Privacy protection feature retrieval intersection method and device, electronic equipment and storage medium

Abstract

The invention provides a privacy protection feature retrieval intersection method and device, electronic equipment and a storage medium, relates to the technical field of information security, is applied to a receiver in a privacy feature protection retrieval PSI protocol, and comprises the following steps: obtaining a plurality of predicates in one-to-one correspondence with a plurality of dimensions, the method comprises the following steps: firstly, judging a feature threshold value in a predicate corresponding to any target dimension in a plurality of dimensions to carry out privacy protection retrieval to obtain a first retrieval result, and then obtaining a first polynomial coefficient according to a condition comparison operator in the predicate corresponding to the target dimension and a set first corresponding relationship between the condition comparison operator and the polynomial coefficient; according to the first polynomial coefficient and the first retrieval result, obtaining a condition retrieval result of the target feature vector on the target dimension, and obtaining a first security retrieval result based on a set AND logic connector and an OR logic connector; the problem that in the prior art, due to the fact that the predicate expression ability in a privacy protection feature retrieval PSI protocol is limited, the retrieval ability is limited is solved.

Description

Technical Field

[0001] This invention relates to the field of information security technology, and in particular to a method and apparatus for retrieving and finding intersections of privacy-preserving features, an electronic device, and a storage medium. Background Technology

[0002] In the Private Set Intersection (PSI) protocol, the communicating parties are referred to as the receiver and the sender. Each party holds a set of private data. The PSI protocol enables the receiver and sender to securely compute the intersection of their private data without revealing additional information. In some PSI scenarios, the receiver typically wants the private data held by the sender to satisfy predicates designed by the receiver (in PSI, a predicate is a set retrieval condition). That is, the receiver sets predicates based on the requirements of a data mining task, and the sender, based on the predicates set by the receiver, retrieves relevant data from its own data that matches the data mining task. Currently, the receiver may lack the relevant features needed to construct the predicates, thus requiring a sender with these features to execute the PSI protocol. If the receiver is willing to directly disclose the predicates to the sender, the current PSI protocol can handle the corresponding data mining task; that is, the sender filters out data that meets the conditions based on the predicates, and then the sender and receiver execute the PSI protocol to obtain the intersection. However, the receiver typically does not want the sender to know the content of the predicates. For example, in financial risk control companies, risk control strategies are their core assets. These strategies are used to generate predicates to target users, and naturally, the company doesn't want these predicates to be known by outsiders, especially competitors. To achieve this more accurate and privacy-preserving data mining method, a protocol called "Privacy-Preserving Feature Retrieval (PSI)" has been proposed. The PSI protocol allows the receiver to obtain the intersection of data that satisfies the predicates without disclosing the predicate content.

[0003] In practice, the predicate expression capabilities supported by the privacy-preserving feature retrieval PSI protocol are limited. For example, it can only support multiple predicates connected by logical connectors, but cannot support predicates connected by OR logical connectors, which limits retrieval capabilities. Summary of the Invention

[0004] This invention provides a privacy-preserving feature retrieval intersection method and apparatus, electronic device and storage medium, to solve the defect of limited retrieval capability caused by the limited predicate expression capability in the privacy-preserving feature retrieval PSI protocol in the prior art, and to achieve the purpose of enriching the predicate expression capability in the privacy-preserving feature retrieval PSI protocol to improve retrieval capability.

[0005] This invention provides a privacy-preserving feature retrieval intersection method, applied to the receiver in the privacy-preserving feature retrieval PSI protocol, comprising the following steps.

[0006] Obtain multiple predicates corresponding one-to-one with multiple dimensions; each predicate includes a feature threshold and a conditional comparison operator; based on the feature threshold in the predicate corresponding to any target dimension, perform privacy protection retrieval on the multiple feature vectors to be retrieved in the sender of the privacy feature protection retrieval PSI protocol, and obtain the first retrieval result of the feature of any target feature vector in the target dimension; based on the conditional comparison operator in the predicate corresponding to the target dimension, and the first correspondence between the set conditional comparison operator and the polynomial coefficients, obtain the first polynomial coefficients; substitute the first polynomial coefficients and the first retrieval result into the set retrieval quadratic polynomial to obtain the conditional retrieval result of the target feature vector in the target dimension, and obtain multiple conditional retrieval results corresponding one-to-one with the target feature vector in multiple dimensions; the conditional retrieval result includes: 0 or 1; When setting logical connectors for any adjacent predicates among multiple predicates, the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates are obtained according to the second correspondence between the set multiple logical connector types and multiple second polynomial coefficients; the multiple logical connector types include: AND logical connectors and OR logical connectors; a first secure retrieval result is obtained based on the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates and multiple conditional retrieval results; wherein, the receiver secretly shares the obtained multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates with the sender in the privacy feature protection retrieval PSI protocol, so as to instruct the sender to obtain a second secure retrieval result based on the multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates.

[0007] According to a privacy-preserving feature retrieval intersection method provided by the present invention, after obtaining multiple conditional retrieval results corresponding one-to-one with the target feature vector in multiple dimensions, the method further includes: when setting a quantity threshold for the target feature vector, summing the multiple conditional retrieval results corresponding one-to-one with the target feature vector in multiple dimensions to obtain a summation result; and when the summation result is greater than the quantity threshold, outputting the multiple conditional retrieval results as a third security retrieval result.

[0008] According to the privacy-preserving feature retrieval intersection method provided by the present invention, the second polynomial coefficients include the first coefficient of the first term and the second coefficient of the second term; based on the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates and multiple conditional retrieval results, a first secure retrieval result is obtained, including: constructing a secure polynomial corresponding to two target conditional retrieval results connected by the first logical connector among the multiple conditional retrieval results; the two target conditional retrieval results include: the first target conditional retrieval result and the second target conditional retrieval result; the first term in the secure polynomial is the sum of the first target conditional retrieval result and the second target conditional retrieval result, and the second term in the secure polynomial is the product of the target conditional retrieval result and the second target conditional retrieval result; determining the first coefficient as the coefficient of the first term in the secure polynomial, and determining the second coefficient as the coefficient of the second term in the secure polynomial, to obtain the retrieval result of the two target conditional retrieval results, and obtaining the final retrieval result of the two conditional retrieval results connected by each logical connector; and taking the final retrieval result of the two conditional retrieval results connected by all logical connectors as the first secure retrieval result.

