Range-based privacy protection spatial data set Top-k query method

Abstract

The invention belongs to the field of information retrieval and big data, and discloses a range-based privacy protection spatial data set Top-k query method, which comprises the following steps of: in a data preprocessing stage, performing grid division on an area where a position point of each spatial data set in a data warehouse is located, generating a counting vector set for each grid, and constructing a safe spatial index structure; uploading the encrypted data warehouse and the security space index structure to a cloud server; in a Top-k query stage, a query user generates an encrypted query vector set and an encrypted parameter vector for a query range request, constructs a query trap door and uploads the query trap door to a cloud server, and the cloud server completes correlation calculation in a privacy protection state by using the query trap door and a security space index structure. And finally, returning the first k encrypted spatial data sets with the highest relevancy as query results. According to the method, the risk of privacy disclosure is reduced, the full-period safety of the data is guaranteed, and the result accuracy and applicability in a complex scene are improved.

Description

Technical Field

[0001] This application belongs to the field of information retrieval and big data, and specifically relates to a range-based Top-k query method for privacy-preserving spatial datasets. Background Art

[0002] With the rapid development of smart cities and location-based services, the scale of spatial data is growing exponentially. Efficient management and retrieval of data such as geographic information, traffic flow, and real-time user locations have become critical requirements. Range-based top-k spatial dataset query technology can filter the top k most relevant datasets within a specified area, providing a core basis for decision support. However, when spatial data is outsourced to cloud servers, the plaintext storage and computation of raw data poses a serious risk of privacy leakage. For example, sensitive information such as location distribution and dataset size may be exposed. Therefore, achieving efficient query while protecting data privacy has become a pressing technical challenge.

[0003] In traditional solutions, directly encrypting spatial datasets can ensure privacy but compromises data availability, making it impossible to support correlation calculation and ranking in encrypted environments. Existing research primarily achieves ranking by simply counting the number of locations within the query range or by using coverage relationships. However, these methods have limitations. For example, an efficient spatial data range search processing method was proposed in the prior art (“Z.Li, H.Dai, J.Sun, H.Zhou, P.Li, and G.Yang, 'Esdrs: Efficient spatial dataset range search processing,' in The 26th IEEE International Conference on High Performance Computing, Communications and Systems (HPCC 2024), pp. 249–254, 2024”). However, this query method relies solely on local location point density and cannot dynamically balance the contribution weights of dataset size and local location point density, making it difficult to meet query requirements in complex scenarios. The existing technology also proposes a method for a nosql-based big spatial data management system and an adaptive grid approximation model ("X.Yang, X.Guan, Z.Pang, X.Kui, and H.Wu,'Gridmesa:A nosql-based big spatialdata management system with an adaptive grid approximation model,'FutureGeneration Computer Systems(FG CS),vol.155,pp.324–339,2024"), but this query method focuses on spatial item retrieval within a single dataset and cannot be extended to Top-K application scenarios of multiple datasets.

[0004] A query method for efficiently searching spatial datasets over multiple data sources has also been proposed in the prior art ("W. Yang, S. Wang, Y. Sun, Z. Chen, and Z. Peng,'Efficient spatial dataset searchover multiple data sources,'arXiv preprint arXiv:2311.13383, 2023"). The index structure in this query method is based on plaintext design and does not integrate encryption computing capabilities. As a result, the cloud server needs to access the original data to complete the relevance calculation, further exacerbating the risk of privacy leakage.

[0005] As can be seen, existing technical solutions focus on querying spatial datasets in plaintext, and the designed technical means cannot solve the problem of range-based top-k querying of privacy-preserving spatial datasets. Therefore, research on methods that can implement top-k querying of privacy-preserving spatial datasets in an encrypted environment has important practical significance and value. Summary of the Invention

[0006] In order to solve the above technical problems, this application proposes a range-based privacy-preserving spatial dataset Top-k query method. This technology constructs a secure spatial index structure through Bloom filter encoding and vector encryption, converts the original spatial data into an encrypted form that supports ciphertext calculation, so that the cloud server can directly calculate the correlation between the encrypted dataset and the query range without decryption, and return the Top-k query results in a privacy-preserving state based on the ciphertext calculation results.

