technical challenges such as high computational complexity and heavy storage pressure. At the same time, trajectory data contains sensitive information such as users' travel habits and location records, and improper handling may lead to the risk of privacy leaks.

[0003] Current research on privacy-preserving close contact trajectory mining methods mostly struggles to balance efficiency and accuracy. Chinese patent application ZL2024102645673 discloses a privacy-preserving close contact object detection method based on trajectory encoding. While this method mentions close contact object detection, its accuracy is low, it is prone to missing solutions, and it cannot achieve high-precision close contact object detection. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a privacy-preserving close contact trajectory mining method based on grid coding. This method enables high-precision close contact trajectory mining with privacy protection capabilities by performing grid coding and encryption processing on trajectory data.

[0005] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0006] This invention is a privacy-preserving close contact trajectory mining method based on grid coding. The method consists of two stages: object trajectory security preprocessing and privacy-preserving close contact trajectory query. In the object trajectory security preprocessing stage, the object trajectory undergoes grid coding to obtain a secure object trajectory code, and the object trajectory is then encrypted. In the privacy-preserving close contact trajectory query stage, the query trajectory undergoes the same grid coding process to obtain a secure query trajectory code. Then, based on close contact determination criteria, close contact trajectory mining is performed, ultimately yielding a close contact trajectory mining result set. Specifically:

[0007] The object trajectory safety preprocessing stage specifically includes the following steps:

[0008] Step 1-1: Use a uniform grid partitioning method to spatially divide the region Z containing the object trajectory and the query trajectory, generating a basic grid set G. Merge four adjacent basic grids into one grid group, and all grid groups constitute the grid group set G. s Assign a unique code to each grid group, and assign each trajectory u in the object trajectory set U. h The trajectory points ∈U are mapped to the corresponding grid group codes to obtain the object trajectory security code sec. h The object trajectory and the query trajectory are both obtained through n location samplings, with the set of sampling time points being T = {t1, t2, v, t}. n}, the set of object trajectories U = {u1, u2, ..., u} m}, u h ={(l h,1 ,t1),(l h,2 ,t2),…,(l h,n ,t n )},l h,p Represents the trajectory u h At sampling time point t p Two-dimensional latitude and longitude coordinates;

[0009] Steps 1-2: Generate key K for each trajectory u in the object trajectory set U. h Symmetric encryption is performed on ∈U to obtain the encrypted trajectory. Finally, an encrypted object trajectory dataset is generated.

[0010] The privacy protection and close contact tracking process includes the following steps:

[0011] Step 2-1: Query trajectory u q The trajectory points in the graph are mapped to the corresponding grid group codes to obtain the query trajectory security code sec. q The query trajectory is denoted as u. q ={(l q,1 ,t1),(l q,2 ,t2),…,(l q,n ,t n )};

[0012] Step 2-2: Based on the close contact determination distance threshold d, and the encrypted object trajectory dataset... All object trajectories are securely encoded to perform close-contact trajectory queries, uncovering all records related to the queried trajectory. q The trajectories of objects with close relationships are obtained, and the result set R of close trajectory mining is finally obtained.

[0013] A further improvement of the present invention is that step 1-1 specifically includes the following steps:

[0014] Step 1-1-1: Divide the region Z containing the object trajectory and the query trajectory into equal-sized square base grids, the base grid set G = {g 1,1 ,g 1,2 ,…,g α,β}, where α is the number of grid rows and β is the number of grid columns;

[0015] Step 1-1-2: Merge four adjacent basic grids into one grid group s a And the diagonal length of each grid group is d, and the grid group s a ={g i,j ,g i,j+1 ,g i+1,j ,g i+1,j+1 Let there be k grid groups, and let G be the set of grid groups formed by them. s ={s1,s2,…,s k};

[0016] Step 1-1-3, constructing a structure consisting of 2 σ A set C consisting of binary codes of length σ, wherein The generation rules are as follows:

[0017]

[0018] Randomly select k codes from C and randomly map them to the grid set G. s In the k grid groups, divide grid group s a The assigned code is denoted as c. a Ensure that each grid group s a It has a unique code;