[0009] According to the privacy-preserving feature retrieval intersection method provided by the present invention, when the logical connector type of the first logical connector is AND logical connector, the first coefficient is 0 and the second coefficient is 1; when the logical connector type of the first logical connector is OR logical connector, the first coefficient is -1 and the second coefficient is 1.

[0010] According to the privacy-preserving feature retrieval intersection method provided by the present invention, the condition comparison operators include a greater than operator, a greater than or equal to operator, a less than operator, a less than or equal to operator, an equal to operator, and an operator that does not perform feature retrieval; the first polynomial coefficients include quadratic coefficients, linear coefficients, and constant coefficients; when the condition operator is a greater than operator, the quadratic coefficient is -1, the linear coefficient is 2, and the constant coefficient is 0; when the condition operator is a less than or equal to operator, the quadratic coefficient is -0.5, the linear coefficient is 0.5, and the constant coefficient is 0. The coefficients of the quadratic term are 1; when the conditional operator is less than, the coefficient of the quadratic term is 0.5, the coefficient of the linear term is -0.5, and the coefficient of the constant term is 0; when the conditional operator is less than or equal to, the coefficient of the quadratic term is 1, the coefficient of the linear term is -2, and the coefficient of the constant term is 1; when the conditional operator is equal to, the coefficient of the quadratic term is 0.5, the coefficient of the linear term is -1.5, and the coefficient of the constant term is 1; when the conditional operator is not to perform feature retrieval, the coefficient of the quadratic term is 0, the coefficient of the linear term is 0, and the coefficient of the constant term is 1.

[0011] According to the privacy-preserving feature retrieval intersection method provided by the present invention, based on the feature threshold in the predicate corresponding to any target dimension among multiple dimensions, a privacy-preserving retrieval is performed on multiple feature vectors to be retrieved in the sender of the privacy feature retrieval PSI protocol to obtain a first retrieval result of the feature of any target feature vector in the target dimension. The method includes: calculating the feature of the target feature vector in the target dimension based on the greater than comparison operation mechanism in Secure Multi-Party Computation (MPC) according to the feature threshold in the predicate corresponding to any target dimension among multiple dimensions, to obtain a first comparison result; calculating the feature of the target feature vector in the target dimension based on the less than comparison operation mechanism in MPC according to the feature threshold in the predicate corresponding to any target dimension among multiple dimensions, to obtain a second comparison result; and obtaining the first retrieval result based on the first comparison result and the second comparison result.

[0012] The present invention also provides a privacy-preserving feature retrieval intersection device, applied to the receiver in privacy-preserving feature retrieval PSI, comprising the following modules: a first acquisition module, a retrieval module, a second acquisition module, a third acquisition module, a fourth acquisition module, and a processing module.

[0013] The first acquisition module is used to acquire multiple predicates that correspond one-to-one with multiple dimensions; each predicate includes a feature threshold and a condition comparison operator.

[0014] The retrieval module is used to perform privacy protection retrieval on multiple feature vectors to be retrieved in the sender of the privacy feature protection retrieval PSI protocol based on the feature threshold in the predicate corresponding to any target dimension among multiple dimensions, and to obtain the first retrieval result of the feature of any target feature vector in the target dimension.

[0015] The second acquisition module is used to acquire the first polynomial coefficients based on the conditional comparison operator in the predicate corresponding to the target dimension and the first correspondence between the conditional comparison operator and the polynomial coefficients.

[0016] The third acquisition module is used to substitute the coefficients of the first polynomial and the first search result into the set search quadratic polynomial to obtain the conditional search results of the target feature vector in the target dimension, and obtain multiple conditional search results that correspond one-to-one with the target feature vector in multiple dimensions; the conditional search results include: 0 or 1.

[0017] The fourth acquisition module is used to acquire the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates when setting logical connectors for any adjacent predicates among multiple predicates, based on the second correspondence between the set multiple logical connector types and the multiple second polynomial coefficients; the multiple logical connector types include: AND logical connector and OR logical connector.

[0018] The processing module is used to obtain a first secure search result based on the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates and multiple conditional search results; wherein, the receiver secretly shares the obtained multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates with the sender in the privacy feature protection retrieval PSI protocol, so as to instruct the sender to obtain a second secure search result based on the multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates.

[0019] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the privacy-preserving feature retrieval and intersection method as described above.

[0020] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the privacy-preserving feature retrieval and intersection method as described above.

[0021] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the privacy-preserving feature retrieval and intersection method as described above.

[0022] The privacy-preserving feature retrieval intersection method, apparatus, electronic device, and storage medium provided by this invention are applied to the receiver in the privacy-preserving feature retrieval (PSI) protocol. They involve obtaining multiple predicates corresponding to multiple dimensions; each predicate includes a feature threshold and a conditional comparison operator; performing privacy-preserving retrieval on multiple feature vectors to be retrieved in the sender of the PSI protocol based on the feature threshold in the predicate corresponding to any target dimension, obtaining a first retrieval result for the feature of any target feature vector in the target dimension; obtaining a first polynomial coefficient based on the conditional comparison operator in the predicate corresponding to the target dimension and a first correspondence between the conditional comparison operator and polynomial coefficients; and substituting the first polynomial coefficient and the first retrieval result into a set retrieval quadratic polynomial to obtain the conditional retrieval result of the target feature vector in the target dimension, thus obtaining the target feature vector in multiple dimensions. The above corresponds to multiple conditional search results; the conditional search results include: 0 or 1; when setting logical connectors for any adjacent predicates among multiple predicates, according to the second correspondence between the set multiple logical connector types and multiple second polynomial coefficients, the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates are obtained; the multiple logical connector types include: AND logical connector and OR logical connector; according to the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates and the multiple conditional search results, the first secure search result is obtained; wherein, the receiver secretly shares the obtained multiple predicates, the first polynomial coefficients, and the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates with the sender in the privacy feature protection search PSI protocol, so as to instruct the sender to obtain the second secure search result according to the multiple predicates, the first polynomial coefficients, and the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates. The logical connector types in this invention include not only AND logical connectors but also OR logical connectors, thereby solving the defect that the limited predicate expression capability in the privacy-preserving feature retrieval PSI protocol in the prior art leads to limited retrieval capability, and achieving the goal of enriching the predicate expression capability in the privacy-preserving feature retrieval PSI protocol to improve retrieval capability. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0024] Figure 1 This is one of the flowcharts illustrating the privacy-preserving feature retrieval and intersection method provided by the present invention.

