Method and device for realizing cross-institutional joint risk control based on secure multi-party graph computation, processor and computer readable storage medium thereof

Abstract

The present application relates to a kind of method for realizing cross-institutional joint risk control based on secure multi-party graph computing, comprising the following steps: graph-based staining mark and query under privacy protection;Graph-based huddling query under privacy protection;Complete huddling query based on graph under non-privacy protection;Complete huddling query based on graph under privacy protection.The method, device, processor and computer readable storage medium thereof for realizing cross-institutional joint risk control based on secure multi-party graph computing of the present application guarantee the privacy of users, the authenticity and integrity of messages and are suitable for two-party graph joint keyword privacy query scheme;Further, the participating party completes the huddling query of oneself using the participating party graph;On the premise of privacy protection, double-participating-party complete huddling query operation is realized, which not only can protect the security and privacy of each node of the graph data of double participating party, but also can protect the privacy of the association between them.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of privacy security, in particular to the field of multi-party secure graph computation, and specifically refers to a method and device for realizing cross-institutional joint risk control based on secure multi-party graph computation, a processor and a computer readable storage medium thereof. BACKGROUND

[0002] Multi-party secure graph computation is a research direction that has emerged in recent years. Before 2024, the focus was mainly on the privacy security of data itself, and subsequent research gradually expanded to new fields such as association security and heterogeneous network security. At the technical application level, researchers have proposed a variety of innovative solutions: a blockchain-based SMPC protocol solves the privacy protection problem in the organ donation scenario; the MPC-DP hybrid method for decision trees strikes a balance between precision and cost; in the field of graph neural networks, a traffic prediction scheme combining differential privacy and a vertical federated architecture integrating federated learning and homomorphic encryption have emerged. In addition, scholars have also explored new technologies such as graph sparsification secure aggregation and heterogeneous graph privacy protection, among which composite circuits and obfuscation schemes provide new ideas for server-assisted protocols. To meet the special needs of dynamic network graphs, a lightweight privacy protection framework has been proposed for epidemic modeling, enabling the localized processing of contact data. These studies have collectively pushed the development of graph data secure computation from static scenarios to dynamic and lightweight directions.

[0003] At the same time, in recent years, graph databases have been widely used as a new and powerful semi-structured storage mode, and related work has been proposed for how to efficiently and securely perform joint queries on multi-party graph databases. Based on the oblivious sorting algorithm, an oblivious scatter and gather algorithm is proposed, and an efficient parallel algorithm is designed to utilize the computing power of multiple processors. Under the scatter and gather algorithm framework, problems such as PageRank depth-first search and maximum independent set can be efficiently computed. Araki et al. noticed that the oblivious sorting results in GraphSC can be reused, and thus designed a query framework based on the secret sharing framework and random permutation protocol. For the attribute graph matching problem, an efficient and secure algorithm is designed based on three mutually non-colluding semi-honest computing parties. The core idea is to use function secret sharing to complete range queries and size comparison operations, encrypt and randomly permute the graph node IDs that meet the conditions, and then publicly disclose them. Through the oblivious algorithm, further condition filtering is performed on the neighbor nodes of the nodes that meet the conditions, and finally the matching result is obtained.

[0004] In summary, the current existing technologies mainly include PSI, PSU, BPSI, PIR, keyword anonymous searchable encryption, vertical and horizontal federated learning technology framework, etc. The current application can solve the problems of small-scale graph data in two parties, such as secret data aggregation, secret matching, secret maximum independent set calculation and secret depth-first search. However, under the current association security requirements, there is still a lack of relevant research on the secure association query and association cluster marking of two parties on the graph, including: 1. The current solution only involves two-party collaborative anonymous query based on KeyWords, which can only guarantee node privacy security, cannot guarantee association relationship security, and does not involve knowledge graph based on complex network and other related prerequisites; 2. The cluster query and complete cluster query under the current privacy protection cannot include the possible association nodes into the marked group, and can only guarantee the node privacy security, and cannot guarantee the association relationship security. SUMMARY

[0005] The purpose of the present application is to overcome the shortcomings of the prior art, and to provide a method, device, processor and computer readable storage medium based on secure multi-party graph calculation for realizing cross-institutional joint risk control, which satisfies high security, good privacy and wide application range.

[0006] In order to achieve the above purpose, the method, device, processor and computer readable storage medium based on secure multi-party graph calculation for realizing cross-institutional joint risk control of the present application are as follows:

[0007] The method for realizing cross-institutional joint risk control based on secure multi-party graph calculation mainly comprises the following steps:

[0008] (1) It is judged whether it is under privacy protection, if yes and black list dyeing query is needed, step (2) is continued; if yes and cluster marking query is needed, step (3) is continued; if yes and complete cluster marking query is needed, step (4) is continued; if no, step (5) is continued;

[0009] (2) Dyeing marking and query based on graph under privacy protection, participant A queries the related information of the user with ID associated with the user in the participant through the graph structure of participant B, and the step is ended;

[0010] (3) Cluster query based on graph under privacy protection, participant A queries the association between the user with ID and the user through the graph structure of participant B, and the step is ended;

[0011] (4) Complete cluster query based on graph under privacy protection, step (5) is continued;

[0012] (5) In the non-privacy protection, the complete cluster query is carried out based on the graph, and the participant A carries out the complete cluster marking task through the graph structure of the participant B.

[0013] Preferably, the step (2) specifically comprises the following steps:

[0014] (2.1) The participant B constructs the set of the associated edges of the participant A in the encrypted graph, and the participant B adds the associated edges of the participant A in the encrypted graph to the set of the associated edges of the participant B in the encrypted graph to obtain the set of the associated edges of the participant B in the encrypted graph. The user involved in the step (2.1) is the user involved in the step (1.1). The corresponding data set of the user involved in the step (2.1) is the corresponding data set of the user involved in the step (1.1). The step (2.1) further comprises the following steps: The step (2.1) further comprises the following steps:

[0015] (2.2) The participant A generates a public-private key pair and the corresponding encryption and decryption functions thereof based on the CKKS homomorphic encryption.

[0016] (2.3) The participant A sends the ciphertext to the participant B.

[0017] (2.4) The participant B calculates and and sends them to the participant A.

[0018] (2.5) After receiving the messages, the participant A decrypts and , and if is 0, then is the associated index primary key to be queried and , wherein is the tth associated word m in the plurality of associated words m, that is, the effective query keyword corresponding to the correct associated word.

[0019] Preferably, the step (3) specifically comprises the following steps:

[0020] (3.1) The participant B establishes the set of the associated edge pairs of any two nodes in the graph by , and the participant A publishes a set of public-private key pairs and , wherein is the index primary key set of the participant B. ​​​a set of graph association relations of the index primary key of the participant B;

[0021] (3.2) Participant B hashes and signs all data in the set of graph association relations of the index primary key of participant A and constructs a product polynomial;

[0022] (3.3) Define a temporary edge pair as Hash and sign the edge pair to obtain Encrypt the edge pair into by the public key of participant A and send it to participant B.