[0019] Step 1-1-4: For any grid group s a Each grid consists of four basic grids, and adjacent grid groups overlap. Let the four grid groups be s. a s b s c and s d Overlapping base mesh g i,j That is, g i,j =s a ∩s b ∩s c ∩s d For any basic mesh g i,j =s a ∩s b ∩s c ∩s dIt is uniquely represented as set e i,j ={c a ,c b ,c c ,c d}, where c a c b c c and c d They are grid groups s a s b s c and s d Grid group encoding;

[0020] Step 1-1-5, for the object trajectory u h ∈U, let its t p The base grid where the time trajectory point is located is Through step 1-1-4, we obtain The grid group encoding set is

[0021] Step 1-1-6: Transfer the object trajectory u h If the trajectory points are represented using grid group coding, then the object trajectory security code obtained after grid coding processing is:

[0022] A further improvement of the present invention is that, in step 1-1-1, in the square base grid of equal size, the diagonal length of each grid is... d is the distance threshold for determining close contact.

[0023] A further improvement of the present invention is that step 2-1 specifically includes the following steps:

[0024] Step 2-1-1: For the query trajectory u q Let its t p The base grid where the time trajectory point is located is Through step 1-1-4, we obtain The grid group encoding set is

[0025] Step 2-1-2: Query trajectory u q If the trajectory points are represented using grid group coding, then the query trajectory code obtained after grid coding processing is:

[0026] A further improvement of the present invention is that step 2-2 specifically involves: giving a detection time range [t] u ,t v For query trajectory security encoding sec q Examining dense object trajectory datasets Each object trajectory is securely encoded in sec h Let the sliding window W = {t} l+1 ,t l+2 ,…,t l+δ It consists of δ consecutive time sampling points, for any t p ∈W, if the object trajectory security encoding sec h In t p Time-based grid encoding set With query trajectory security code sec q In t p Time-based grid encoding set satisfy Then encrypt its trajectory Add the close contact trajectory mining result set R, where δ is the sliding window width.

[0027] The beneficial effects of this invention are:

[0028] This invention is based on the secure coding of object trajectories and combines it with the secure coding of query trajectories for trajectory mining. It achieves a trajectory mining method with privacy protection capabilities without the secure coding containing trajectory information.

[0029] In the process of grid encoding the trajectory, the present invention employs a fine gridding method with a simple structure, which is easy to implement and apply, and provides high query accuracy.

[0030] The mining results of this invention are returned in the form of encrypted trajectory, without directly exposing the plaintext trajectory, thus ensuring strong security.

[0031] In summary, this invention achieves efficient privacy-preserving close contact tracing while improving query accuracy. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the grid defined in this invention.

[0033] Figure 2 This is a schematic diagram illustrating the close contact determination criteria of the present invention.

[0034] Figure 3 This is a flowchart of the trajectory data mining method in this invention. Detailed Implementation

[0035] The embodiments of the present invention will be disclosed below with reference to the drawings. For clarity, many practical details will be described in the following description. However, it should be understood that these practical details are not intended to limit the invention. That is, in some embodiments of the invention, these practical details are not essential.

[0036] First, we define the close relationships of query trajectories:

[0037] Assume that both the object trajectory and the query trajectory are obtained through n location samplings, and the set of sampling time points is T = {t1, t2, ..., t}. n The query trajectory is denoted as u. q ={(l q,1 ,t1),(l q,2 ,t2),…,(l q,n ,t n )}, where l q,p Indicates query trajectory u q At sampling time point t p Two-dimensional latitude and longitude coordinates.

[0038] Definition 1 Closely connected trajectory: Given a set of object trajectories, denoted as U = {u1, u2, ..., u...} m}, where u h ={(l h,1 ,t1),(l h,2 ,t2),…,(l h,n ,t n )},l h,p Represents the trajectory u h At time sampling point t p Two-dimensional latitude and longitude coordinates.

[0039] Given a detection time range [t] u ,t v ], for querying trajectory u q The set of trajectories of the objects under consideration is U = {u1, u2, ..., u...} m Each trajectory u in} h ∈U, if there exists a sliding window W = {t} consisting of several consecutive time sampling points. l+1 ,t l+2 ,…,t l+δ If the following conditions are met, the trajectory of this object is called a closely spaced trajectory.