[0025] Figure 2 This is the second flowchart of the privacy-preserving feature retrieval and intersection method provided by the present invention.

[0026] Figure 3 This is a schematic diagram of the privacy-preserving feature retrieval and intersection device provided by the present invention.

[0027] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0029] The following is combined with Figures 1-2 This invention describes the privacy-preserving feature retrieval and intersection method.

[0030] Figure 1 This is one of the flowcharts illustrating the privacy-preserving feature retrieval intersection method provided by this invention. This method is applied to the receiver in the privacy-preserving feature retrieval PSI protocol. For example... Figure 1 As shown, the method includes the following steps S110~S160.

[0031] S110: Obtain multiple predicates that correspond one-to-one with multiple dimensions; where each predicate includes: feature threshold and condition comparison operator.

[0032] In this embodiment of the invention, the receiver understands the feature information held by the sender, including the dimensions, order, and type of the features (because the receiver should know what services the sender can provide). Based on the feature information held by the sender that the receiver understands, the receiver sets multiple predicates corresponding to multiple dimensions. Each predicate may include: feature thresholds and conditional comparison operators.

[0033] For example, the receiver can propose a set of predicates. , . It can include ~ , d represents the total number of dimensions across multiple dimensions, where d is a positive integer. It can include ~ , Let be the feature threshold for the j-th dimension (j is a positive integer, 0 ≤ j ≤ d). For the conditional comparison operator of the j-th dimension, and Together, they determine the predicate for the j-th dimension. The predicate for the j-th dimension is the predicate used to retrieve the sender's j-th dimension feature. Let be the feature threshold in the predicate of the j-th dimension. For the conditional comparison operator in the predicate of the j-th dimension.

[0034] For example, if the j-th dimension feature of the sender to be retrieved is personal income, and the retrieval condition is that personal income must be greater than or equal to 20,000, then the following settings can be configured: =20000, =≥.

[0035] In practice , , For an integer ring that is a power of 2, d is the set of feature values ​​corresponding to the ID set of sender Y. The feature dimensions. It refers to a set consisting of d subsets that correspond one-to-one with d feature dimensions, where each subset contains elements of the same type. One of the elements. It refers to a set consisting of d subsets that correspond one-to-one with d feature dimensions, where each subset contains elements of the same type. One of the elements.

[0036] It can be a logical connector that connects two predicates (see the corresponding introduction to logical connectors in S150~S160, which will not be explained in detail here); or, The number threshold can be set for multiple predicates (see the corresponding introduction to the number threshold for multiple predicates in the later section, which will not be explained in detail here).

[0037] In practice, if the receiver does not restrict the sender's j-th dimension feature, that is, the sender's j-th dimension feature is not included in the retrieval, it can be set as follows: It is a random value, and is set = .

[0038] To obtain multiple predicates that correspond one-to-one with multiple dimensions, one can directly receive the predicates corresponding to each of the multiple dimensions set by the receiver.

[0039] After obtaining multiple predicates that correspond one-to-one with multiple dimensions, the receiver secretly shares these predicates with the sender.

[0040] S120: Based on the feature threshold in the predicate corresponding to any target dimension among multiple dimensions, perform privacy protection retrieval on multiple feature vectors to be retrieved in the sender of the privacy feature protection retrieval PSI protocol, and obtain the first retrieval result of any target feature vector in the target dimension.

[0041] The basic functional mechanisms in secure multi-party computation (MPC) include: addition mechanism. Multiplication mechanism Greater than comparison operation mechanism Less than comparison operation mechanism And the greater than or equal to comparison mechanism .

[0042] In some embodiments, in each of the receiver and the sender, the features of the target feature vector in the target dimension can be calculated based on the feature threshold in the predicate corresponding to any target dimension among multiple dimensions, and the greater than comparison operation mechanism in Secure Multi-Party Computation (MPC) can be used to obtain a first comparison result. Then, based on the feature threshold in the predicate corresponding to any target dimension among multiple dimensions, the features of the target feature vector in the target dimension can be calculated based on the less than comparison operation mechanism in MPC to obtain a second comparison result.

[0043] For example, the target dimension is the j-th dimension, and the sender and receiver use a privacy-preserving retrieval PSI protocol to retrieve any one of multiple target feature vectors from the sender's multiple feature vectors. Input and the feature threshold in the predicate corresponding to the target dimension. , call and The following results (1) and (2) can be obtained.

[0044] Result (1): ,express ;in, Indicates secret sharing, Indicates the size of the sender's input. Represents the target feature vector Features in the j-th dimension.

[0045] Result (2): ,express .

[0046] The above result (1) is the first comparison result, and the above result (2) is the second comparison result.

[0047] Then, the receiver and the sender each obtain the first search result based on the first comparison result and the second comparison result.

[0048] The receiver and sender perform the following calculations: = +2 ; .

[0049] for The j-th dimension of the i-th feature vector is processed as above, and the following three results can be obtained as shown in formula (1).

[0050] Formula (1): .

[0051] for The feature of the i-th eigenvector in the j-th dimension First search results ,like , then it means That is, the feature of the i-th feature vector in the j-th dimension is equal to the feature threshold of the i-th feature vector in the j-th dimension; if , then it means That is, the feature of the i-th feature vector in the j-th dimension is greater than the feature threshold of the i-th feature vector in the j-th dimension; if , then it means That is, the feature of the i-th feature vector in the j-th dimension is less than the feature threshold of the i-th feature vector in the j-th dimension.

[0052] The above process was completed With all A security comparison between them. Next, based on... To evaluate the retrieval quadratic polynomial, the result of the evaluation of the retrieval quadratic polynomial can be used to determine... Does the feature of the i-th eigenvector in the j-th dimension satisfy...? and For the evaluation process of the retrieved quadratic polynomial, see S130~S140.

[0053] S130: Obtain the first polynomial coefficients based on the conditional comparison operator in the predicate corresponding to the target dimension and the first correspondence between the conditional comparison operator and the polynomial coefficients.