[0023] (3.4) Participant B substitutes the constructed product polynomial to calculate Send c back to participant A;

[0024] (3.5) Participant A decrypts c, if is 0, there is association, otherwise, there is no association.

[0025] Preferably, the step (5) specifically comprises the following steps:

[0026] (5.1) Define the index primary key set of participant A as , the set of graph association relations of the index primary key as , the index primary key set of participant B as , and the set of graph association relations of the index primary key as ;

[0027] (5.2) Define the index primary key set of participant A at risk as , the index primary key set of participant B at risk as , and both are empty sets initially;

[0028] (5.3) Participant A takes the marked point from and puts it into until all points in are traversed;

[0029] (5.4) Participant A defines a temporary index primary key association relation set , takes each time until the elements in remain unchanged for a cycle, and obtains , which is the index primary key set of participant A at risk;

[0030] (5.5) Participant A takes and ​The message is sent to participant B. After receiving the message, participant B defines a temporary set of indexed primary keys. and a temporary set of indexed primary key relationships. , definition take and ,Pick ,until The elements in the set remain unchanged and continue for one cycle, resulting in a new set of risk points. ,Will Return to participant A;

[0031] (5.6) Participant A calculates after receiving the message. .

[0032] Preferably, in step (5.4), each time... Specifically, it includes the following steps:

[0033] If a and b belong to... Then add the other one as well. In the middle, and from this edge Delete;

[0034] If both a and b belong to Then directly remove this edge from Delete;

[0035] If neither a nor b belongs to Then place the edge on The last part of the sequence.

[0036] Preferably, in step (5.5), each time... Specifically, it includes the following steps:

[0037] If there exists only one of a and b that belongs to and another one that does not belong to Then it will not belong to Add point The queue, and its edge from Delete; if the point belongs to Then add it. middle;

[0038] If both a and b belong to Then directly remove this edge from Delete;

[0039] If neither a nor b belongs to Or in other cases, place the edge... The last part of the sequence.

[0040] Preferably, the step (4) specifically comprises the following steps:

[0041] (4.1) defining the index primary key set of the participant A as , the graph association relationship set of the index primary key as , the index primary key set of the participant B as , the graph association relationship set of the index primary key as , the hash value set of the index primary key of the participant A as , the hash value set of the graph association relationship of the index primary key as , the hash value set of the index primary key of the participant B as , the hash value set of the graph association relationship of the index primary key as ;

[0042] (4.2) defining the index primary key set at risk of the participant A as , the index primary key set at risk of the participant B as , both of which are initially empty sets;

[0043] (4.3) the participant A takes out the marked point from and puts it into until all points in are traversed;

[0044] (4.4) the participant A defines a temporary index primary key association relationship set , each time taking until the elements in remain unchanged for a cycle, obtaining , which is the index primary key set at risk of the participant A;

[0045] (4.5) the participant A and the participant B generate a temporary shared key based on the timestamp T and the identities and of the users of the two parties through key exchange;

[0046] (4.6) the participant A randomly selects a , , and XORs the hash value sets and of and with the temporary shared key to form two sets and sends them to the participant B;

[0047] (4.7) Party B defines n temporary index primary key sets and n temporary index primary key association sets After receiving the information, Party B decrypts it with the shared key, performs integrity and authenticity verification, and obtains the real concatenated information after verification, and recovers and ;

[0048] (4.8) Party B directly takes n hash value sets by redefining and ;

[0049] (4.9) Starting from , take , until the elements in start to remain unchanged and last for a cycle, to obtain a new risk point set , and all are concatenated in order to obtain , and returned to Party A;

[0050] (4.10) Party A takes the tth calculation according to the t value after receiving the message, finds the corresponding , completes the complete hugging query activity, All index primary keys involved in the hugging mark in are .

[0051] Preferably, in step (4.9), the , specifically comprising the following steps:

[0052] If and only one of them belongs to and the other does not belong to , the point not belonging to is added to the queue of , and its edge is deleted from , if the point belongs to , it is added to ;

[0053] If both a and b belong to , directly delete its edge from ;

[0054] If neither a nor b belongs to or other conditions, the edge is placed at the end of the sequence.

[0055] The device for realizing cross-institutional joint risk control based on secure multi-party graph calculation has the following main features: the device comprises

[0056] a processor configured to execute computer executable instructions;

[0057] a memory storing one or more computer executable instructions, which, when executed by the processor, realize the steps of the method for realizing cross-institutional joint risk control based on secure multi-party graph calculation.

[0058] The processor for realizing cross-institutional joint risk control based on secure multi-party graph calculation has the following main features: the processor is configured to execute computer executable instructions, which, when executed by the processor, realize the steps of the method for realizing cross-institutional joint risk control based on secure multi-party graph calculation.

[0059] The computer readable storage medium has the following main features: a computer program is stored on the computer readable storage medium, and the computer program can be executed by a processor to realize the steps of the method for realizing cross-institutional joint risk control based on secure multi-party graph calculation.

[0060] The method, device, processor and computer readable storage medium for realizing cross-institutional joint risk control based on secure multi-party graph calculation of the application guarantee the privacy of users, the authenticity and integrity of messages and are suitable for a two-party graph joint keyword privacy query scheme; further, participating parties can complete self-party huddling query by using a participating party graph; by using the greedy idea, an accurate and complete huddling query on a plaintext can be realized; based on the huddling query scheme on the plaintext, a group marking under privacy protection is realized by combining PIR technology based on PIS, OT, a hash function and other technologies, so that a double-participating-party complete huddling query operation is realized under the premise of privacy protection. The scheme is analyzed in terms of correctness and security. In the above-mentioned privacy protection scheme process, the safety and privacy of each node of the graph data of the double participating parties can be protected, and the privacy of the association relationship between them can also be protected; on the basis of privacy protection, accurate and complete huddling query calculation can also be realized. BRIEF DESCRIPTION OF DRAWINGS

[0061] Figure 1 The structure diagram of the dyeing marking involved in the dyeing marking and query based on a graph under privacy protection of the method for realizing cross-institutional joint risk control based on secure multi-party graph calculation of the application.

[0062] Figure 2 The structure diagram of the huddling query involved in the huddling query based on a graph under privacy protection of the method for realizing cross-institutional joint risk control based on secure multi-party graph calculation of the application.

[0063] Figure 3a The structural diagram of the complete huddling query involved in the complete huddling query under the non-privacy protection based on the graph for the method of realizing cross-institutional joint risk control based on secure multi-party graph calculation of the application.