[0007] In order to achieve the above objectives, this application is implemented through the following technical solutions:

[0008] This application is a range-based privacy-preserving spatial dataset Top-k query method, which is applied to data owners, cloud servers, and query users. The privacy-preserving spatial dataset Top-k query method specifically includes two stages: data preprocessing stage: completed independently by the data owner; Top-k query stage: completed collaboratively by the cloud server and the query user. Assume that the spatial data warehouse is denoted as D = {D1, D2, ..., D n}, where D i For a spatial data set containing several two-dimensional location points, the coverage area of all location points in the spatial data warehouse D is evenly divided to generate m×m grids, forming a grid set G = {g1, g2, ..., g m×m};

[0009] The data preprocessing stage specifically includes the following steps:

[0010] Step 1a: The data owner has a set of grids G. i , any dataset D in the statistical spatial data warehouse D j Falling on grid g i Number of internal position points According to the number of location points Generate a 2D vector Specifically:

[0011]

[0012] The final generated mesh g i The corresponding count vector set

[0013] Step 1b: The data owner generates a count vector set based on step 1a. Constructing a sequence of counting vector sets

[0014] Step 1c: The data owner constructs a secure space index structure L = {L1, L2, ..., L v}, where v≤m×m, any item in L i ∈L is a tuple consisting of an encrypted grid vector and an encrypted count vector set. Then, each spatial data set in the spatial data warehouse D is encrypted to generate an encrypted data warehouse. And the secure space index structure L and the encrypted data warehouse Upload to cloud server for storage;

[0015] The Top-k query phase includes the following steps:

[0016] Step 2a: The query range request initiated by the query user is Q = (R, b, k), where R is the query range, b is the user-defined security parameter, 0≤b≤1, and k is the number of spatial datasets returned by the request. Based on the query range R, a minimum grid set G covering the query range R is generated. R , and initialize the encrypted query vector set Constructing a query trapdoor in To encrypt the parameter vector, the query trapdoor td is sent to the cloud server as a query instruction;

[0017] Step 2b: The cloud server calculates the query range R and the encrypted data warehouse in a privacy-preserving state based on the query trapdoor td and the security space index structure L. Each encrypted dataset The correlation between them is calculated, and finally the top k encrypted space datasets with the highest correlation are returned as the query result Result.

[0018] A further improvement of the present application is that: Step 1c for each grid g in the grid set G i and grid g i The corresponding count vector set in the count vector set sequence S The count vector set For a set of non-zero vectors, generate a pair L of encrypted grid vectors and encrypted count vectors. i , specifically including the following steps:

[0019] Step 1c1: Use Bloom filter BF for grid g i Generate binary grid code vector Bi , and encode the binary grid vector B i Encrypt and generate the corresponding encrypted grid code vector Among them, BF uses w hash functions to generate a binary encoding vector of length h;

[0020] Step 1c2: count vector set Each count vector in Perform encryption processing to generate an encrypted count vector Finally, the corresponding encrypted count vector set is generated

[0021] Step 1c3: Use the encrypted grid code vector generated in step 1c1 above And the encrypted count vector set generated in step 1c2 Construct a pair of encrypted grid vector and encrypted count vector set Encrypted grid vector-encrypted count vector set tuple L i With grid g i correspond.

[0022] A further improvement of the present application is that step 2a of constructing a query trapdoor specifically includes the following steps:

[0023] Step 2a1: Minimize the grid set G R Each grid g in i , using Bloom filter BF to generate binary query encoding vector Z i , and then encode the binary query vector Z i Encrypt and generate the corresponding encrypted query vector And encrypt the query vector Add query vector collection

[0024] Step 2a2: For the user-defined security parameter b, generate a two-dimensional parameter vector V p = [1, b], then for the two-dimensional parameter vector V p Perform encryption processing to generate encryption parameter vector

[0025] Step 2a3: Use the query vector set generated in step 2a1 above and the encryption parameter vector generated in step 2a2 Constructing a query trapdoor

[0026] A further improvement of the present application is that: Step 2b, determining the final query result Result according to the query trapdoor td and the security space index structure L, specifically comprises the following steps:

[0027] Step 2b1: Encrypt the data warehouse Each encrypted dataset Initialize a relevance Rel i =0;

[0028] Step 2b2: For the encrypted query vector set Any encrypted query vector in And any constructed encrypted grid vector-encrypted count vector set tuple in the secure space index structure L like This indicates that the successfully matched grid g t Within the query range R, at this time, for the set of encrypted count vectors that are successfully matched The encrypted count vector for each successful match in In terms of Update the encrypted dataset Relevance Rel within the query range R i :

[0029]

[0030] Step 2b3, select the relevance Rel i The top k encrypted data sets with the highest values constitute the final query result set Result.