[0054] In some embodiments, conditional comparison operators may include: greater than operator >, greater than or equal to operator ≥, less than operator <, less than or equal to operator ≤, equal to operator =, and an operator that does not perform feature retrieval. The coefficients of the first polynomial include the coefficients of the quadratic term, the coefficients of the linear term, and the coefficient of the constant term.

[0055] The first correspondence can be seen in Table 1 below.

[0056] Table 1

[0057] As shown in Table 1, when the conditional operator is greater than the operator, the coefficient of the quadratic term is -1, the coefficient of the linear term is 2, and the coefficient of the constant term is 0.

[0058] As shown in Table 1, when the conditional operator is less than or equal to the operator, the coefficient of the quadratic term is -0.5, the coefficient of the linear term is 0.5, and the coefficient of the constant term is 1.

[0059] As shown in Table 1, when the conditional operator is less than, the coefficient of the quadratic term is 0.5, the coefficient of the linear term is -0.5, and the coefficient of the constant term is 0.

[0060] As shown in Table 1, when the conditional operator is less than or equal to the operator, the coefficient of the quadratic term is 1, the coefficient of the linear term is -2, and the coefficient of the constant term is 1.

[0061] As shown in Table 1, when the conditional operator is the equality operator, the coefficient of the quadratic term is 0.5, the coefficient of the linear term is -1.5, and the coefficient of the constant term is 1.

[0062] As shown in Table 1, when the conditional operator is the one that does not perform feature retrieval, the coefficient of the quadratic term is 0, the coefficient of the linear term is 0, and the coefficient of the constant term is 1.

[0063] S140: Substitute the coefficients of the first polynomial and the first search result into the set search quadratic polynomial to obtain the conditional search results of the target feature vector in the target dimension, and obtain multiple conditional search results that correspond one-to-one with the target feature vector in multiple dimensions; the conditional search results include: 0 or 1.

[0064] The retrieval quadratic polynomial can be set as follows: .

[0065] Conditional retrieval results based on the features of the i-th feature vector in the j-th dimension obtained from S120 and S130 based on The obtained a, b, and c will be Substitute a, b, and c into In this process, we can obtain the conditional retrieval result of the i-th feature vector in the j-th dimension.

[0066] A conditional search result of 0 indicates that the i-th feature vector does not satisfy the condition in the j-th dimension. and A conditional retrieval result of 1 indicates that the i-th feature vector satisfies the condition in the j-th dimension. and .

[0067] S150: When setting logical connectors for any adjacent predicates among multiple predicates, obtain the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates according to the set multiple logical connector types and the second correspondence of multiple second polynomial coefficients; the multiple logical connector types include: AND logical connector and OR logical connector.

[0068] The following describes the setting of the logical connector for connecting two predicates in the embodiments of the present invention.

[0069] Specifically, This is the logical connector that joins two predicates. In this case, AND represents the AND logical connector, and OR represents the OR logical connector. express: It can be ( There are d logical connectors used to connect features of d dimensions in a feature vector.

[0070] The process of obtaining the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates may include: obtaining the logical connector type of any two adjacent predicates; obtaining the second polynomial coefficients corresponding to the logical connector types of the two adjacent predicates according to the second correspondence between multiple logical connector types and multiple second polynomial coefficients; and determining the second polynomial coefficients as the second polynomial coefficients corresponding to the logical connectors of the two adjacent predicates.

[0071] The second correspondence between logical connector types and multiple second polynomial coefficients .

[0072] For example, if the logical connector type of two adjacent predicates is AND (i.e., the above AND), according to the second correspondence mentioned above, the coefficient of the second polynomial corresponding to the AND logical connector is... The coefficients of the second polynomial corresponding to the logical connectors of the two adjacent predicates are If the logical connectors of two adjacent predicates are of type OR (i.e., the OR above), then according to the second correspondence above, the coefficient of the second polynomial corresponding to the OR logical connector is... The coefficients of the second polynomial corresponding to the logical connectors of the two adjacent predicates are .

[0073] As in the example above, the coefficients of the second polynomial include the first coefficient of the first term and the second coefficient of the second term. For example, the coefficients of the second polynomial are... The first coefficient of the first term is 0, and the second coefficient of the second term is 1; for example, the coefficient of the second polynomial is... The first coefficient of the first term is -1, and the second coefficient of the second term is 1.

[0074] After obtaining the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates, the receiver secretly shares these coefficients with the sender. This prevents the sender from knowing the specific type of the logical connector, but allows the sender to perform logical connector operations based on the second polynomial coefficients.

[0075] S160: Obtain a first secure search result based on the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates and multiple conditional search results; wherein, the receiver secretly shares the obtained multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates with the sender in the PSI, so as to instruct the sender to obtain a second secure search result based on the multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates.

[0076] A secure polynomial can be constructed based on two target condition search results connected by a first logical connector among multiple condition search results. The two target condition search results include: a first target condition search result and a second target condition search result. The first term in the secure polynomial is the sum of the first target condition search result and the second target condition search result, and the second term in the secure polynomial is the product of the target condition search result and the second target condition search result.

[0077] For example, the search results for the first target condition are: , The second target condition search result is , .

[0078] The initial safety polynomial of the two target condition search results is .

[0079] When the first logical connector is of type AND, the first coefficient is 0 and the second coefficient is 1; when the first logical connector is of type OR, the first coefficient is -1 and the second coefficient is 1.

[0080] The first coefficient is determined as the coefficient of the first term in the safety polynomial, and the second coefficient is determined as the coefficient of the second term in the safety polynomial. The retrieval results of the two target conditions are obtained, and the retrieval results of the two conditions connected by each logical connector are obtained.

[0081] When the first logical connector is an AND logical connector, the coefficients of the second polynomial corresponding to the AND logical connector (i.e.: Substituting the initial safety polynomial above, we can obtain the safety polynomials for the two target condition retrieval results as follows: .

[0082] When the first logical connector is an OR logical connector, the coefficients of the second polynomial corresponding to the OR logical connector (i.e.: Substituting the initial safety polynomial above, we can obtain the safety polynomials for the two target condition retrieval results as follows: .

[0083] like =0, =0, the first logical connector is the AND logical connector, which means that the search result of the two target conditions is 0. Therefore, the final search result of the two conditions connected by the first logical connector is 0.