[0040] Condition 1: |W|=δ

[0041] Condition 2:

[0042] Where δ is the width of the sliding window W, i.e., the number of consecutive time sampling points in the window, l h,p For the trajectory u h In t p Two-dimensional latitude and longitude coordinates of time, l q,p To query trajectory u q In t p Two-dimensional latitude and longitude coordinates of a given moment. Δ(l) h,p ,l q,p ) represents tp The Euclidean distance between the coordinates of the object's trajectory at a given time and the coordinates of the queried trajectory. d is the close proximity determination distance threshold.

[0043] The above definition of close contact trajectory mining is performed in plaintext, without considering the privacy and security of the trajectory data. To protect the privacy of trajectory data, it is necessary to perform grid-based encoding and encryption on the trajectory data, and to support the mining of close contact trajectories.

[0044] like Figure 3 As shown, this invention is a privacy-preserving close contact trajectory mining method based on grid coding. The method consists of two stages: object trajectory security preprocessing and privacy-preserving close contact trajectory query. In the object trajectory security preprocessing stage, the object trajectory undergoes grid coding to obtain a secure object trajectory code, and the object trajectory is then encrypted. In the privacy-preserving close contact trajectory query stage, the query trajectory undergoes the same grid coding process to obtain a secure query trajectory code. Based on the close contact determination distance threshold d, trajectory mining is performed based on the object trajectory security code, ultimately yielding a close contact trajectory mining result set.

[0045] The object trajectory security preprocessing stage specifically includes the following steps:

[0046] Step 1-1: Use a uniform grid partitioning method to spatially divide the region Z containing the object trajectory and the query trajectory, generating a basic grid set G. Merge four adjacent basic grids into one grid group, and all grid groups constitute the grid group set G. s Assign a unique code to each grid group, and assign each trajectory u in the object trajectory set U. h The trajectory points ∈U are mapped to the corresponding grid group codes to obtain the object trajectory security code sec. h The object trajectory and the query trajectory are both obtained through n location samplings, with the set of sampling time points being T = {t1, t2, ..., t...}. n}, the set of object trajectories U = {u1, u2, ..., u} m}, u h ={(l h,1 ,t1),(l h,2 ,t2),…,(l h,n ,t n )},l h,p Represents the trajectory u h At sampling time point t p Two-dimensional latitude and longitude coordinates.

[0047] Specifically, the following steps are included:

[0048] Step 1-1-1: Divide the region Z containing the object trajectory and the query trajectory into α×β equal-sized square basic grids, and generate the basic grid set G = {g 1,1 ,g 1,2 ,…,g α,β}, where α is the number of grid rows and β is the number of grid columns;

[0049] Step 1-1-2: Merge four adjacent basic grids into one grid group s a And the diagonal length of each grid group is d, and the grid group s a ={g i,j ,g i,j+1 ,g i+1,j ,g i+1,j+1 Let there be k grid groups, and let G be the set of grid groups formed by them. s ={s1,s2,…,s k};

[0050] Step 1-1-3, constructing a structure consisting of 2 σ A set C consisting of binary codes of length σ, wherein The generation rules are as follows:

[0051]

[0052] Randomly select k codes from C and randomly map them to the grid set G. s In the k grid groups, divide grid group s a The assigned code is denoted as c. a Ensure that each grid group s a It has a unique code;

[0053] Step 1-1-4: For any grid group s a Each grid consists of four basic grids, and adjacent grid groups overlap. Let the four grid groups be s. a s b s c and s d Overlapping base mesh g i,j That is, g i,j =s a ∩s b ∩s c ∩s d Since each grid group has a unique code, and each basic grid is uniquely represented by four grid groups, therefore for any basic grid g... i,j =s a ∩s b ∩s c ∩s d It is uniquely represented as set ei,j ={c a ,c b ,c c ,c d}, where c a c b c c and c d They are grid groups s a s b s c and s d Grid group encoding;

[0054] Step 1-1-5, for the object trajectory u h ∈U, let its t p The base grid where the time trajectory point is located is Through step 1-1-4, we obtain The grid coding set is

[0055] Step 1-1-6: Transfer the object trajectory u h If the trajectory points are represented by grid coding, then the object trajectory security code obtained after grid coding is:

[0056] Steps 1-2: Generate key K for each trajectory u in the object trajectory set U. h Symmetric encryption is performed on ∈U to obtain the encrypted trajectory. Finally, an encrypted object trajectory dataset is generated.