[0031] The beneficial effects of this application are:

[0032] This application constructs a secure space index structure and query trapdoors to perform Top-k query operations without exposing the actual location, reducing the risk of privacy leakage.

[0033] This application is based on a vector encryption strategy to ensure that ciphertext cannot be reversed and to guarantee data security throughout its lifecycle.

[0034] This application uses user-defined security parameter b to dynamically adjust the weight, taking into account data distribution characteristics and query requirements, and improving the accuracy and applicability of results in complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flowchart of the data preprocessing stage and Top-k query stage of this application.

[0036] Figure 2 This is the security space index structure construction diagram of this application.

[0037] Figure 3 This is a schematic diagram of the application scenario model of this application.

[0038] Figure 4 This is a query time experiment diagram under different grid division strategies of this application.

[0039] Figure 5 This is a query time experiment diagram under different query range strategies of this application. DETAILED DESCRIPTION

[0040] The following diagrams illustrate embodiments of the present invention. For clarity, many practical details are included in the following description. However, it should be understood that these practical details are not intended to limit the present invention. In other words, in some embodiments of the present invention, these practical details are not essential.

[0041] For the convenience of description, the relevant symbols are defined as follows:

[0042] Assume that the spatial data warehouse is denoted as D = {D1, D2, ..., D n}, where D i For a spatial data set containing several two-dimensional location points, the coverage area of all location points in the spatial data warehouse D is evenly divided to generate m×m grids, forming a grid set G = {g1, g2, ..., g m×m}; BF is a Bloom filter, which uses w hash functions to generate a binary code vector of length h; K1={s1,M 11 ,M 12} and K2={s2,M 21 ,M 22} is the key used for vector encryption, where s1 is an h-dimensional random Boolean vector, M 11 and M 12 is an h×h dimensional random reversible matrix, s2 is a 2-dimensional random Boolean vector, M 21 and M 22 is a 2×2 dimensional random reversible matrix.

[0043] Figure 3For the application scenario model of this application, consider an application scenario consisting of a data owner, a cloud server, and a query user. Among them, the cloud server is "honest and curious", providing data storage and query services, but trying to infer data privacy; the data owner and the query user are completely trusted parties. In order to protect the spatial dataset from being obtained by the cloud server, the data owner performs vector encryption processing on the spatial data, and then outsources the encrypted data to the cloud server for storage, and shares the key with the query user; when the query user submits a query request, a query trap is constructed based on the vector encryption strategy and uploaded to the cloud server; the cloud server completes the correlation calculation between the spatial dataset in the privacy protection state and the query range based on the query trap and the secure spatial index structure, and returns the Top-k query result Result to the query user.

[0044] Figure 2 The global space area is divided into 4×4 equal-sized grids and denoted as g0 to g 15 , each non-empty grid g i Corresponding to a count vector set Any item in this collection Record the corresponding data set D j In the grid g i The distribution characteristics within. Then for the non-empty grid g i Generate grid code vector B i , and generate encrypted grid encoding vectors based on vector encryption strategy and encrypted count vector sets Finally based on and Construct a secure space index structure L.

[0045] The range-based privacy-preserving spatial dataset Top-k query method proposed in this application mainly includes two stages: the first stage is the data preprocessing stage. First, the area where the location points of each spatial dataset in the data warehouse are located is gridded, and a counting vector set is generated for each grid; then, Bloom filter encoding and vector encryption are used to build a secure spatial index structure; finally, the encrypted data warehouse and the secure spatial index structure are uploaded to the cloud server together. The second stage is the Top-k query stage. The query user generates an encrypted query vector set and an encrypted parameter vector for the query range request, and then constructs a query trap and uploads it to the cloud server. The cloud server uses the query trap and the secure spatial index structure to complete the correlation calculation between the spatial dataset in the privacy-preserving state and the query range, and finally returns the top k encrypted spatial datasets with the highest correlation as the query results.