[0084] like =0, =1, the first logical connector is an AND logical connector, which means that the search result for two target conditions is 0. Therefore, the final search result for the two conditions connected by the first logical connector is 0.

[0085] like =1, =1, the first logical connector is an AND logical connector, which means that the search result for two target conditions is 1. Therefore, the final search result for the two conditions connected by the first logical connector is 1.

[0086] like =0, =0, the first logical connector is an OR logical connector, which means that the search result for two target conditions is 0. Therefore, the final search result for the two conditions connected by the first logical connector is 0.

[0087] like =0, =1, the first logical connector is an OR logical connector, which means that the search result for two target conditions is 1. Therefore, the final search result for the two conditions connected by the first logical connector is 1.

[0088] like =1, =1, the first logical connector is an OR logical connector, which means that the search result for two target conditions is 1. Therefore, the final search result for the two conditions connected by the first logical connector is 1.

[0089] The final search result of two conditions connected by all logical connectors is taken as the first safe search result.

[0090] It should be noted that the receiver secretly shares the acquired predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates with the sender in the Privacy Feature Protection Retrieval (PSI) protocol. This instructs the sender to obtain a second secure retrieval result based on the predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates. The process by which the sender obtains the second secure retrieval result based on the predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates can be referred to the process by which the receiver obtains the first secure retrieval result, as described above. This embodiment of the invention will not elaborate further on this process.

[0091] In practice, the receiver and sender can obtain , indicating each ∈ Does it satisfy the predicate set? .in, It is the output result of the sender during the retrieval process. It is the output result of the receiver during the retrieval process.

[0092] In some embodiments, the present invention can also perform threshold-based security retrieval (or threshold-based security retrieval).

[0093] After executing S140, as Figure 2 As shown, S210~S220 can also be executed.

[0094] S210: When a quantity threshold is set for the target feature vector, sum the results of multiple conditional searches that correspond one-to-one with the target feature vector in multiple dimensions to obtain the summation result.

[0095] S220: If the summation result is greater than the quantity threshold, output multiple conditional search results as the third secure search result.

[0096] The quantity threshold refers to the condition that if the number of features in the target feature vector that satisfy a certain predicate is greater than or equal to the quantity threshold, then the target feature vector is considered to satisfy that predicate.

[0097] In the The feature of the i-th eigenvector in the j-th dimension The conditional test result is ,like Then it means that the i-th feature vector satisfies the following in the j-th dimension. and ,like This indicates that the i-th eigenvector does not satisfy the condition in the j-th dimension. and .

[0098] In this case, the addition mechanism in MPC can be used directly. right Summing each row, that is, summing each row... The conditional retrieval results of features across all dimensions of the i-th feature vector are summed, and then a comparison mechanism greater than or equal to is used. Determine if the summation result is greater than or equal to a quantity threshold. If the result is yes, meaning the summation result is greater than or equal to the quantity threshold, then the comparison mechanism is activated. The output result is the third-party security search result.

[0099] When a quantity threshold is set for the target feature vector, the above example illustrates the process by which the receiver sums the multiple conditional search results corresponding one-to-one with the target feature vector across multiple dimensions, obtains the summation result, and outputs the multiple conditional search results as the third secure search result when the summation result exceeds the quantity threshold. The process by which the sender sums the multiple conditional search results corresponding one-to-one with the target feature vector across multiple dimensions, obtains the summation result, and outputs the multiple conditional search results as the fourth secure search result when the summation result exceeds the quantity threshold, can be referred to the above-described process for obtaining the third secure search result; this embodiment of the invention will not elaborate further on this.

[0100] In practice, the feature set input by the sender S Receiver R input At the end of the method in this embodiment of the invention, S and R respectively obtain outputs. and , satisfying for any ,like Then there is + =1, where, ( , ) ( , ). + =0 indicates that the predicate is not satisfied. refers to samples Does the predicate satisfy? .

[0101] The method in this embodiment of the invention can also be applied to the PSI protocol based on DH (Diffie-Hellman) key negotiation.

[0102] Privacy-preserving feature retrieval PSI protocols based on DH key negotiation require combining standard DH key negotiation-based PSI protocols with secure retrieval protocols. The simplest approach is to use additive homomorphic encryption to bind the output of the secure retrieval to a mask. However, this naive approach incurs the intersection size |I| and the original intersection size. This information is leaked to the sender. X is the set of IDs for the receivers, and Y is the set of IDs for the sender.

[0103] In this embodiment of the invention, Cuckoo hashing (a highly efficient hash table structure that uses multiple hash functions to put elements into a fixed-size hash bucket, and can "kick out" existing elements and relocate them when there is a collision) can be used as a bridge to seamlessly integrate the standard DH-based PSI protocol with the secure retrieval protocol of this embodiment of the invention, thereby constructing a privacy-preserving feature retrieval PSI protocol that does not disclose any cardinality information.

[0104] Specifically, the objective of this embodiment of the invention is to construct the following mask for the receiver and the sender: for The mask is ; It is an elliptic curve group. To map the input to the group hash function on, For the recipient's secret, It is the sender's secret.

[0105] for The mask is .

[0106] when At that time, it will necessarily satisfy ,in For the recipient's secret, It is the sender's secret. However, the receiver cannot keep it. Corresponding to the recipient In connection, the recipient cannot determine the specific details for each individual. Should be used ,because index and Alignment, not with Alignment.

[0107] This invention utilizes Cuckoo hashing to align data between two parties, as proposed in this embodiment. Unlike related technologies where the receiver constructs the Cuckoo hashing, this invention involves the sender constructing the Cuckoo hash table. Specifically, the sender uses... Building the Cuckoo Hash Table and the corresponding feature values Append to according to hash index Then, the sender and receiver execute a secure retrieval protocol based on these indexes. In this way, the receiver can retrieve data for each index. When constructing the mask, it is only necessary to select from k candidate masks. In this embodiment, homomorphic encryption is no longer relied upon, thereby significantly reducing computational overhead.

[0108] In some embodiments of the present invention, the method can also be applied to the Privacy Equality Test (PEQT).

[0109] PEQT is a secure protocol that allows two parties to determine if their input values ​​are equal without revealing the input itself. Specifically, if they are equal, both parties receive equal random values; if they are not equal, both parties receive unequal random values.

[0110] This invention demonstrates a general method for constructing a privacy-preserving feature retrieval (PSI) protocol based on OPRF.