[0057] The privacy protection close contact tracking query stage specifically includes the following steps:

[0058] Step 2-1: Query trajectory u q The trajectory points in the graph are mapped to the corresponding grid group codes to obtain the query trajectory security code sec. q The query trajectory is denoted as u. q ={(l q,1 ,t1),(l q,2 ,t2),…,(l q,n ,t n The specific steps include the following:

[0059] Step 2-1-1: For the query trajectory u q Let its t p The base grid where the time trajectory point is located is Through step 1-1-4, we obtain The grid coding set is

[0060] Step 2-1-2: Query trajectory u q If the trajectory points are represented by grid coding, then the query trajectory code obtained after grid coding processing is:

[0061] Figure 2 This is a schematic diagram illustrating the close contact determination conditions of the present invention. The trajectory represented by the circle as the trajectory point indicates the query trajectory u. q The trajectory represented by asterisks as trajectory points indicates the trajectory of the object u. h d is the distance threshold for determining close relationships, the rectangular box represents the sliding time window W, and δ is the minimum sliding window threshold for determining close relationships, that is, the sliding window must contain at least δ consecutive sampling points.

[0062] Step 2-2: Based on the close contact determination distance threshold d, and the encrypted object trajectory dataset... All object trajectories are securely encoded to perform close-contact trajectory queries, uncovering all records related to the queried trajectory. q Trajectories of objects with close relationships are obtained, ultimately leading to a close trajectory mining result set R. Specifically, the close trajectory query method is as follows: given a detection time range [t]... u ,t v For query trajectory security encoding sec q Examining dense object trajectory datasets Each object trajectory is securely encoded in sec h Let the sliding window W = {t} l+1 ,t l+2 ,…,t l+δ It consists of δ consecutive time sampling points, for any t p ∈W, if the object trajectory security encoding sec h In t p Time-based grid encoding set With query trajectory security code sec q In t p Time-based grid encoding set satisfy Then encrypt its trajectory Add the close contact trajectory mining result set R, where δ is the sliding window width.

[0063] To verify the present invention, it was compared with two other methods, and the results are shown in Table 1 below.

[0064] Table 1

[0065] method Accuracy (%) Time (s) Method 1 92.54 115.05 Method 2 87.7 477.4 This method 93 8.74

[0066] Method 1 is a Chinese patent application: ZL2024102645673. Method 2 is from Li, M., Zeng, Y., Zheng, L., Chen, L., Li, Q.: Accurate and efficient trajectory-based contact tracing with secure computation and geo-indistinguishability. In: International Conference on Database Systems for Advanced Applications.pp.300-316. Springer (2023).

[0067] As shown in Table 1, the method of the present invention has high accuracy and short time. The present invention achieves efficient privacy protection and close contact trajectory mining while improving query accuracy.

[0068] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A privacy-preserving close contact trajectories mining method based on grid coding, characterized in that: The privacy-preserving close contact trajectory mining method consists of two stages: object trajectory security preprocessing and privacy-preserving close contact trajectory query. In the object trajectory security preprocessing stage, the object trajectory is subjected to grid encoding to obtain object trajectory security encoding, and the object trajectory is encrypted. In the privacy-preserving close contact trajectory query phase, the query trajectory undergoes the same grid encoding process to obtain a security code for the query trajectory. Then, based on the close contact determination criteria, close contact trajectory mining is performed, ultimately yielding a close contact trajectory mining result set. Specifically: The object trajectory security preprocessing stage specifically includes the following steps: Step 1-1: Use a uniform grid partitioning method to spatially divide the region Z containing the object trajectory and the query trajectory, generating a basic grid set G. Merge four adjacent basic grids into one grid group, and all grid groups constitute the grid group set G. s Assign a unique code to each grid group, and assign each trajectory u in the object trajectory set U. h The trajectory points ∈U are mapped to the corresponding grid group codes to obtain the object trajectory security code sec. h The object trajectory and the query trajectory are both obtained through n location samplings, with the set of sampling time points being T = {t1, t2, ..., t...}. n }, the set of object trajectories U = {u1, u2, ..., u} m }, u h ={(l h,1 ,t1),(l h,2 ,t2),…,(l h,n ,t n )},l h,p Represents the trajectory u h At sampling time point t p Two-dimensional latitude and longitude coordinates; Steps 1-2: Generate key K for each trajectory u in the object trajectory set U. h Symmetric encryption is performed on ∈U to obtain the encrypted trajectory. Finally, an encrypted object trajectory dataset is generated. The privacy protection close contact tracking query stage specifically includes the following steps: Step 2-1: Query trajectory u q The trajectory points in the graph are mapped to the corresponding grid group codes to obtain the query trajectory security code sec. q The query trajectory is denoted as u. q ={(l q,1 ,t1),(l q,2 ,t2),…,(l q,n ,t n )}; Step 2-2: Based on the close contact determination distance threshold d, and the encrypted object trajectory dataset... All object trajectories are securely encoded to perform close-contact trajectory queries, uncovering all records related to the queried trajectory. q The trajectories of objects with close relationships are obtained, and the result set R of close trajectory mining is finally obtained.

