Elastic reverse k-nearest neighbor query privacy computing system and method for multi-attribute data of dense space

Abstract

The invention discloses an elastic reverse k-nearest neighbor query privacy computing system and method for multi-attribute data in a secret state space, and belongs to the technical field of data security. The method comprises the following steps: initializing a system through a trusted center and generating an improved symmetric homomorphic encryption key; a plurality of sensing users execute a non-interactive multi-party key exchange algorithm to generate a shared key; the sensing user encrypts the multi-attribute data containing the spatial position and the reputation value by using the shared key; the query mechanism encrypts the query request to generate a query trap door; and the central server executes RkNN query qualification verification and reputation verification in a secret state environment based on the encrypted data and the query trap door, and screens sensing users meeting conditions. According to the method, the defects that in the prior art, a user needs to be online continuously, reputation indexes are ignored and privacy is leaked are overcome, and safe and efficient multi-attribute RkNN query under the intermittent offline scene of the user is achieved.

Description

Technical Field

[0001] This invention belongs to the field of data security and privacy computing technology, specifically relating to an elastic reverse k-nearest neighbor query privacy computing system and method for dense-state multi-attribute data. Background Technology

[0002] With the rapid development of data mining technologies, the acquisition and processing of large-scale sensing data has significantly improved the efficiency of secure queries for dense spatial data. In such services, reverse k-nearest neighbor (RkNN) queries are a widely used query type, identifying target records from a dataset by evaluating the similarity between a data record and its k nearest neighbors. However, while numerous studies have focused on RkNN queries, most work primarily addresses spatial data or feature values, neglecting the importance of user reputation metrics. Furthermore, privacy breaches have attracted widespread attention, especially when users urgently need to submit their personal sensing data and query requests to third parties for secure processing. Existing RkNN query privacy computation schemes typically distribute keys interactively to ensure data security, but this requires users to be continuously online. However, in application scenarios such as ride-hailing services, passengers (query agents) send ride requests to the platform, while willing drivers (sensing users) submit their information to the platform for passengers to choose a suitable driver. Due to factors such as network outages or driving safety concerns, sensing users often cannot be continuously online and may be intermittently offline. The interactive key distribution method described above lacks universality in such environments because it relies on the user's continuous online status to obtain the key.

[0003] Non-interactive key exchange (NIKE) is a cryptographic algorithm designed to enable multiple entities to negotiate and generate shared keys without direct communication. The concept was first proposed by Diffie and Hellman. In 2013, Freire et al. proposed an innovative NIKE security model, providing corresponding protocols under both the Random Oracle (ROM) model and the standard model. However, these schemes rely on bilinear pairing-based construction, introducing additional computational overhead and limiting their scalability in large-scale systems. Subsequently, Joux proposed the Bilinear Diffie–Hellman hypothesis based on bilinear pairing groups, achieving single-round communication and avoiding the traditional two-round exchange process. However, the Joux scheme still relies on bilinear pairing operations, and in large-scale, high-efficiency applications, the computational complexity of bilinear pairings can become a bottleneck for system performance. To avoid bilinear pairing-based construction, Guo et al. proposed a variant of the NIKE protocol based on the LWE (Learning with Errors) problem, but it still faces technical challenges in terms of efficiency and feasibility. Recently, Afshar et al. proposed a more fine-grained NIKE protocol and explored the construction of three-party and four-party NIKE schemes under weaker assumptions. By combining a random oracle model with algebraic structures, they enhanced the security of the protocol while avoiding the use of bilinear pairings, but also pointed out the limitations of relying solely on algebraic methods when constructing multi-party NIKE protocols.

[0004] RkNN query privacy computation algorithm: The RkNN query privacy computation algorithm utilizes encryption technology to ensure the privacy of sensitive user information during RkNN query execution. In recent years, Zheng et al. proposed RkNN privacy computation schemes based on dense-state spatial feature data, employing MIUR-tree and improved M-tree structures respectively. However, both schemes suffer from access path privacy vulnerabilities and do not introduce a reputation value indicator, which is crucial in task allocation scenarios. To address the limitations of RkNN query processing under high-dimensional data, Zheng et al. proposed the PHRkNN scheme, utilizing the PFR-tree structure to process high-dimensional datasets. Although this scheme improves performance, it remains vulnerable to covert channel attacks due to the involvement of dual cloud servers. Furthermore, Zheng et al. proposed the PRRkNN scheme to support range-based queries. This scheme achieves privacy computation by encoding feature data, but it has shortcomings in multi-attribute data scalability, and the encoding introduces additional computational overhead for high-dimensional matrices.