[0046] Specific examples Figure 1As shown, the data preprocessing stage specifically includes the following steps:

[0047] Step 1a: The data owner has a set of grids G. i , any dataset D in the statistical spatial data warehouse D j Falling on grid g i Number of internal position points According to the number of location points Generate a 2D vector Specifically:

[0048]

[0049] The final generated mesh g i The corresponding count vector set

[0050] Step 1b: The data owner generates a count vector set based on step 1a. Constructing a sequence of counting vector sets

[0051] Step 1c: The data owner constructs a secure space index structure L = {L1, L2, ..., L v}, where v≤m×m, any item L in L i ∈L is a tuple consisting of an encrypted grid vector and an encrypted count vector set. Then, each spatial data set in the spatial data warehouse D is encrypted to generate an encrypted data warehouse. And the secure space index structure L and the encrypted data warehouse Upload to the cloud server for storage.

[0052] In this step, a safe space index structure L = {L1, L2, ..., L v}, any one of which L i ∈L is a tuple consisting of an encrypted grid vector and an encrypted count vector set. Specifically, for each grid g in the grid set G i and the corresponding counting vector set in the counting vector set sequence S The count vector set For a set of non-zero vectors, generate a pair L of encrypted grid vectors and encrypted count vectors i , specifically including the following steps:

[0053] Step 1c1: Use Bloom filter BF for grid g i Generate binary grid code vector B i, using the key K1={s1,M 11 ,M 12 The random Boolean vector s1 in} will be the binary grid encoding vector B i Split into B i ′ and B i ″, recorded as B i =(B i ′,B i ″), the splitting rules are as follows:

[0054]

[0055] Then use the reversible matrix M in the key K1 11 and M 12 To B i Encrypt and generate the corresponding encrypted grid code vector

[0056] Step 1c2: count vector set Each count vector in Using the key K2 = {s2, M 21 ,M 22 The random Boolean vector s2 in} will be Split into and Recorded as The splitting rules are as follows:

[0057]

[0058] Then use the reversible matrix M in the key K2 21 and M 22 right Encrypt and generate the corresponding encrypted count vector Finally, the encrypted count vector set is generated

[0059] Step 1c3: Use the encrypted grid code vector generated in step 1c1 above And the encrypted count vector set generated in step 1c2 Construct a pair of encrypted grid vector and encrypted count vector set Encrypted grid vector-encrypted count vector set tuple L i With grid g i correspond.

[0060] The Top-k query phase includes the following steps:

[0061] Step 2a: The query range request initiated by the query user is Q = (R, b, k), where R is the query range, b is the user-defined security parameter, 0≤b≤1, and k is the number of spatial datasets returned by the request. Based on the query range R, a minimum grid set G covering the query range R is generated. R , and initialize the encrypted query vector set Constructing a query trapdoor in To encrypt the parameter vector, the query trapdoor td is sent to the cloud server as a query instruction;

[0062] In this step, the query user requests the query range Q = (R, b, k) and the minimized grid set G R Process and generate query trapdoor The specific steps are as follows:

[0063] Step 2a1: Minimize the grid set G R Each grid g in i , using Bloom filter BF to generate binary query encoding vector Z i ;

[0064] Reuse key K1 = {s1,M 11 ,M 12 The random Boolean vector s1 in} will encode the binary query vector Z i Split into Z i ′ and Z i ″, denoted as Z i =(Z i ′,Z i ″), the splitting rules are as follows:

[0065]

[0066] Then use the reversible matrix M in the key K1 11 and M 12 Encode the binary query vector Z i Encrypt and generate the corresponding encrypted query vector and will Add query vector collection

[0067] Step 2a2: For the user-defined security parameter b, generate a two-dimensional parameter vector V p = [1, b], then use the key K2 = {s2, M 21 ,M 22 The random Boolean vector s2 in V p Split into V p ′ and V p ″, denoted as Vp =(V p ′,V p ″), the splitting rules are as follows:

[0068]

[0069] Then use the reversible matrix M in the key K2 21 and M 22 For the two-dimensional parameter vector V p Encrypt and generate the corresponding encryption parameter vector

[0070] Step 2a3: Use the query vector set generated in step 2a1 above and the encryption parameter vector generated in step 2a2 Constructing a query trapdoor

[0071] Step 2b: The cloud server calculates the query range R and the encrypted data warehouse in a privacy-preserving state based on the query trapdoor td and the security space index structure L. Each encrypted dataset The correlation between them is calculated, and finally the top k encrypted space datasets with the highest correlation are returned as the query result. The specific steps include the following:

[0072] Step 2b1: Encrypt the data warehouse Each encrypted dataset Initialize a relevance Rel i =0;

[0073] Step 2b2: For the encrypted query vector set Any encrypted query vector in And any constructed encrypted grid vector-encrypted count vector set tuple in the secure space index structure L like This indicates that the successfully matched grid g t Within the query range R, at this time, for the set of encrypted count vectors that are successfully matched The encrypted count vector for each successful match in In terms of Update the encrypted dataset Relevance Rel within the query range R i :

[0074]

[0075] Step 2b3, select the relevance Rel iThe top k encrypted data sets with the highest values constitute the final query result set Result.

[0076] like Figure 4 As shown, this application plots the runtime changes of Top-k queries under different grid partitioning strategies. In the experiment, the number of datasets n in the spatial data warehouse D remains unchanged, the number of grids corresponding to the query grid coverage (GR--gridcounts) remains unchanged, and only the granularity of the grid partitioning is changed, which is recorded as G-gridcounts. It can be seen that with the increase of grid partitioning granularity, the runtime of Top-k queries shows an overall upward trend, with some local cases of first decreasing and then increasing. Finer grid partitioning can more accurately represent the distribution characteristics of spatial data, allowing the cloud server to more efficiently filter out qualified grids when calculating relevance, reducing unnecessary calculations. However, when the grid partitioning is too fine, the number of grids increases sharply, resulting in an increase in the size of the encrypted query vector set. This increases the computational cost of the cloud server when performing relevance calculations, making the query time longer. In addition, although coarser grid partitioning can reduce the number of grids, it will reduce the accuracy of spatial data representation, requiring the cloud server to calculate more irrelevant grids, which in turn increases the query time.

[0077] like Figure 5 As shown in Figure 1, this application plots the running time variation of Top-k queries under different query range strategies. In the experiment, the number of data sets n in the spatial data warehouse D and the grid division granularity G-gridcounts remain unchanged, and only the size of the query range R is changed, that is, the number of grids covered by the query range GR--gridcounts is gradually increased. As the query range R expands, the running time of Top-k queries shows an upward trend. A larger query range requires covering more grids, and the cloud server needs to process more encrypted grid vector-encrypted count vector set tuples L. i Matching and correlation calculations are performed, which increases the amount of calculation. In addition, a larger query range may contain more spatial datasets. The cloud server sorts the datasets and filters out the top k datasets with the highest correlation based on the correlation calculation results in an encrypted environment, further increasing the computational cost of sorting and comparison. However, although a smaller query range can reduce the number of grids and the number of datasets that need to be processed, it may lead to inaccurate query results because it may not cover enough relevant spatial datasets, resulting in missing important data.

[0078] This application reduces the risk of privacy leakage, ensures data security throughout the entire cycle, and improves the accuracy and applicability of results in complex scenarios.

[0079] The foregoing is merely an embodiment of the present invention and is not intended to limit the present invention. It will be apparent to those skilled in the art that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are intended to be included within the scope of the claims of the present invention.