2. The privacy-preserving close contact trajectory mining method based on grid coding according to claim 1, characterized in that: Step 1-1 specifically includes the following steps: Step 1-1-1: Divide the region Z containing the object trajectory and the query trajectory into equal-sized square base grids, the base grid set G = {g 1,1 ,g 1,2 ,…,g α,β }, where α is the number of grid rows and β is the number of grid columns; Step 1-1-2: Merge four adjacent basic grids into one grid group s a And the diagonal length of each grid group is d, and the grid group s a ={g i,j ,g i,j+1 ,g i+1,j ,g i+1,j+1 Let there be k grid groups, and let G be the set of grid groups formed by them. s ={s1,s2,…,s k }; Step 1-1-3, constructing a structure consisting of 2 σ A set C of binary codes of length σ, wherein The generation rules are as follows: k codes are randomly selected from the binary code set C and randomly mapped to the grid group set G. s In the k grid groups, divide grid group s a The assigned code is denoted as c. a Ensure that each grid group s a It has a unique code; Step 1-1-4: For any grid group s a Each grid consists of four basic grids, and adjacent grid groups overlap. Let the four grid groups be s. a s b s c and s d Overlapping base mesh g i,j That is, g i,j =s a ∩s b ∩s c ∩s d For any basic mesh g i,j =s a ∩s b ∩s c ∩s d It is uniquely represented as set e i,j ={c a ,c b ,c c ,c d }, where grid group s b ={g i,j+1 ,g i,j+2 ,g i+1,j+1 ,g i+1,j+2 }, grid group s c ={g i+1,j ,g i+1,j+1 ,g i+2,j ,g i+2,j+1 }, grid group s d ={g i+1,j+1 ,g i+1,j+2 ,g i+2,j+1 ,g i+2,j+2 }, c a c b c c and c d They are grid groups s a s b s c and s d Grid group encoding; Step 1-1-5, for the object trajectory u h ∈U, let its t p The base grid where the time trajectory point is located is Through step 1-1-4, we obtain The grid group encoding set is Step 1-1-6: Transfer the object trajectory u h If the trajectory points are represented using grid group coding, then the object trajectory security code obtained after grid-based coding is:

3. The privacy-preserving close contact trajectory mining method based on grid coding according to claim 2, characterized in that: In step 1-1-1, within the equally sized square base grid, the diagonal length of each grid cell is... d is the distance threshold for determining close contact.

4. The privacy-preserving close contact trajectory mining method based on grid coding according to claim 2, characterized in that: In step 2-1, the queried trajectory u q The trajectory points in the graph are mapped to the corresponding grid group codes to obtain the query trajectory security code sec. q Specifically, it includes the following steps: Step 2-1-1: For the query trajectory u q Let its t p The base grid where the time trajectory point is located is Through step 1-1-4, we obtain The grid group encoding set is Step 2-1-2: Query trajectory u q If the trajectory points are represented using grid group encoding, then the query trajectory encoding obtained after grid-based encoding is:

5. The privacy-preserving close contact trajectory mining method based on grid coding according to claim 1, characterized in that: Step 2-2 specifically involves: giving a detection time range [t] u ,t v For query trajectory security encoding sec q Examining the trajectory dataset of closely related objects Each object trajectory is securely encoded in sec h Let the sliding window W = {t} l+1 ,t l+2 ,…,t l+δ It consists of δ consecutive time sampling points, for any t p ∈W, if the object trajectory security encoding sec h In t p Time-based grid group encoding set With query trajectory security code sec q In t p Time-based grid group encoding set satisfy Then encrypt its trajectory Add the close contact trajectory mining result set R, where δ is the sliding window width.

Images