[0064] Figure 3b The structural diagram of the complete huddling query involved in the complete huddling query under the privacy protection based on the graph for the method of realizing cross-institutional joint risk control based on secure multi-party graph calculation of the application.

[0065] Figure 4 The flowchart of the dyeing marking and query under the privacy protection based on the graph for the method of realizing cross-institutional joint risk control based on secure multi-party graph calculation of the application.

[0066] Figure 5 The flowchart of the huddling query under the privacy protection based on the graph for the method of realizing cross-institutional joint risk control based on secure multi-party graph calculation of the application.

[0067] Figure 6 The flowchart of the complete huddling query under the non-privacy protection and the privacy protection based on the graph for the method of realizing cross-institutional joint risk control based on secure multi-party graph calculation of the application.

[0068] Figure 7 The workflow diagram for the method of realizing cross-institutional joint risk control based on secure multi-party graph calculation of the application.

[0069] Figure 8 The specific interaction flowchart for the method of realizing cross-institutional joint risk control based on secure multi-party graph calculation of the application. DETAILED DESCRIPTION

[0070] In order to enable the technical content of the application to be more clearly described, further description will be made in combination with specific embodiments.

[0071] The method of realizing cross-institutional joint risk control based on secure multi-party graph calculation of the application, wherein the following steps are included:

[0072] (1) judging whether under privacy protection, if yes and black list dyeing query is needed, then continue to step (2); if yes and huddling marking query is needed, then continue to step (3); if yes and complete huddling marking query is needed, then continue to step (4); if no, then continue to step (5);

[0073] (2) dyeing marking and query under the privacy protection based on the graph, the participant A queries the related information of the user with the participant ID of the user in the participant associated with the user

[0074] (3) Under privacy protection, group queries are performed based on the graph. Participant A queries the ID through the graph structure of participant B. users and The steps are completed by verifying the connections between users.

[0075] (4) Perform a full cluster query based on the graph under privacy protection, and continue to step (5);

[0076] (5) Under non-privacy protection, perform full cluster query based on graph, and participant A performs full cluster labeling task through the graph structure of participant B.

[0077] In a preferred embodiment of the present invention, step (2) specifically includes the following steps:

[0078] (2.1) Participant B will The users involved Corresponding dataset Construct a product polynomial, perform Lagrange interpolation on the dataset to obtain the specific coefficients, where... This refers to the set of associated edges in the dense graph transmitted by participant A, augmented by participant B through the graph database; x refers to the keyword that participant A needs to query; and m refers to the associated word m corresponding to the keyword x that participant A queries through participant B's Lagrange interpolation function.

[0079] (2.2) Participant A generated a public-private key pair and its corresponding encryption and decryption functions based on CKKS homomorphic encryption. and ;

[0080] (2.3) Participant A will send the encrypted message Send to participant B;

[0081] (2.4) Participant B's calculation and And send it to participant A;

[0082] (2.5) After receiving the message, participant A decrypts it. and ,if If it is 0, then After decryption, the primary key of the related index to be queried is obtained. ,in, The term refers to the t-th related word m among multiple related words, i.e., the effective query keyword. The correct conjunctions.

[0083] In a preferred embodiment of the present invention, step (3) specifically includes the following steps:

[0084] (3.1) Party B establishes an association edge pair between any two nodes in ; ; , Party A publishes a set of public and private key pairs and , wherein, denotes the index primary key set of Party B, denotes the graph association relationship set of the index primary key of Party B;

[0085] (3.2) Party B hashes and signs all data in and constructs a product polynomial;

[0086] (3.3) A temporary edge pair is defined as , and the edge pair is hashed and signed to obtain , which is encrypted by the public key of Party A into , and sent to Party B.

[0087] (3.4) Party B substitutes the constructed product polynomial to calculate , and sends c back to Party A;

[0088] (3.5) Party A decrypts c, and if is 0, there is an association, otherwise, there is no association.

[0089] As a preferred embodiment of the present application, the step (4) specifically comprises the following steps:

[0090] (4.1) The index primary key set of Party A is defined as , the graph association relationship set of the index primary key is defined as , the index primary key set of Party B is defined as , and the graph association relationship set of the index primary key is defined as ;

[0091] (4.2) The index primary key set of Party A at risk is defined as , and the index primary key set of Party B at risk is defined as , both of which are initially empty sets;

[0092] (4.3) Party A takes out the marked point from and puts it into until all points in are traversed;

[0093] (4.4) Party A defines a temporary index primary key association relationship set , until the elements in remain unchanged for one cycle, obtaining , that is, the set of all risk-exposed index primary keys in the participant A;

[0094] (4.5) The participant A sends and to the participant B, and after the participant B receives the message, defines a temporary set of index primary keys and a temporary set of index primary key association relations , defines and , takes , until the elements in remain unchanged for one cycle, obtaining a new set of risk-increasing points , and returns to the participant A;

[0095] (4.6) After receiving the message, the participant A calculates .

[0096] As a preferred embodiment of the present application, the step (4.4) of taking each time specifically comprises the following steps:

[0097] If only one of a and b belongs to , the other one is also added to , and the edge of the other one is deleted from ;

[0098] If both a and b belong to , the edge of the other one is directly deleted from ;

[0099] If neither a nor b belongs to , the edge is placed at the end of ;

[0100] As a preferred embodiment of the present application, the step (4.5) of taking each time specifically comprises the following steps:

[0101] If only one of a and b belongs to and the other one does not belong to , the point not belonging to is added to the queue of , and the edge of the point is deleted from ; if the point belongs to , the point is added to​ In;

[0102] If both a and b belong to , directly delete this edge from ;

[0103] If neither a nor b belongs to or other cases, place this edge at the end of sequence.

[0104] As a preferred embodiment of the present application, the step (5) specifically comprises the following steps:

[0105] (5.1) defining the index primary key set of the participant A as , the graph association relationship set of the index primary key as , the index primary key set of the participant B as , the graph association relationship set of the index primary key as , defining the hash value set of the index primary key of the participant A as , the hash value set of the graph association relationship of the index primary key as , the hash value set of the index primary key of the participant B as , the hash value set of the graph association relationship of the index primary key as ;

[0106] (5.2) defining the index primary key set at risk of the participant A as , the index primary key set at risk of the participant B as , both of which are initially empty sets;

[0107] (5.3) the participant A takes out the marked point from and places it into until all points in are traversed;

[0108] (5.4) the participant A defines a temporary index primary key association relationship set , takes each time until the elements in remain unchanged and last for a cycle, and obtains , which is the index primary key set at risk of the participant A;

[0109] (5.5) the participant A and the participant B generate a temporary shared key based on the timestamp T and the identities and of the users of the two parties through key exchange;

[0110] (5.6) Party A randomly selects one , , the hash value set and of each element of the temporary shared key, form two sets, sent to party B; and ;

[0111] (5.7) Party B defines n temporary index primary key set and n temporary index primary key association set , after receiving the information, the shared key is decrypted, integrity and authenticity verification is passed, the real cascade information is obtained, and and are restored;

[0112] (5.8) Party B directly takes n hash value set and by redefining;

[0113] (5.9) from , take , until the element in starts to be constant and lasts for a cycle, and the new risk point set is obtained , all are cascaded in order to obtain , and returned to party A;

[0114] (5.10) party A according to t value, after receiving the message, take out the tth calculation , find out the corresponding , complete the complete hugging query activity, all the index primary keys involved in the hugging mark in , the irrelevant index primary key set is .