Claims

1. A range-based top-k query method for privacy-preserving spatial datasets, applicable to data owners, cloud servers, and querying users, characterized by: The privacy-preserving spatial dataset Top-k query method specifically includes two stages: data preprocessing stage: completed independently by the data owner, and Top-k query stage: completed collaboratively by the cloud server and the query user. Assume that the spatial data warehouse is denoted as D = {D1, D2, ..., D n }, where D i For a spatial data set containing several two-dimensional location points, the coverage area of all location points in the spatial data warehouse D is evenly divided to generate m×m grids, forming a grid set G = {g1, g2, ..., g m×m }; The data preprocessing stage specifically includes the following steps: Step 1a: The data owner has a set of grids G. i , any dataset D in the statistical spatial data warehouse D j Falling on grid g i Number of internal position points According to the number of location points Generate a two-dimensional vector V j i , specifically: The final generated mesh g i The corresponding count vector set Step 1b: The data owner generates a count vector set based on step 1a. Constructing a sequence of counting vector sets Step 1c: The data owner constructs a secure space index structure L = {L1, L2, ..., L v }, where v≤m×m, any item L in L i ∈L is a tuple consisting of an encrypted grid vector and an encrypted count vector set. Then, each spatial data set in the spatial data warehouse D is encrypted to generate an encrypted data warehouse. And the secure space index structure L and the encrypted data warehouse Upload to cloud server for storage; The Top-k query phase includes the following steps: Step 2a: The query range request initiated by the query user is Q = (R, b, k), where R is the query range, b is the user-defined security parameter, 0≤b≤1, and k is the number of spatial datasets returned by the request. Based on the query range R, a minimum grid set G covering the query range R is generated. R , and initialize the encrypted query vector set Constructing a query trapdoor in To encrypt the parameter vector, the query trapdoor td is sent to the cloud server as a query instruction; Step 2b: The cloud server calculates the query range R and the encrypted data warehouse in a privacy-preserving state based on the query trapdoor td and the security space index structure L. Each encrypted dataset The correlation between them is calculated, and finally the top k encrypted space datasets with the highest correlation are returned as the query result Result.

2. The range-based privacy-preserving spatial dataset Top-k query method according to claim 1, characterized in that: Step 1c: For each grid g in the grid set G i and grid g i The corresponding count vector set in the count vector set sequence S The count vector set For a set of non-zero vectors, generate a pair L of encrypted grid vectors and encrypted count vectors i , specifically including the following steps: Step 1c1: Use Bloom filter BF for grid g i Generate binary grid code vector B i , and encode the binary grid vector B i Encrypt and generate the corresponding encrypted grid code vector Among them, BF uses w hash functions to generate a binary encoding vector of length h; Step 1c2: count vector set Each count vector V in j i Perform encryption processing to generate an encrypted count vector Finally, the corresponding encrypted count vector set is generated Step 1c3: Use the encrypted grid code vector generated in step 1c1 above And the encrypted count vector set generated in step 1c2 Construct a pair of encrypted grid vector and encrypted count vector set Encrypted grid vector-encrypted count vector set tuple L i With grid g i correspond.

3. The range-based privacy-preserving spatial dataset Top-k query method according to claim 1, characterized in that: Step 2a: Constructing a query trapdoor includes the following steps: Step 2a1: Minimize the grid set G R Each grid g in i , using Bloom filter BF to generate binary query encoding vector Z i , and then encode the binary query vector Z i Encrypt and generate the corresponding encrypted query vector And encrypt the query vector Add query vector collection Step 2a2: For the user-defined security parameter b, generate a two-dimensional parameter vector V p = [1, b], then for the two-dimensional parameter vector V p Perform encryption processing to generate encryption parameter vector Step 2a3: Use the query vector set generated in step 2a1 above and the encryption parameter vector generated in step 2a2 Constructing a query trapdoor 4. The range-based privacy-preserving spatial dataset Top-k query method according to claim 1, characterized in that: The step 2b, determining the final query result Result according to the query trapdoor td and the security space index structure L, specifically includes the following steps: Step 2b1: Encrypt the data warehouse Each encrypted dataset Initialize a relevance Rel i =0; Step 2b2: For the encrypted query vector set Any encrypted query vector in And any constructed encrypted grid vector-encrypted count vector set tuple in the secure space index structure L like This indicates that the successfully matched grid g t Within the query range R, at this time, for the set of encrypted count vectors that are successfully matched The encrypted count vector for each successful match in In terms of Update the encrypted dataset Relevance Rel within the query range R i : Step 2b3, select the relevance Rel i The top k encrypted data sets with the highest values constitute the final query result set Result.

Images