[0111] After executing the secure retrieval protocol (the method described above in this embodiment of the invention), the sender and receiver respectively obtain... and , and Satisfy: For any ,like Then there must be =1; otherwise + =0.

[0112] After this, the sender and receiver will each input... - and Send to PEQT function module ,get and .like and If they are equal, then there is a corresponding and They are equal, otherwise and For random values ​​(i.e.) and (For different random results). For a large finite field, For length is The unit vector.

[0113] Finally, the receiver and sender execute the OPRF protocol, with the receiver using... Get F for input X The sender calculates F using Y as input. Y , will F Y With the previously obtained The sum is then sent to the recipient. The recipient then subtracts the corresponding amount. Get F Y ', will F Y 'With F X Perform a comparison and determine the comparison result as the output result.

[0114] The privacy-preserving feature retrieval intersection method provided by this invention is applied to the receiver in the Privacy-Preserving Feature Retrieval (PSI) protocol. It obtains multiple predicates corresponding to multiple dimensions. Each predicate includes a feature threshold and a conditional comparison operator. Based on the feature threshold in the predicate corresponding to any target dimension, a privacy-preserving retrieval is performed on the multiple feature vectors to be retrieved in the sender of the PSI protocol, yielding a first retrieval result for the feature of any target feature vector in the target dimension. Based on the conditional comparison operator in the predicate corresponding to the target dimension and a predefined first correspondence between the conditional comparison operator and polynomial coefficients, a first polynomial coefficient is obtained. Substituting the first polynomial coefficient and the first retrieval result into a predefined retrieval quadratic polynomial, a conditional retrieval result for the target feature vector in the target dimension is obtained, resulting in a one-to-one correspondence between the target feature vector and the polynomial coefficients in multiple dimensions. Multiple conditional search results; the conditional search results include: 0 or 1; when setting logical connectors for any adjacent predicates among multiple predicates, the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates are obtained according to the second correspondence between the set multiple logical connector types and multiple second polynomial coefficients; the multiple logical connector types include: AND logical connector and OR logical connector; based on the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates and the multiple conditional search results, a first secure search result is obtained; wherein, the receiver secretly shares the obtained multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates with the sender in the privacy feature protection search PSI protocol, so as to instruct the sender to obtain a second secure search result based on the multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates. The logical connector types in this invention include not only AND logical connectors but also OR logical connectors. Furthermore, this invention can also achieve threshold-based secure retrieval, thereby solving the defect in the prior art where the predicate expression capability in the privacy-preserving feature retrieval PSI protocol is limited, resulting in restricted retrieval capabilities. This invention aims to enrich the predicate expression capability in the privacy-preserving feature retrieval PSI protocol to improve retrieval capabilities.

[0115] The privacy-preserving feature retrieval and intersection device provided by the present invention is described below. The privacy-preserving feature retrieval and intersection device described below can be referred to in correspondence with the privacy-preserving feature retrieval and intersection method described above.

[0116] Figure 3 This is a schematic diagram of the privacy-preserving feature retrieval and intersection device provided by the present invention. Figure 3As shown, the privacy-preserving feature retrieval and intersection device 300 includes the following modules: a first acquisition module 301, a retrieval module 302, a second acquisition module 303, a third acquisition module 304, a fourth acquisition module 305, and a processing module 306.

[0117] The first acquisition module 301 is used to acquire multiple predicates that correspond one-to-one with multiple dimensions; wherein each predicate includes a feature threshold and a condition comparison operator.

[0118] The retrieval module 302 is used to perform privacy protection retrieval on multiple feature vectors to be retrieved in the sender of the privacy feature protection retrieval PSI protocol based on the feature threshold in the predicate corresponding to any target dimension among multiple dimensions, and to obtain the first retrieval result of the feature of any target feature vector in the target dimension.

[0119] The second acquisition module 303 is used to acquire the first polynomial coefficients based on the conditional comparison operator in the predicate corresponding to the target dimension and the first correspondence between the conditional comparison operator and the polynomial coefficients.

[0120] The third acquisition module 304 is used to substitute the coefficients of the first polynomial and the first search result into the set search quadratic polynomial to obtain the conditional search results of the target feature vector in the target dimension, and obtain multiple conditional search results that correspond one-to-one with the target feature vector in multiple dimensions; the conditional search results include: 0 or 1.

[0121] The fourth acquisition module 305 is used to acquire the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates when setting logical connectors for any adjacent predicates among multiple predicates, based on the second correspondence between the set multiple logical connector types and the multiple second polynomial coefficients; the multiple logical connector types include: AND logical connector and OR logical connector.

[0122] Processing module 306 is used to obtain a first secure search result based on the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates and multiple conditional search results; wherein, the receiver secretly shares the obtained multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates with the sender in the privacy feature protection retrieval PSI protocol, so as to instruct the sender to obtain a second secure search result based on the multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates.

[0123] The privacy-preserving feature retrieval intersection device provided by this invention is applied to the receiver in the privacy-preserving feature retrieval (PSI) protocol. It obtains multiple predicates corresponding to multiple dimensions. Each predicate includes a feature threshold and a conditional comparison operator. Based on the feature threshold in the predicate corresponding to any target dimension, a privacy-preserving retrieval is performed on multiple feature vectors to be retrieved in the sender of the PSI protocol, yielding a first retrieval result for the feature of any target feature vector in the target dimension. Based on the conditional comparison operator in the predicate corresponding to the target dimension and a first correspondence between the conditional comparison operator and polynomial coefficients, a first polynomial coefficient is obtained. Substituting the first polynomial coefficient and the first retrieval result into a set retrieval quadratic polynomial, a conditional retrieval result for the target feature vector in the target dimension is obtained, resulting in a one-to-one correspondence between the target feature vector and the polynomial coefficients in multiple dimensions. Multiple conditional search results; the conditional search results include: 0 or 1; when setting logical connectors for any adjacent predicates among multiple predicates, the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates are obtained according to the second correspondence between the set multiple logical connector types and multiple second polynomial coefficients; the multiple logical connector types include: AND logical connector and OR logical connector; based on the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates and the multiple conditional search results, a first secure search result is obtained; wherein, the receiver secretly shares the obtained multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates with the sender in the privacy feature protection search PSI protocol, so as to instruct the sender to obtain a second secure search result based on the multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates. The logical connector types in this invention include not only AND logical connectors but also OR logical connectors, thereby solving the defect that the limited predicate expression capability in the privacy-preserving feature retrieval PSI protocol in the prior art leads to limited retrieval capability, and achieving the goal of enriching the predicate expression capability in the privacy-preserving feature retrieval PSI protocol to improve retrieval capability.