[0005] In summary, although the NIKE algorithm has potential application value in the field of cryptography, existing research is still mainly focused on two-party or specific application scenarios, and the construction of multi-party NIKE algorithms remains a challenge. In addition, although the above schemes have made significant progress in RkNN query privacy computation research, they still face challenges in multi-attribute data processing and system scalability, especially in application scenarios where users dynamically join or leave the system. Therefore, how to design a secure and non-interactive RkNN query privacy computation method for multi-attribute data in the cryptographic space still faces great challenges. (1) Existing research mainly focuses on the design of non-interactive key exchange algorithms to overcome the limitations of traditional interactive key distribution schemes, especially in cases where users may dynamically join or leave the system. Many scholars have studied NIKE algorithms under different security models, from two-party to multi-party protocol design and their security assumptions. However, in RkNN query scenarios, due to the large number of perceived users involved, achieving a secure and efficient multi-party NIKE algorithm remains a key challenge; (2) For privacy computation of RkNN queries for dense spatial perceived data, various algorithms have been proposed, including several tree-based methods. These methods achieve efficient and accurate RkNN queries by encoding feature vectors or spatial data. However, they usually focus on spatial or feature attributes, ignoring the impact of user reputation indicators. In addition, the construction of tree structures inevitably introduces additional computation and storage overhead. Therefore, constructing an efficient RkNN query algorithm that supports multi-attribute data has become an urgent technical problem to be solved. Summary of the Invention

[0006] To address the problems existing in the prior art, the purpose of this invention is to provide a privacy-preserving computation system and method for elastic reverse k-nearest neighbor query for multi-attribute data in dense space, specifically implemented through the following technical solutions:

[0007] A resilient reverse k-nearest neighbor query privacy computing system for dense-state multi-attribute data, comprising four participants:

[0008] 1) Trusted Center (TA): Responsible for generating the keys required for the operation of the entire system and distributing the corresponding keys to various entities in the system;

[0009] 2) Central Server (CS): Responsible for receiving the perception data and identity ID sent by the data owner and the query request initiated by the querying agency, executing the RkNN query algorithm and sending the final result to the querying agency;

[0010] 3) Perceived Users (SUs): Each perceived user in the system has a set of perceived data and a corresponding reputation value. Each perceived user needs to execute the 2K-NIKE algorithm to share a key. Perceived users respond to received query requests. Calculate its value from all shared datasets. The similarity between neighbors, the perception users who are willing to accept the task encode the dense spatial data to be uploaded to the cloud according to the RkNN query algorithm, encrypt the corresponding information, and finally send it to the central server;

[0011] 4) Query Agency QA: Each query agency in the system is responsible for initiating an RkNN query task to the central server. The query agency encrypts the data into a query trap and sends it to the central server as required.

[0012] Furthermore, the system also includes a system initialization module, a multi-party non-interactive key generation module, and an algorithm construction module.

[0013] A privacy-preserving computation method for elastic reverse k-nearest neighbor query for multi-attribute data in dense-state space includes the following steps:

[0014] 1) System initialization: The trusted center generates public parameters and private keys for the improved symmetric homomorphic encryption algorithm, and distributes the public parameters and keys to authorized users and querying agencies in the system;

[0015] 2) Non-interactive key generation: Performed by multiple perceptual users, a shared symmetric key is generated through negotiation based on their own generated random numbers by running the 2K-NIKE non-interactive key exchange algorithm without real-time communication.

[0016] 3) Data encryption and uploading: This is performed by each sensing user. The original data of the sensing user is encrypted using a shared symmetric key. The original data includes at least spatial location information and user reputation value. The generated ciphertext data is then uploaded to the central server.

[0017] 4) Secure RkNN query: This is executed collaboratively by the query agency and the central server. The query agency uses its key to encrypt the query request into a query trapdoor and sends it to the central server. In a secure environment, the central server uses the ciphertext data and the query trapdoor, along with the homomorphic encryption computation characteristics, to perform reverse k-nearest neighbor query qualification verification and reputation verification, selects a set of qualified users, and returns the encrypted query results to the query agency.