[0115] As a preferred embodiment of the present application, the step (5.9) takes , specifically comprising the following steps:

[0116] If and only one of them belongs to and the other does not belong to , the point not belonging to is added to the queue of , and the edge is deleted from , if the point belongs to , it is added to ;

[0117] If both a and b belong to , then directly delete this edge from ;

[0118] If neither a nor b belongs to , or other cases, then place this edge at the end of sequence.

[0119] The device for realizing cross-institution joint risk control based on secure multi-party graph calculation, wherein the device comprises:

[0120] a processor configured to execute computer executable instructions;

[0121] a memory storing one or more computer executable instructions, which, when executed by the processor, realize the steps of the method for realizing cross-institution joint risk control based on secure multi-party graph calculation.

[0122] The processor for realizing cross-institution joint risk control based on secure multi-party graph calculation, wherein the processor is configured to execute computer executable instructions, which, when executed by the processor, realize the steps of the method for realizing cross-institution joint risk control based on secure multi-party graph calculation.

[0123] The computer readable storage medium of the present application, which stores a computer program, wherein the computer program can be executed by a processor to realize the steps of the method for realizing cross-institution joint risk control based on secure multi-party graph calculation.

[0124] The application discloses a double-participant joint risk control method and system based on secure multi-party graph calculation, and the method comprises the following four solutions.

[0125] Solution 1: risk list dyeing query based on a graph under privacy protection: taking a risk list dyeing query based on a graph under privacy protection of a participant A and a participant B as an example, the participant A needs to query, by means of a graph structure of the participant B, information that a user with an ID of associates with a user with an ID of in a graph of the other party and the information is authorized to be viewed;

[0126] Solution 2: risk list dyeing query based on a graph under privacy protection: taking a risk list dyeing query based on a graph under privacy protection of a participant A and a participant B as an example, the participant A needs to query, by means of a graph structure of the participant B, information that a user with an ID of associates with a user with an ID of ;

[0127] Scheme 3: Non-privacy protection based on graph complete clustering query: Taking the complete clustering query of party A and party B based on the graph as an example, party A needs to complete the complete clustering labeling task with the help of the graph structure of party B;

[0128] Scheme 4: Privacy protection based on graph complete clustering query: Taking the complete clustering query of party A and party B under privacy protection as an example, the privacy protection mechanism is gradually added to the step description of scheme 2, and it is described in more detail.

[0129] Scheme 1, scheme 2 and scheme 4 are carried out under privacy protection, and scheme 3 is carried out under non-privacy protection, which is the basic scheme of scheme 4.

[0130] In schemes 1, 2 and 4, all messages transmitted to the other party by both parties need to be encrypted and signed based on identity by the party itself. Therefore, the schemes involved have anti-repudiation.

[0131] Schemes 1, 2 and 4 (scheme 3 as the precursor of scheme 4, without privacy protection), determine whether the external enemy can obtain the single graph node information or the association information between the graph nodes of the graph of the party in the initial state. The security of these information must meet: the privacy, confidentiality and integrity of the encrypted information, and the anti-forgery identity authentication of the two participants A and B.

[0132] Scheme 2 has higher privacy and business capability requirements than scheme 1, and scheme 4 has higher privacy and business capability requirements than scheme 2, that is, they belong to a progressive relationship, but schemes 1, 2 and 4 can be executed by 3. Only scheme 3 (i.e. step (5)) does not need to be under privacy protection, and the others need to be under privacy protection.

[0133] The execution conditions of the four schemes are as follows:

[0134] The privacy protection requirement of scheme 1 (i.e. step (2)) is to determine whether the external enemy can obtain the single graph node information or the association information between the graph nodes of the graph of the party in the initial state. The security of these information must meet: the privacy, confidentiality and integrity of the encrypted information, and the anti-forgery identity authentication of the two participants A and B. It is also necessary to ensure the structural security of the CKKS homomorphic encryption algorithm used to send information.

[0135] The privacy protection requirement of scheme 2 (i.e. step (3)) is that, in addition to the privacy protection requirement of scheme 2, it is also necessary to ensure that the temporary index primary key association relationship set also meets the privacy, confidentiality and integrity security required by the former.

[0136] Scheme 3 (i.e., step (5)) is a complete business scheme of the cluster mark query that must be performed when scheme 4 is performed (rather than a privacy protection scheme similar to schemes 1, 2, and 4).

[0137] The privacy protection requirement of scheme 4 (i.e., step (4)) is consistent with scheme 2.

[0138] In the specific embodiment of the present application, it is assumed that each piece of data of the participant A and the participant B needs to be signed and traceable, and the specific scheme is as follows:

[0139] Scheme 1, graph-based dyeing mark and query under privacy protection:

[0140] Taking the graph-based dyeing and query of the participant A and the participant B under privacy protection as an example, the participant A needs to query the relevant information of the user associated with the user in the participant by means of the graph structure of the participant B, and the intermediate process and identity authentication are irrelevant to the scheme, so this part is omitted, and a more detailed description is as follows:

[0141] Step S101: The participant B constructs a product polynomial of the user corresponding to the data set involved in , where s represents the number of possible associated users, and represents all the 3-degree associated member information:

[0142] .

[0143] After expansion, it is as follows:

[0144] .

[0145] Then, a polynomial as shown below is constructed:

[0146] .

[0147] The is subjected to Lagrange interpolation polynomial construction, and the specific coefficient is obtained, so that the specific is obtained.

[0148] Step S102: The participant A generates a public-private key pair and its corresponding encryption and decryption functions based on CKKS homomorphic encryption.

[0149] Step S103: For the to-be-searched keyword , the participant A sends the ciphertext to the participant B.

[0150] Step S104: After receiving the message, Party B calculates With , and sends it to Party A.

[0151] Step S105: After receiving the message, Party A decrypts With , if the former is 0, the latter is the associated index primary key that needs to be queried and , otherwise the content of the latter is meaningless.