[0124] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communications bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other through the communications bus 440. The processor 410 can call logical instructions in the memory 430 to execute a privacy-preserving feature retrieval intersection method. This method includes: obtaining multiple predicates corresponding one-to-one with multiple dimensions; wherein each predicate includes a feature threshold and a conditional comparison operator; performing privacy-preserving retrieval on multiple feature vectors to be retrieved in the sender of the privacy-preserving feature retrieval PSI protocol based on the feature threshold in the predicate corresponding to any target dimension, obtaining a first retrieval result for the feature of any target feature vector in the target dimension; obtaining a first polynomial coefficient based on the conditional comparison operator in the predicate corresponding to the target dimension and a first correspondence between the conditional comparison operator and the polynomial coefficients; substituting the first polynomial coefficients and the first retrieval result into a set retrieval quadratic polynomial to obtain the conditional retrieval result of the target feature vector in the target dimension, thus obtaining a one-to-one correspondence between the target feature vector and the polynomial coefficients in multiple dimensions. The search results are obtained based on multiple conditions. The conditions include 0 or 1. When logical connectors are set for any adjacent predicates among multiple predicates, the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates are obtained according to the second correspondence between the set logical connector types and the multiple second polynomial coefficients. The multiple logical connector types include AND and OR logical connectors. A first secure search result is obtained based on the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates and the multiple conditions. The receiver secretly shares the obtained multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates with the sender in the Privacy Feature Protection Search (PSI) protocol, instructing the sender to obtain a second secure search result based on the multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates.

[0125] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0126] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the privacy-preserving feature retrieval intersection method provided by the above methods. The method includes: obtaining multiple predicates that correspond one-to-one with multiple dimensions; wherein each predicate includes: a feature threshold and a conditional comparison operator; performing privacy-preserving retrieval on multiple feature vectors to be retrieved in the sender of the privacy-preserving feature retrieval PSI protocol according to the feature threshold in the predicate corresponding to any target dimension among the multiple dimensions, and obtaining a first retrieval result of the feature of any target feature vector in the target dimension; obtaining a first polynomial coefficient according to the conditional comparison operator in the predicate corresponding to the target dimension and a first correspondence between the set conditional comparison operator and the polynomial coefficient; and substituting the first polynomial coefficient and the first retrieval result into a set retrieval quadratic polynomial to obtain the target feature vector. The retrieval results for the target feature vector in multiple dimensions are obtained by using conditional retrieval results in the target dimension. The conditional retrieval results include 0 or 1. When setting logical connectors for any adjacent predicates among multiple predicates, the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates are obtained according to the second correspondence between the set logical connector types and the second polynomial coefficients of multiple second polynomials. The multiple logical connector types include AND logical connectors and OR logical connectors. The first secure retrieval result is obtained based on the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates and the multiple conditional retrieval results. The receiver secretly shares the obtained multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates with the sender in the Privacy Feature Protection Retrieval (PSI) protocol, so as to instruct the sender to obtain the second secure retrieval result based on the multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates.

[0127] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the privacy-preserving feature retrieval intersection method provided by the above methods. This method includes: obtaining multiple predicates corresponding one-to-one with multiple dimensions; wherein each predicate includes a feature threshold and a conditional comparison operator; performing a privacy-preserving retrieval on multiple feature vectors to be retrieved in the sender of the privacy-preserving feature retrieval PSI protocol based on the feature threshold in the predicate corresponding to any target dimension among the multiple dimensions, to obtain a first retrieval result of the feature of any target feature vector in the target dimension; obtaining a first polynomial coefficient based on the conditional comparison operator in the predicate corresponding to the target dimension and a first correspondence between the conditional comparison operator and polynomial coefficients; and substituting the first polynomial coefficient and the first retrieval result into a set retrieval quadratic polynomial to obtain a conditional retrieval result of the target feature vector in the target dimension. The system obtains multiple conditional retrieval results corresponding one-to-one with the target feature vector across multiple dimensions; the conditional retrieval results include: 0 or 1; when setting logical connectors for any adjacent predicates among multiple predicates, it obtains the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates according to the second correspondence between the set multiple logical connector types and multiple second polynomial coefficients; the multiple logical connector types include: AND logical connector and OR logical connector; based on the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates and the multiple conditional retrieval results, it obtains a first secure retrieval result; wherein, the receiver secretly shares the obtained multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates with the sender in the Privacy Feature Protection Retrieval (PSI) protocol, to instruct the sender to obtain a second secure retrieval result based on the multiple predicates, first polynomial coefficients, and second polynomial coefficients corresponding to the logical connectors of all adjacent predicates.

[0128] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0129] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A privacy-preserving feature retrieval and intersection method, characterized in that, The method, applied to a recipient in a privacy-preserving retrieval PSI protocol, includes: Obtain multiple predicates that correspond one-to-one with multiple dimensions; wherein each of the multiple predicates includes: a feature threshold and a conditional comparison operator; Based on the feature threshold in the predicate corresponding to any of the multiple dimensions, a privacy-preserving retrieval is performed on the multiple feature vectors to be retrieved in the sender of the privacy feature protection retrieval PSI protocol, and the first retrieval result of the feature of any target feature vector in the target dimension is obtained. Based on the conditional comparison operator in the predicate corresponding to the target dimension, and the first correspondence between the conditional comparison operator and the polynomial coefficients, the first polynomial coefficients are obtained. Substitute the coefficients of the first polynomial and the first search result into the set search quadratic polynomial to obtain the conditional search results of the target feature vector in the target dimension, and obtain multiple conditional search results corresponding one-to-one with the target feature vector in multiple dimensions; the conditional search results include: 0 or 1. When setting logical connectors for any adjacent predicates among the plurality of predicates, the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates are obtained according to the second correspondence between the set plurality of logical connector types and the plurality of second polynomial coefficients; the plurality of logical connector types include: AND logical connector and OR logical connector; Based on the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates and the multiple conditional search results, a first secure search result is obtained; wherein, the receiver secretly shares the obtained multiple predicates, the first polynomial coefficients, and the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates with the sender in the Privacy Feature Protection Search (PSI) protocol, so as to instruct the sender to obtain a second secure search result based on the multiple predicates, the first polynomial coefficients, and the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates.