[0018] Furthermore, the specific steps of the 2K-NIKE non-interactive key exchange algorithm in step 2) are as follows:

[0019] 2-1) Sensing User Message Generation: Each sensing user inputs its corresponding random value, treats its size as a 2K vector, calculates and broadcasts its message;

[0020] 2-2) Sensing User Group Message Generation: Each sensing user group finds the common output and calculates the group message based on all broadcast messages;

[0021] 2-3) Multi-party non-interactive key generation: K groups of senses are used to execute the Diffie-Hellman key exchange protocol. After K rounds of iteration, the final shared symmetric key is generated.

[0022] Furthermore, the privacy computation algorithm for the reverse k-nearest neighbor query eligibility verification in step 4) is as follows:

[0023] make , (same ), (same ), (same Then we can deduce that: .

[0024] make , ,in:

[0025]

[0026]

[0027] It can be inferred that: = ,therefore: ;

[0028] Considering the security requirements of privacy-preserving computation, the designed 2K-NIKE algorithm is used to calculate the shared key, and a symmetric encryption algorithm is used to secure it. Encryption is performed, and iSHE is introduced for data re-encryption. The proposed PREC algorithm mainly includes the following algorithms: ,

[0029] a) PREC.Setup( ) Calculate the shared key iSHE Key ;

[0030] b) PREC.TokenGen First, the encryptor uses a symmetric encryption algorithm to encrypt the original data. ,get:( ) 2K-NIKE.Enc( Then, the encryptor expands. get ( ),in , , { } is a random number in the real number field and , Finally, iSHE encryption is invoked. get: ;

[0031] c) PREC.TrapdoorGen Q) First, the encryptor selects a series of random numbers from the real number field. , },in , Next, the encryptor extends Q. ,get ,in , Finally, the iSHE algorithm is used for encryption. Obtain the ciphertext ;

[0032] d) PREC.Filter( ) Enter ciphertext and query traps Actuator calculation: .

[0033] Furthermore, the privacy computation algorithm for reverse k-nearest neighbor query reputation verification in step 4) is as follows:

[0034] make , , Therefore, we can deduce that: [ ] ;

[0035] If we let:

[0036]

[0037] It can be inferred that: ( = ,therefore: ( ;

[0038] The proposed PRRV algorithm employs symmetric encryption and iSHE homomorphic encryption to encrypt data, and generates random numbers to perturb the data. The algorithm mainly includes the following components: ;

[0039] a) PR .Setup( ) : Key generated by calling the PREC algorithm PREC.Setup( );

[0040] b) PR .TokenGen( ) First, the encryptor processes the original data. Encrypt to obtain 2K-NIKE.Enc( Then, the encryptor selects a series of random numbers from the real number field. right After perturbating the data, we get:

[0041] ;

[0042] Next, the encryptor... Re-encrypt to obtain ;

[0043] c)PR .TrapdoorGen( ) The encryptor selects a series of random numbers from the real number field. right After perturbating the data, we get:

[0044] ;

[0045] Then, the encryptor... Encrypt to obtain ;

[0046] d)PR .Refine( ) Enter ciphertext and query traps Actuator calculation: .

[0047] Based on in-depth research and analysis of existing RkNN query algorithms and their high-performance query privacy computing technology and research results, this invention realizes an elastic RkNN query privacy computing method for multi-attribute data in the dense space. This invention innovates a method that can support efficient and secure RkNN query in the intelligent ride-hailing scenario, effectively protecting the privacy computing security of the dense dataset, user query requests and query results. The main innovations include: (1) Non-interactive multi-party key exchange mechanism: In order to meet the needs of sensing users in the intermittent online state to still meet the requirements of key generation, protection of sensing data and user identity privacy, a multi-party non-interactive key exchange algorithm (2K-NIKE) is proposed. This algorithm combines the NIKE-based protocol and the Diffie-Hellman two-party key exchange mechanism, which not only realizes efficient key reception among multiple sensing users, but also uses the generated key and symmetric encryption algorithm to encrypt the sensing data and user identity identifier; (2) Secure RkNN query and reputation verification algorithm: In order to enhance the security of dense space data, the judgment conditions of RkNN query are formally defined first, and the original sensing data and query request are encoded. Based on this, utilizing the homomorphic properties of the improved symmetric homomorphic encryption algorithm (iSHE), the RkNN query qualification verification privacy computation algorithm (PREC) and the RkNN query reputation verification privacy computation algorithm (PRRV) were designed. The above algorithms ensure the privacy of multi-attribute data during the RkNN query process and can verify the correctness of the query results in the dense space; (3) Secure dense space elastic RkNN query privacy computation scheme: In order to realize the elastic RkNN secure query in the dense space, an elastic RkNN query privacy computation method based on multi-attribute data in the dense space was designed. The formal security of the method in the simulation environment was rigorously proved; at the same time, for the proposed 2K-NIKE algorithm, it was proven that it satisfies - Passive security, meaning that when an attacker initiates a maximum of q queries, the advantage of its cracking algorithm is less than... . Attached Figure Description