[0152] wherein refers to the set of associated edges that Party B augments through the graph database in the cryptogram graph transmitted by Party A; x refers to the keyword that Party A needs to query; and m refers to the Lagrange interpolation function set by Party B, so that Party A can query the associated word m corresponding to the keyword x through the function of Party B. This means that Party A will issue multiple associated words m at a time, and only the tth associated word m is the valid query keyword , which can return the corresponding correct associated keyword , so as to ensure that other keywords x will not return the corresponding associated word m, and only will return the corresponding .

[0153] Scheme 2: Grouping query under privacy protection based on graph:

[0154] Taking the grouping query under privacy protection based on graph between Party A and Party B as an example, Party A needs to query whether the user with ID and the user with ID are associated, and a more detailed description is as follows:

[0155] Step S201: Party B establishes the set of associated edge pairs between any two nodes in . Party A publishes a set of public and private key pairs and . Step S202:

[0156] The number of elements in is denoted as , and Party B hashes and signs all the data in and constructs a product polynomial:

[0157] .

[0158] Step S203: Party A finds the index primary keys corresponding to the two users With At this time Then define a temporary edge pair as And first hash signature of the edge pair to get Through the public key of the party A encrypted into Send to party B.

[0159] Step S204: After receiving the message, party B directly substitutes Calculate

[0160] Get And send c back to party A.

[0161] Step S205: Party A decrypts c to get, if Then there is relevance, otherwise there is no relevance.

[0162] Wherein, Is the index primary key set of party B, Is the graph association relationship set of the index primary key of party B.

[0163] Simple correctness verification:

[0164] The main part to be verified is the correctness verification of the homomorphic encryption calculation completed by party B, and the derivation process is as follows:

[0165] ;

[0166] Scheme 3: complete huddling query based on graph under non-privacy protection:

[0167] Next, taking the complete huddling query based on graph of party A and party B as an example, party A needs to complete the complete huddling marking task by means of the graph structure of party B, and the more detailed description is as follows:

[0168] Step S301: According to the assumption, the index primary key set of party A is , the graph association relationship set of the index primary key is , the index primary key set of party B is , and the graph association relationship set of the index primary key is .

[0169] Step S302: Suppose that the index primary key set of party A at risk is , and the index primary key set of party B at risk is , which are initially empty sets.

[0170] Step S303: Party A first gets The point that the lieutenant general has already marked Take it out and put it in. In the middle, at this time there is until the entire process is completed. All points.

[0171] Step S304: Based on this, participant A first defines a temporary set of indexed primary key association relationships. Initial command Each time take If a and b belong to , Then add the other one as well. (Place) (the last of the traversal sequence), and remove this edge from Delete; if both a and b belong to Then directly remove this edge from Delete; if neither a nor b belongs to Then place the edge on The end of the sequence. This process repeats until... The elements in the loop remain unchanged from the beginning, and a loop is completed while maintaining this state. This operation (the second operation) can be considered the end of the traversal. At this point, the result is... This is the set of indexed primary keys for all participants A that are at risk.

[0172] Step S305: Participant A will and The message is sent to participant B. Upon receiving the message, participant B first defines a temporary set of indexed primary keys. Initially empty, and includes a temporary set of indexed primary key relationships. Initial command Redefining the take And its initial value is and And its initial value is .Pick If at this point, there exists a and b such that only one of them belongs to the same group. and another one that does not belong to Then it will not belong to Add point The queue (placed at the end), and its edge from Delete if the point belongs to Then it also needs to be added If both a and b belong to... Then directly remove this edge from Delete; if neither a nor b belongs to Or in other cases, place the edge... The end of the sequence. This process repeats until... The elements in the loop remain unchanged from the beginning, and a loop is completed while maintaining this state. This operation (the second operation) can be considered the termination of the traversal. A new set of risk points (the telephone set) is obtained. Ultimately, Return to participant A.

[0173] Step S306: Participant A calculates after receiving the message. This will complete the entire group query activity. The set of all indexed primary keys that have been marked as grouped is: The set of irrelevant index primary keys is .

[0174] The correctness of this solution can be easily verified using basic set operations.

[0175] Option 4: Privacy-protected graph-based fully collaborative query activity:

[0176] The following example illustrates a fully collaborative query activity between participant A and participant B under privacy protection. Privacy protection mechanisms are gradually incorporated into the steps of Scheme 2, and will be explained in more detail below:

[0177] Step S401: Based on step S301, add "According to the assumption, the hash value set of the primary key index of participant A is..." The set of hash values ​​for the graph associations of the indexed primary key is The set of hash values ​​for the indexed primary key of participant B is: The set of hash values ​​for the graph associations of the indexed primary key is ".

[0178] Step S402: Same as step S302.

[0179] Step S403: Same as step S303.

[0180] Step S404: Based on step S204, add "record" The set of hash values ​​is , The set of hash values ​​is ".

[0181] Step S405: Participant A and Participant B, based on timestamp T and the identities of both users... and A temporary shared key is generated through key exchange. This key cannot be reconstructed and can only be generated once within a certain period of time.

[0182] Step S406: Participant A randomly selects one , ,Will and Hash value set and XOR each element with a temporary shared key Two sets are formed and Send to participant B.

[0183] Step S407: Before receiving the information, participant B first defines n temporary sets of indexed primary keys. Initially empty, and contains n temporary sets of index primary key associations. Initial command The set of hash value pairs formed by the hash values ​​of the two endpoints. After receiving the information, participant B first decrypts it using the shared key to obtain... and Then, use integrity and authenticity verification tools. and The two parts were verified, and the actual cascading information was obtained after the verification was successful. and Thus recovering and .

[0184] Step S408: Participant B directly retrieves a set of n hash values ​​by redefining the hash value set. Their initial values ​​are respectively Similarly, obtain the hash value set. initial value ,in Therefore, only when Only when and .

[0185] Step S409: Next, from Begin, take If at this time the conditions are met and There is only one that belongs to and another one that does not belong to Then it will not belong to Add point The queue (placed at the end), and its edge from Delete if the point belongs to Then it also needs to be added If both a and b belong to , then directly delete this edge from ; if neither a nor b belongs to or other cases, then place this edge at the end of . Repeat this process until the elements in start to remain unchanged and complete a cycle (k times of operations) in this state, so that it can be considered that the traversal terminates, and a new risk-increasing point set (phone set) is obtained as . Finally, all are cascaded in order to obtain and returned to party A.

[0186] Step S410: Party A relies on the t value in hand to take out the tth calculation directly find the corresponding , which can complete this time the complete huddling query activity, and all index primary key sets involved in the huddling marked in are .