2. The privacy-preserving feature retrieval and intersection method according to claim 1, characterized in that, After obtaining the multiple conditional retrieval results that correspond one-to-one with the target feature vector across the multiple dimensions, the method further includes: With a quantity threshold set for the target feature vector, the summation of multiple conditional retrieval results corresponding one-to-one with the target feature vector in multiple dimensions is obtained; If the summation result is greater than the quantity threshold, the multiple conditional search results are output as the third secure search result.

3. The privacy-preserving feature retrieval and intersection method according to claim 1, characterized in that, The second polynomial coefficients include the first coefficient of the first term and the second coefficient of the second term; obtaining the first secure search result based on the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates and the multiple conditional search results includes: Based on two target conditional search results connected by a first logical connector among the multiple conditional search results, a secure polynomial is constructed corresponding to the two target conditional search results; the two target conditional search results include: a first target conditional search result and a second target conditional search result; the first term in the secure polynomial is the sum of the first target conditional search result and the second target conditional search result, and the second term in the secure polynomial is the product of the target conditional search result and the second target conditional search result; The first coefficient is determined as the coefficient of the first term in the security polynomial, and the second coefficient is determined as the coefficient of the second term in the security polynomial, to obtain the retrieval results of the two target condition retrieval results, and to obtain the final retrieval result of the two condition retrieval results connected by each logical connector; The final search result of the two conditions connected by all logical connectors is taken as the first secure search result.

4. The privacy-preserving feature retrieval and intersection method according to claim 3, characterized in that, When the first logical connector is of type AND, the first coefficient is 0 and the second coefficient is 1; when the first logical connector is of type OR, the first coefficient is -1 and the second coefficient is 1.

5. The privacy-preserving feature retrieval and intersection method according to claim 1, characterized in that, The conditional comparison operators include the greater than operator, greater than or equal to operator, less than operator, less than or equal to operator, equal to operator, and no feature retrieval operator; The coefficients of the first polynomial include the coefficients of the quadratic term, the coefficients of the linear term, and the coefficients of the constant term; When the conditional operator is the greater than operator, the coefficient of the quadratic term is -1, the coefficient of the linear term is 2, and the coefficient of the constant term is 0. When the conditional operator is the less than or equal to operator, the coefficient of the quadratic term is -0.5, the coefficient of the linear term is 0.5, and the coefficient of the constant term is 1. When the conditional operator is the less than operator, the coefficient of the quadratic term is 0.5, the coefficient of the linear term is -0.5, and the coefficient of the constant term is 0. When the condition operator is the less than or equal to operator, the coefficient of the quadratic term is 1, the coefficient of the linear term is -2, and the coefficient of the constant term is 1. When the conditional operator is the equals operator, the coefficient of the quadratic term is 0.5, the coefficient of the linear term is -1.5, and the coefficient of the constant term is 1. When the conditional operator is the "no feature retrieval" operator, the coefficient of the quadratic term is 0, the coefficient of the linear term is 0, and the coefficient of the constant term is 1.

6. The privacy-preserving feature retrieval and intersection method according to claim 1, characterized in that, The step of performing privacy-preserving retrieval on multiple feature vectors to be retrieved in the sender of the privacy-preserving retrieval PSI protocol based on the feature threshold in the predicate corresponding to any one of the multiple feature vectors, and obtaining a first retrieval result of the feature of any one of the multiple feature vectors on the target dimension, includes: Based on the feature threshold in the predicate corresponding to any of the multiple dimensions, and based on the greater than comparison operation mechanism in Secure Multi-Party Computation (MPC), the features of the target feature vector on the target dimension are calculated to obtain the first comparison result. Based on the feature threshold in the predicate corresponding to any of the multiple dimensions, and based on the less than comparison operation mechanism in the MPC, the features of the target feature vector on the target dimension are calculated to obtain the second comparison result; Based on the first comparison result and the second comparison result, the first search result is obtained.

7. A privacy-preserving feature retrieval and intersection device, characterized in that, A receiver used in privacy-preserving feature retrieval PSI, the apparatus comprising: The first acquisition module is used to acquire multiple predicates that correspond one-to-one with multiple dimensions; wherein, each of the multiple predicates includes: a feature threshold and a condition comparison operator; The retrieval module is used to perform privacy protection retrieval on multiple feature vectors to be retrieved in the sender of the privacy feature protection retrieval PSI protocol based on the feature threshold in the predicate corresponding to any target dimension among the multiple dimensions, and to obtain the first retrieval result of the feature of any target feature vector in the target dimension among the multiple feature vectors; The second acquisition module is used to acquire the first polynomial coefficients based on the conditional comparison operator in the predicate corresponding to the target dimension and the first correspondence between the conditional comparison operator and the polynomial coefficients. The third acquisition module is used to substitute the coefficients of the first polynomial and the first search result into a set search quadratic polynomial to obtain the conditional search results of the target feature vector in the target dimension, and to obtain multiple conditional search results that correspond one-to-one with the target feature vector in multiple dimensions; the conditional search results include: 0 or 1. The fourth acquisition module is used to, when setting logical connectors for any adjacent predicates among the plurality of predicates, acquire the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates according to the second correspondence between the set plurality of logical connector types and the plurality of second polynomial coefficients; the plurality of logical connector types include: AND logical connectors and OR logical connectors; The processing module is configured to obtain a first secure search result based on the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates and the multiple conditional search results; wherein, the receiver secretly shares the obtained multiple predicates, the first polynomial coefficients, and the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates with the sender in the Privacy Feature Protection Search (PSI) protocol, so as to instruct the sender to obtain a second secure search result based on the multiple predicates, the first polynomial coefficients, and the second polynomial coefficients corresponding to the logical connectors of all adjacent predicates.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the privacy-preserving feature retrieval and intersection method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the privacy-preserving feature retrieval and intersection method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the privacy-preserving feature retrieval and intersection method as described in any one of claims 1 to 6.

Images