[0048] Figure 1 This is a system framework diagram of the present invention;

[0049] Figure 2 The diagram shows the algorithm block diagram and data flow diagram of this invention. Detailed Implementation

[0050] The present invention will be further described below with reference to the accompanying drawings to provide a better understanding of the technical solution.

[0051] This invention is based on common intelligent ride-hailing application scenarios, and the overall system architecture is as follows: Figure 1 As shown, it includes the following four participants:

[0052] (1) Trusted Center (TA): Responsible for generating the keys required for the operation of the entire system and distributing the corresponding keys to various entities in the system.

[0053] (2) Central Server (CS): Provides RkNN query services to querying agencies. Specifically, the CS receives the perception data and identity ID sent by the data owner and the query request sent by the querying agency, executes the RkNN query algorithm, and sends the final result to the querying agency.

[0054] (3) Perceiving Users (SU): There are multiple perceiving users in the system. Each Each has a set of perceived data and a corresponding reputation value, represented as: { }, ,in , ,and Furthermore, each sensing user in the system needs to perform the 2K-NIKE algorithm to share the key. Then, the sensing user responds to the received query request. Calculate its value from all shared datasets. Neighbor Similarity between Users willing to accept the task encode the dense spatial data to be uploaded to the cloud according to the RkNN query algorithm, encrypt the corresponding information, and finally send it to the central server.

[0055] (4) Inquiry Agency (QA): There are multiple inquiry agencies in the system. Each query agency is responsible for initiating an RkNN query task to the CS. ,[ ]},in, Indicates the location where the task is executed. This represents the characteristics that need to be met to perform a task. This indicates that QA requires a minimum credit score for users who can perform the task. This represents the credit score corresponding to the maximum fee that QA pays to the user performing the task. However, to ensure query privacy, QA is required to encrypt the data into a query trap before sending it to CS.

[0056] The system consists of the following three modules, and the algorithm flowchart and data flow diagram are as follows: Figure 2 As shown:

[0057] (1) System initialization module

[0058] The Trusted Center performs the following operations to initialize the system:

[0059] 1) Symmetric homomorphic encryption key generation

[0060] a) Generate security parameters based on the key generation algorithm of the improved symmetric homomorphic encryption algorithm (iSHE). ( ),make The message space is set to Then, randomly select two. Find prime numbers p and q in -bit, and let Furthermore, randomly select a prime number. Finally, choose -bit random number and obtain common parameters. private key .

[0061] b) Output system common parameters private key .

[0062] 2) Data encryption

[0063] Based on iSHE's data encryption algorithm, select random numbers. and ,in , Using iSHE's key The message m can be encrypted, such as mod .

[0064] 3) Key distribution

[0065] Trust Center will To make public Send them to the authorized SU and QA respectively.

[0066] (2) Multi-party non-interactive key generation module

[0067] 1) Algorithm initialization 2K-NIKE.Setup( )

[0068] Without loss of generality, Represented as ,in and They are two isomorphic groups, with sizes of respectively and Enter security parameters Random Oracle Size is positive integer group (Generator is) ), and a large prime number .make , which represents the number of elements in the minimum intersection between sets.

[0069] 2) Generating 2K-NIKE.Msg by sensing user messages ( )

[0070] For each perceptive user Enter its corresponding random value Size is The specific message generation algorithm process is as follows:

[0071] a) will Consider it as a size of vector For each Treat its corresponding element as ,in Corresponding to The first K and last K elements. Then, calculate , .

[0072] b) Output and publish the generated message: .