[0187] Simple correctness verification:

[0188] The main part of the scheme that needs to be verified for correctness is the ciphertext set operation module completed by party B, which needs to be consistent with the plaintext calculation. The derivation process is as follows:

[0189] When , there are and ;

[0190] According to the correctness of the steps of S205 and S308, it is easy to verify, and there is a certain synchronization in the operation of plaintext and ciphertext respectively. Therefore, it can be known that party B in the subsequent operation is exactly the same as the pure plaintext operation process, so that can be obtained smoothly, so that party A can directly restore when is obtained.

[0191] When , there are and ;

[0192] At this time, party B in the subsequent operation is exactly the same as the pure plaintext operation process, so that can be obtained smoothly, so that party A cannot directly restore when is obtained.​

[0193] Simple security analysis of scheme 1:

[0194] Here, since the identity authentication is not responsible for the part of the module, therefore here, only need to discuss whether the scheme has differential privacy:

[0195] Differential privacy: fully homomorphic encryption function To ensure that the participant B cannot use The data is encrypted and compared with To know whether there is participant A to query information and specific content ; In addition, participant A cannot use the received message to restore the coefficient of the Lagrange interpolation-based key information association function In participant B, participant A cannot know the information content irrelevant to the keyword of this query, so this scheme can realize differential privacy.

[0196] Technical effect of scheme 1:

[0197] The process participant A does not expose the node privacy and association privacy, participant B does not expose the information content irrelevant to the query keyword, and completes the query and marking operation under privacy protection.

[0198] Innovation of scheme 1:

[0199] This scheme first uses fully homomorphic encryption and Lagrange interpolation-based keyword anonymous query technology, so that the scheme can guarantee privacy security while not needing to transmit multiple polynomial power ciphertext data like traditional schemes, which can greatly save communication overhead.

[0200] Simple security analysis of scheme 2:

[0201] Since the anti-middleman attack relies on identity authentication, the anti-replay attack relies on time stamp, the message integrity and authenticity rely on the digital signature of each message with identity hash value, the forward and backward security relies on the temporary shared key with identity, the two-party secure calculation does not exist internal collusion attack and other security problems, the key structure security and differential privacy are bound. Therefore, here, the present application only needs to discuss whether the scheme has differential privacy:

[0202] Differential privacy: fully homomorphic encryption function To ensure that the participant B cannot use The data is encrypted and compared with To know whether there is participant A to query information and specific content Therefore, this scheme can realize differential privacy.

[0203] Technical effects of scheme 2:

[0204] Both the process participant A and the participant B do not expose the node privacy and the association privacy of the self party, and complete the huddling query operation under privacy protection.

[0205] Innovation of scheme 2:

[0206] The huddling query scheme based on the secure graph calculation is designed, the security of the systems of the query party and the queried party is ensured, the query party does not know the privacy information and the association graph of the queried party except the query content, and the queried party cannot know the specific data and the association graph of the query party, so that the whole graph joint query process is privacy and security, and the pain point that the traditional privacy query based on the graph exposes the association graph is remedied.

[0207] Technical effects of scheme 3:

[0208] In the process, the idea of greed and graph topology state convergence is used, the accurate group marking operation of the participant and the participant is completed, and the accurate complete huddling query activity under non-privacy protection is supported. However, in the process, although the participant A fully cooperates with the participant B to complete the query and marking operation, the node privacy and the association privacy of the self party are exposed, and the privacy protection property is not possessed. Therefore, on this basis, scheme 3 is designed.

[0209] Innovation of scheme 3:

[0210] Based on the idea of greed and graph topology state convergence, the accurate complete huddling query scheme under non-privacy protection is designed, the accurate group marking operation of the participant and the participant is completed, the accurate complete huddling query operation under non-privacy protection is supported, and the pain point that the traditional technology cannot accurately and completely huddle is remedied.

[0211] Simple security analysis of scheme 4:

[0212] Since the integrity and authenticity of the message depend on the digital signature of the identity hash value of each message, the two-party secure calculation does not have security problems such as internal collusion attack. Therefore, here, the application only needs to discuss whether the scheme has the properties of differential privacy, key structure and ciphertext indistinguishability, anti-middleman attack and replay attack, forward and backward security and the like:

[0213] Differential privacy: since the key is one-time and cannot be sent alone, an external enemy cannot perform a two-time differential privacy attack on it. And the internal is shared, and there is no internal privacy protection.

[0214] Key and ciphertext indistinguishability: is the unknown quantity of party B, which cannot be obtained by and acquiring value or recovering the required , that is, satisfying

[0215] ;

[0216] Therefore, the ciphertext has indistinguishability. The key itself is constructed by requiring the IDs of both parties and the current timestamp and a hash function, so that the synchronous construction of one-time pad can be completed together without key agreement, and the security of this part is attributed to the security of the key against man-in-the-middle attacks.

[0217] Anti-middleman attack and replay attack: in the system, the part mainly attacked by the middleman attack is the key part, and the middleman, even if he obtains the time T, does not know and , and it is difficult to be forged from the outside, so it can resist the middleman attack. Similarly, the two keys cannot be the same and have no any association and rule between them, so it can also resist the replay attack.

[0218] Forward and backward security: the enemy, even if he obtains the key of this round and the time T, does not know and , so he cannot obtain the old key and generate a new key , thereby ensuring the forward and backward security.

[0219] Technical effects of scheme 4:

[0220] The process realizes that the participants and the participants do not expose the node privacy and association privacy of each other, and cooperatively completes the query and marking operation while having the privacy protection property.

[0221] Innovation of scheme 4:

[0222] On the basis of scheme 3, the application designs a huddling query and complete huddling query privacy protection scheme based on secure graph computing, guarantees the security of the complete huddling query initiator and the business cooperation party system, lets the complete huddling query initiator not know the privacy information and association relationship graph of the cooperation party, and lets the cooperation party not know the specific data and association relationship graph of the initiator query, so that the whole complete huddling query process is private and secure, and the pain point that the traditional complete huddling query exposes the association relationship graph is remedied.

[0223] The flowchart of the complete huddling query part in the application is as follows Figure 7As shown, this optimized privacy protection workflow provides reliable response for complete huddling query, and realizes privacy security of both atlas correlation.

[0224] Step one: the participating unit pre-processes the other party's atlas data.

[0225] In this step, the complex network atlas of the participating unit is compressed and mapped into a plane atlas, and the atlas data is processed by the server for convergence, separating the risk existing subset and the risk-free subset.

[0226] Step two: the participating unit randomly selects a hash function signature and sends the atlas data, and establishes an atlas screening intersection with the participating unit.