[0073] 3) Sensing user group message generation 2K-NIKE.Sharedmsg( )

[0074] make For each perceptive user Enter its index The corresponding random value and messages output by other users within the system. The algorithm flow for generating user group messages is as follows:

[0075] a) Call the message generation algorithm to calculate its own message. 2K-NIKE and based on the message The message is parsed using the corresponding index value, that is... .

[0076] b) If it exists ,make If the condition is met, the algorithm will terminate or output 0.

[0077] c) Otherwise, let Indicates lexicographical order The first common output of each pair of perceptual user groups. You can find the corresponding random value ,make Known , The Baby-Step Giant-Step (BSGS) algorithm can be used to find [something related to] ... The corresponding last K random values, i.e. from Found Perceived User Group The public random value is represented as Group messages are represented as s .

[0078] d) K-group of perceptual users Each group will release its own information. .

[0079] 4) Multi-party non-interactive key generation 2K-NIKE.Key( )

[0080] Index of input-aware user groups ,and K groups of users The key generation algorithm for executing the Diffie-Hellman key exchange protocol is as follows:

[0081] a) If Perceive user group Output its shared key .

[0082] b) If Perceive user group Output its shared key according to the Diffie-Hellman two-party key exchange protocol. , .

[0083] c) K needs to be performed. Round key exchange, i.e., awareness user group ( Output the final shared key .

[0084] d) Let Then the shared key .

[0085] 5) Data encryption algorithm 2K-NIKE.Enc( )

[0086] Each perceptive user Execute the key generation algorithm to obtain 2K-NIKE.Key( Finally, a symmetric encryption algorithm is used to verify the original information of the perceived user. Encrypt, obtain .

[0087] 6) Data decryption algorithm 2K-NIKE.Dec( )

[0088] The query agency QA will provide the query results. Decryption, obtain .

[0089] (3) Algorithm construction module

[0090] 1) RkNN Query Qualification Validation (PREC) Privacy Computation Algorithm

[0091] To ensure security, determine whether a user seeking to accept a task meets the similarity requirements of the query request, i.e., whether they meet the requirements. A privacy-preserving computation algorithm for RkNN query eligibility verification (PREC) is designed. First, the conditions for satisfying RkNN filtering queries are defined as follows: .make Therefore, we can deduce that:

[0092] ,

[0093] ,

[0094] ,

[0095] ,

[0096] ,

[0097] .

[0098] make (same ), (same ), (same )but: .

[0099] If let , ,in:

[0100]

[0101]

[0102] It can be inferred that: = ,therefore: .

[0103] Considering the needs of privacy computing applications, a design is adopted. The algorithm calculates the shared key and uses a symmetric encryption algorithm to... Encryption is performed. To further enhance the algorithm's security, iSHE is introduced to re-encrypt the data. The proposed PREC algorithm mainly includes the following algorithms, namely... .

[0104] a) PREC.Setup( ) During the initialization phase, the shared key is first calculated according to the 2K-NIKE algorithm. 2K-NIKE Next, input the iSHE security parameters and the output key. .

[0105] b) PREC.TokenGen During the instruction generation phase, the TokenGen algorithm uses... and encryption First, the encryptor uses a symmetric encryption algorithm to encrypt the original data. Encryption yields: ( ) 2K-NIKE.Enc( Then, the encryptor selects a series of random numbers from the real number field. },in , Next, the encryptor expands. get ( ),in , Finally, the iSHE algorithm is called to encrypt ( The ciphertext was obtained: .

[0106] c) PREC.TrapdoorGen Q) During the query generation phase, the TrapdoorGen algorithm uses i Encrypted query request Q First, the encryptor selects a series of random numbers from the real number field. , },in , Next, the encryptor extends Q. ,get ,in , Finally, the iSHE algorithm is used for encryption. Obtain the ciphertext .

[0107] d) PREC.Filter( ) During the query phase, the filtering algorithm inputs ciphertext. and query traps Actuator calculation: .

[0108] 2) RkNN Query Reputation Verification (PRRV) Privacy Computation Algorithm

[0109] To ensure security, it determines whether the perceived user who wants to accept the task meets the reputation value requirements of the query request, i.e., whether they meet the requirements. [ A privacy-preserving algorithm for RkNN query reputation verification (PRRV) is designed. First, the conditions for satisfying RkNN filtering queries are defined as follows:

[0110] [ ],

[0111] ,

[0112] ,

[0113] .

[0114] make , , but: [ ] .

[0115] If we let:

[0116]

[0117] It can be inferred that: ( = ,therefore: ( .

[0118] The data is encrypted using symmetric encryption and iSHE homomorphic encryption, and random numbers are generated to perturb the data. The designed PRRV algorithm mainly includes the following algorithms: .

[0119] a) PR .Setup( ) During the Setup phase, the Setup algorithm calls the key generated by the PREC scheme. PREC.Setup( ).

[0120] b) PR .TokenGen( ) During the instruction generation phase, the TokenGen algorithm uses... First, the encryptor uses a symmetric encryption algorithm to encrypt the original data. Encrypt to obtain 2K-NIKE.Enc( Then, the encryptor selects a series of random numbers from the real number field. right After perturbating the data, we get:

[0121] ;

[0122] Next, the encryptor... Encrypt to obtain .

[0123] c)PR .TrapdoorGen( ) During the query generation phase, the TrapdoorGen algorithm uses i Encrypted query First, the encryptor selects a series of random numbers from the real number field, including... right After perturbating the data, we get:

[0124] ;

[0125] Then, the encryptor... Encrypt to obtain .

[0126] d)PR .Refine( ) During the query phase, the algorithm input ciphertext is refined. and query traps Actuator calculation: .

[0127] Technical features of the system of this invention:

[0128] (1) In the key generation stage: On the one hand, to meet the needs of sensing users to complete key generation even when they are intermittently online, a multi-party non-interactive key exchange algorithm (2K-NIKE) was designed. This algorithm combines the NIKE-based protocol with the Diffie-Hellman two-party key exchange mechanism to achieve efficient key reception among multiple sensing users; on the other hand, the generated shared key and symmetric encryption algorithm are used to encrypt the sensing data and user identity, thereby ensuring the privacy and security of the sensing data and users while achieving efficient key distribution.

[0129] (2) In the secure RkNN query and reputation verification stage: On the one hand, the original perceived data and query requests are encoded by formalizing the judgment conditions of RkNN queries to support privacy computation of multi-attribute queries; on the other hand, by utilizing the homomorphic characteristics of the improved symmetric homomorphic encryption algorithm (iSHE), the RkNN query qualification verification privacy computation algorithm (PREC) and the RkNN query reputation verification privacy computation algorithm (PRRV) are designed, which realizes the correctness verification of query results in the dense space, thereby effectively ensuring the security and integrity of multi-attribute data in the RkNN query process.

[0130] (3) In the secure and resilient RkNN query stage: On the one hand, a resilient RkNN query privacy computation method based on multi-attribute data in the dense space is designed, realizing secure and robust RkNN queries in the ciphertext environment; on the other hand, the formal security of this method in the simulated environment is rigorously proven, and it is also proven that the proposed 2K-NIKE algorithm satisfies (ϵ,q)-passive security, that is, when the attacker initiates a maximum of q queries, the advantage of its cracking algorithm is less than ϵ. This scheme not only has formal security guarantees, but also maintains high efficiency and scalability in a multi-user resilient access environment.

Claims

1. An elastic reverse k-nearest neighbor query privacy computing system for ciphertext space multi-attribute data, characterized in that, The system comprises four participants: 1) a trusted center TA: responsible for generating the keys required for the operation of the entire system, and distributing the corresponding keys to each entity in the system; 2) a central server CS: responsible for receiving the perception data and identity ID sent by the data owner and the query request initiated by the query agency, executing the RkNN query algorithm and sending the final result to the query agency; 3) Perceived Users (SUs): Each perceived user in the system has a set of perceived data and a corresponding reputation value. Each perceived user needs to execute the 2K-NIKE algorithm to share a key. Perceived users respond to received query requests. Calculate its value from all shared datasets. The similarity between neighbors, the perception users who are willing to accept the task encode the dense spatial data to be uploaded to the cloud according to the RkNN query algorithm, encrypt the corresponding information, and finally send it to the central server; 4) a query agency QA: each query agency in the system is responsible for initiating the RkNN query task to the central server, and the query agency encrypts the data into a query trapdoor according to the requirements and sends it to the central server.

2. The privacy computing system for elastic reverse k-nearest neighbor query on ciphertext spatial multi-attribute data according to claim 1, wherein, The system also comprises a system initialization module, a multi-party non-interactive key generation module and an algorithm construction module.