[0227] In this step, after the participating unit sends the tthhash function signature data, the participating unit adopts each possible hash function to calculate the hash of the data node that may participate in "edge correlation" to obtain multiple hash atlases, and associates and screens the atlas nodes sent by the participating unit, adds the data existing in one end and not existing in the other end in "edge correlation" to the temporary risk queue, if this data is in the risk-free set sent by the participating unit, it is directly added to the risk set, and the data without association is put into the screening queue, so on and so forth, until a period of stable and unchanged cycle is performed, so as to indicate that the atlas converges to a stable state. At this time, the participating unit cannot distinguish which atlas is the atlas needed by the participating unit, and all the latest risk set atlases are packaged and prepared to be sent to the participating unit.

[0228] Step three: the participating unit selects the final atlas sent by the participating unit to obtain the required final data.

[0229] In this step, the participating unit directly selects the tthatlas sent, directly compares the risk atlas of oneself with the overall atlas, finds the ciphertext data of the new atlas obtained through this joint correlation activity from the database of oneself, and thus completes the complete huddling query activity.

[0230] The core technologies adopted by the four scenes of the scheme are as follows:

[0231] Scheme 1. In the privacy protection based on graph for dyeing label and query, the core technology we use is CKKS fully homomorphic encryption technology, Lagrange polynomial interpolation method based on anonymous keyword binding. Both of these two technologies are existing technologies, but we first found that Lagrange polynomial interpolation technology can be used in the encrypted domain graph black list dyeing query of two parties or even more parties. This can realize the content of the query to be safely hidden in the polynomial, and the query party does not know whether the queried party has information other than the authorized query and the specific content. Since the query operation does not cross the domain, the queried party also does not know the query keyword of the query party.

[0232] Scheme 2. In the privacy protection based on graph for huddling query, the core technology we use is PIR anonymous trace query technology based on passive transmission technology, Lagrange polynomial interpolation method based on anonymous keyword binding. Both of these two technologies are also existing technologies. Like scheme 1, through comparison with the achievements of predecessors, we are the first to excavate that these two technologies can be combined in the encrypted domain graph huddling query of two parties or even more parties. This can make the query party query the encrypted domain graph data of the queried party without obtaining the original domain graph data of the queried party, thereby ensuring the data security of the queried party. And the queried party also does not know the keyword information of the query party in the process of assisting the query.

[0233] Scheme 3. In the non-privacy protection based on graph for complete huddling query, the graph convergence solution involved is proposed by us by referring to the data structure "union set" thought involved in algorithm competition. Through the "temporary index primary key association set", the "quantity and element state" does not change after one traversal (entering the stable state), and it is considered that the complete huddling query process is completed, and the huddling of those who can huddle is completed.

[0234] In scheme 4, based on the graph under privacy protection, we increase the privacy protection technology on the basis of scheme 3. The core technology we use is the B-PSI technology for large-scale privacy intersection published in the CCS, one of the four top security conferences in the past two years, the temporary shared key based on identity and time, and the anonymous query technology PIR based on the passive transmission technology. These two technologies are also existing technologies. As in scheme 2, by comparing with the previous results, on the basis of the independently proposed scheme 3, we are the first to excavate that these three technologies can be combined in the encrypted sub-domain graph complete clustering query of two or even more parties. In this way, the querying party can query the encrypted sub-domain graph data of the queried party without obtaining the original sub-domain graph data of the queried party, so as to know which user ID is the other party and which is not. The B-PSI technology can also be used to shorten the time consumption of the privacy intersection. The queried party also cannot know the original sub-domain graph data of the querying party in the process of assisting the query, so as to know which user ID is the other party and which is not.

[0235] The specific implementation scheme of the embodiment can be referred to the related description in the above embodiment, which will not be repeated here.

[0236] It can be understood that the same or similar parts in the above embodiments can be mutually referred to, and the contents not described in detail in some embodiments can be referred to the same or similar contents in other embodiments.

[0237] It should be noted that in the description of the present application, the terms "first", "second", etc. are only for the purpose of description, and cannot be understood as indicating or implying relative importance. In addition, in the description of the present application, unless otherwise specified, the meaning of "a plurality of" is at least two.

[0238] Any process or method descriptions in flow charts or otherwise described herein represent embodiments which can be managed as one or more modules of executable code for implementation by various types of processing means. The scope of preferred embodiments of the present application encompasses other implementations wherein the order of execution could be changed, additional or fewer processes are performed, or where things other than being described are done.

[0239] It should be understood that aspects of the application can be implemented in hardware, software, firmware or a combination thereof. In the above embodiments, various steps or methods can be implemented in software or firmware that is stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, and in another embodiment, any of the following techniques can be used to implement the hardware: discrete logic circuitry having logic gates for implementing logic functions upon data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and / or other implementations known to those of skill in the art.

[0240] Those of skill in the art would understand that all or a portion of the exemplary methods described above can be carried out by a programmed computer or a specialized computer system. In an embodiment, the program is stored in a computer readable storage medium, which can be implemented in any of a variety of ways, including as a computer program product stored on a computer storage medium.

[0241] In addition, each functional unit in the various embodiments of the present application can be integrated in one processing module, or each unit can exist physically, or two or more units can be integrated in one module. The above-mentioned integrated module can be realized in the form of hardware or in the form of a software function module. If the integrated module is realized in the form of a software function module and sold or used as an independent product, it can also be stored in a computer readable storage medium.

[0242] The above-mentioned storage medium can be a read-only memory, a magnetic disk or an optical disk, etc.

[0243] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily mean the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0244] The method, device, processor and computer readable storage medium based on secure multi-party graph calculation for realizing cross-institutional joint risk control according to the application guarantee the privacy of users, the authenticity and integrity of messages and are suitable for a two-party graph spectrum joint keyword privacy query scheme; further realize that a participant completes self-party huddling query by using a participant graph spectrum; by using a greedy thought, accurate complete huddling query on a plaintext is found; based on the huddling query scheme on the plaintext, combined with PIR technology based on PIS, OT, hash function and other technologies, group marking under privacy protection is realized, so that finally, double-participant complete huddling query operation is realized under the premise of privacy protection. The scheme is analyzed from the aspects of correctness and security. In the above various privacy protection scheme processes, not only the security and privacy of each node of the graph spectrum data of the double participants can be protected, but also the privacy of the association relationship between them can be protected; on the basis of privacy protection, accurate complete huddling query calculation can also be realized.

[0245] In this specification, the application has been described with reference to particular embodiments thereof. However, it is apparent that various modifications and changes can be made without departing from the spirit and scope of the application. Therefore, the specification and drawings should be considered in an illustrative rather than a restrictive sense.

Claims

1. A method for implementing cross-institutional joint risk control based on secure multi-party graph computation, characterized in that, The method comprises the following steps: (1) judging whether to be under privacy protection, if yes and black list dyeing query is needed, then continue step (2); if yes and huddling marking query is needed, then continue step (3); if yes and complete huddling marking query is needed, then continue step (4); if no, then continue step (5); (2) Under privacy protection, coloring and querying are performed based on the graph. Participant A queries its own ID through the graph structure of Participant B. Users associated with the participating parties Complete the relevant information and end the process; (3) under privacy protection, group query is carried out based on the graph, the participant A queries the association between the user with ID and the user with ID through the graph structure of the participant B, and the step ends; (4) performing complete huddling query based on the graph under privacy protection, and continue step (5); (5) performing complete huddling query based on the graph under non-privacy protection, and participant A performs complete huddling marking task through the graph structure of participant B; The step (2) comprises the following steps: (2.1) Party B will the user involved in the above corresponding data set Constructing the product polynomial, the Lagrange interpolation polynomial is constructed for the data set, and the specific coefficient is obtained, wherein, The set of associated edges of the associated edge in the ciphertext graph transmitted by party A after being augmented by party B through the graph database; x refers to the keyword that party A needs to query, and m refers to the associated keyword m corresponding to the keyword x queried by party A through the Lagrange interpolation function of party B. (2.2) Party A generates a public-private key pair based on CKKS homomorphic encryption and its corresponding encryption and decryption functions and ; (2.3) Party A sends the ciphertext to Party B; (2.4) Party B computes with and sends to Party A; (2.5) After receiving the message, Party A decrypts with if is 0, then after decryption is the associated index primary key that needs to be queried wherein, refers to the tth associated word m in the plurality of associated words m, i.e. the effective query keyword corresponding correct associated word; The step (3) comprises the following steps: (3.1) Participant B establishes an association edge pair between any two nodes in establishes a set of index primary keys of participant B , participant A publishes a set of public and private key pairs and , wherein, denotes the index primary key set of participant B, denotes the graph association relationship set of the index primary key of participant B; (3.2) Party B hashes signs all data in and constructs a product polynomial; (3.3) define a temporary edge pair as and hash sign the edge pair to get encrypt by the public key of the own side to get and send to the participant B; (3.4) Party B substitutes into the constructed product polynomial and computes c is sent back to Party A; (3.5) Party A decrypts c, if is 0, then there is correlation, otherwise, there is no correlation; The step (4) comprises the following steps: (4.1) define the index primary key set of the participant A as , the atlas association relationship set of the index primary key as , the index primary key set of the participant B as , the atlas association relationship set of the index primary key as , define the hash value set of the index primary key of the participant A as , the hash value set of the atlas association relationship of the index primary key as , the hash value set of the index primary key of the participant B as , the hash value set of the atlas association relationship of the index primary key as ; (4.2) define the index primary key set at risk for party A as , and the index primary key set at risk for party B as , both initially empty sets; (4.3) Party A sends from the marked points to Party B, who takes them and puts them in until all points in have been traversed; (4.4) Party A defines a temporary index primary key association relationship set , each time taking , , until the elements in remain unchanged and last for a cycle, obtaining , that is, the index primary key set of all existing risks in Party A; (4.5) Party A and Party B generate a temporary shared key based on the timestamp T and the identities of the users of both parties and by means of a key exchange ; (4.6) Participant A randomly selects one , ,Will and Hash value set and Each element is XORed with a temporary shared key to form two sets, which are then sent to participant B. (4.7) The participant B defines n temporary index primary key sets and n temporary index primary key association sets After receiving the information, the participant B decrypts the information with the shared key, performs integrity and authenticity verification, and obtains the real cascade information after verification, and restores and ; (4.8) Party B directly takes n hash value sets by redefining and ; (4.9) From Begin, take , ,until The elements in the set remain unchanged for one cycle, resulting in a new set of risk points. to all By cascading them in order, we get... and return it to participant A; (4.10) Party A takes out the tth calculation according to the t value after receiving the message , finds the corresponding , completes the complete huddle query activity, All index primary key sets involved in the huddle marked in the , the irrelevant index primary key set is ; The step (5) comprises the following steps: (5.1) define the index primary key set of the participant A as , the atlas association relationship set of the index primary key as , the index primary key set of the participant B as , the atlas association relationship set of the index primary key as ; (5.2) define the index primary key set at risk for party A as , and the index primary key set at risk for party B as , both initially empty sets; (5.3) Party A takes the points that have been marked in all points in the set until all points in the set have been traversed.​​​ (5.4) Party A defines a temporary index primary key association relationship set , each time taking , , until the elements in remain unchanged and last for a cycle, obtaining , that is, the index primary key set of all risks in Party A; (5.5) Participant A will and The message is sent to participant B. After receiving the message, participant B defines a temporary set of indexed primary keys. and a temporary set of indexed primary key relationships. , definition take and ,Pick , ,until The elements in the set remain unchanged and continue for one cycle, resulting in a new set of risk points. ,Will Return to participant A; (5.6) Party A calculates upon receipt of the message 2. The method for realizing cross-institution joint risk control based on secure multi-party graph computation according to claim 1, characterized in that, In step (5.4), each time... Specifically, it includes the following steps: If a and b have exactly one in common then the other one is added to and its edge from is removed. If both a and b belong to , then directly delete this edge from ; If neither a nor b belongs to the set of edges that are placed at the end of the sequence.

3. The method for implementing cross-institution joint risk control based on secure multiparty graph computation according to claim 1, characterized in that, The step (5.5) takes each time , and specifically includes the following steps: If there exists only one of a and b that belongs to and another one that does not belong to Then it will not belong to Add point The queue, and its edge from Delete; if the point belongs to Then add it. middle; If both a and b belong to , then directly delete this edge from ; If neither a nor b belongs to or otherwise, the edge is placed at the end of the sequence.

4. The method for implementing cross-institution joint risk control based on secure multiparty graph computation according to claim 1, characterized in that, The step (4.9) of taking , specifically comprising the following steps: If satisfied and There is only one that belongs to and another one that does not belong to Then it will not belong to Add point The queue, and its edge from Delete if the point belongs to Then add it. middle; If both a and b belong to , then directly delete this edge from ; If neither a nor b belongs to or otherwise, the edge is placed at the end of the sequence.

5. An apparatus for implementing cross-institutional joint risk control based on secure multi-party graph computation, characterized in that, The device comprises: a processor configured to execute computer executable instructions; a memory storing one or more computer executable instructions, which, when executed by the processor, implement the steps of the method for realizing cross-institutional joint risk control based on secure multi-party graph computation according to any one of claims 1 to 4.

6. A processor for implementing cross-institutional joint risk control based on secure multi-party graph computation, characterized in that, The processor is configured to execute computer executable instructions, which, when executed by the processor, implement the steps of the method for realizing cross-institutional joint risk control based on secure multi-party graph computation according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, A computer program is stored thereon, which can be executed by the processor to implement the steps of the method for realizing cross-institutional joint risk control based on secure multi-party graph computation according to any one of claims 1 to 4.

Images