3. An elastic reverse k-nearest neighbor query privacy computing method for ciphertext space multi-attribute data, characterized in that, The method comprises the following steps: 1) system initialization: the improved symmetric homomorphic encryption algorithm public parameters and private keys are generated by the trusted center, and the public parameters and keys are distributed to authorized perception users and query agencies in the system; 2) non-interactive key generation: performed by multiple perception users, a shared symmetric key is generated by running the 2K-NIKE non-interactive key exchange algorithm based on the random numbers generated by each user without real-time communication; 3) data encryption and uploading: performed by each perception user, the original data of the perception user is encrypted using the shared symmetric key, and the original data at least includes spatial location information and user reputation value, and the generated ciphertext data is uploaded to the central server; 4) secure RkNN query: performed by the query agency and the central server, the query agency encrypts the query request into a query trapdoor using its key and sends it to the central server, the central server performs reverse k-nearest neighbor query qualification verification and reputation verification in a ciphertext environment based on the ciphertext data and the query trapdoor, uses the homomorphic encryption calculation characteristics to filter out the perception user set meeting the conditions, and returns the encrypted query result to the query agency.

4. The privacy computing method for elastic reverse k-nearest neighbor query of ciphertext space multi-attribute data according to claim 3, wherein, The specific steps of the 2K-NIKE non-interactive key exchange algorithm in step 2) are as follows: 2-1) perception user message generation: each perception user inputs its corresponding random value, takes its size as a 2K vector, calculates and broadcasts its message; 2-2) perception user group message generation: each perception user group finds the public output according to all broadcasted messages and calculates the group message; 2-3) multi-party non-interactive key generation: K groups of perception users perform the Diffie-Hellman key exchange protocol, and after K rounds of iteration, the final shared symmetric key is generated.

5. The privacy-preserving computation method for elastic reverse k-nearest neighbor query on multi-attribute data in ciphertext space according to claim 1, wherein, The privacy calculation algorithm for reverse k-nearest neighbor query qualification verification in step 4) is as follows: Let , (same as ), (same as ), (same as ), it can be inferred that: ; Let , wherein: It can be inferred that: = Therefore: ;​ Considering the security requirements of privacy computing, the 2K-NIKE algorithm is used to calculate the shared key, and the symmetric encryption algorithm is used to encrypt At the same time, iSHE is introduced to re-encrypt the data, and the PREC algorithm mainly includes the following algorithms, that is , a) PREC.Setup ) : Compute shared key iSHE's key ; b) PREC. TokenGen : First, the encryptor encrypts the original data using a symmetric encryption algorithm , obtaining: ) 2K-NIKE. Enc( ); then, the encryptor extends to obtain ( ), where , , { } are random numbers in the real number field and , ; finally, the iSHE encryption is called to obtain: ; c) PREC. TrapdoorGen (Q) , : First, the encrypter selects a series of random numbers from the real number field , , where , ; then, the encrypter extends Q to obtain , where , ; finally, the iSHE algorithm is called to encrypt to obtain the ciphertext ; d) PREC.Filter ) : input ciphertext and query trapdoor , the executor computes: .

6. The privacy-preserving computation method for elastic reverse k-nearest neighbor query on multi-attribute data in ciphertext space according to claim 1, wherein, The privacy calculation algorithm for reverse k-nearest neighbor query reputation verification in step 4) is as follows: Let , , , it can be inferred that: [ ] ; If It can be concluded that: ( = Therefore: ( ; The data is encrypted by using symmetric encryption technology and iSHE homomorphic encryption technology, and random numbers are generated for data perturbation, and the PRRV algorithm mainly includes the following algorithms, namely ; a) PR .Setup( ) : call the PREC algorithm to generate the key PREC.Setup( ); b) PR .TokenGen( ) : First, the encrypter encrypts the original data to get 2K-NIKE.Enc( ); then, the encrypter selects a series of random numbers in the real number field to disturb the data to get: ； Next, the encrypter re-encrypts , obtaining ; c) PR TrapdoorGen( ) : encrypter selects a series of random numbers from the real numbers field data perturbation on , resulting in: ； Then, the encrypter encrypts to obtain ; d) PR .Refine( ) : input ciphertext and query trapdoor , the executor computes: .

Images