Data processing method and system, and computing device

By secretly sharing and hash bucket judgments of the feature values ​​corresponding to the ID in the intersection, the problem of ID plaintext exposure in the prior art is solved, and high-security multi-party computing and efficient AI model training are achieved.

WO2024234685A1PCT designated stage expired Publication Date: 2024-11-21HUAWEI CLOUD COMPUTING TECHNOLOGIES CO LTD

Patent Information

Application Number
PCT/CN2024/071000
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-08-23
Filing Date
2024-01-08
Publication Date
2024-11-21

AI Technical Summary

Technical Problem

When the prior art uses the Privacy Set Intersection (PSI) protocol to calculate the data intersection of multiple computing parties, the ID plaintext of the intersection will be exposed, resulting in the privacy of customer IDs leaking and poor security.

Method used

By secretly sharing the eigenvalues ​​corresponding to the ID in the intersection, it is only used as the training data of the neural network when the ID is the intersection ID. Otherwise, random numbers are used to ensure that only the eigenvalues ​​corresponding to the intersection ID participate in model training, avoiding Additional overhead, and termination of operation through hash bucket and threshold judgment to improve model training efficiency.

Benefits of technology

Without exposing the privacy of the customer ID, AI model training is carried out based on the data corresponding to the ID in the intersection, which significantly improves the security of multi-party computing, protects the ID privacy of the customer, and improves the efficiency of model training.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2024071000_21112024_PF_FP_ABST
    Figure CN2024071000_21112024_PF_FP_ABST
Patent Text Reader

Abstract

A data processing method and system, and a computing device. The method comprises: a first computing participant and a second computing participant respectively obtain first ciphertext data; the second computing participant obtains, from the first computing participant, a share of a first feature value corresponding to a first ID; the first computing participant obtains second data, wherein if the first ID is an intersection ID, the second data is a share of a second feature value corresponding to the first ID in the second computing participant, and if the first ID is not the intersection ID, the second data is a random number; and if the first ciphertext data indicates that the first ID is the intersection ID, the first computing participant uses the share of the first feature value and the second data held by the first computing participant as training data of a neural network, and the second computing participant uses the share of the first feature value and the share of the second feature value held by the second computing participant as training data of the neural network. The method can improve the security of multi-party computation.
Need to check novelty before this filing date? Find Prior Art

Description

Data processing method, system and computing device

[0001] This application claims priority to the Chinese patent application filed with the State Intellectual Property Office on May 16, 2023, with application number 202310552730.1 and invention name “Method and device for multi-party vertical federated learning with hidden ID”, and the Chinese patent application filed with the State Intellectual Property Office on August 23, 2023, with application number 202311068411.X and invention name “Method, system and computing device for data processing”, all contents of which are incorporated by reference into this application. Technical Field

[0002] The present application relates to the field of secure computing, and more particularly, to a data processing method, system, and computing device. Background Art

[0003] Private set intersection (PSI) is a secure multi-party computation protocol that can compute the intersection of multiple data sets without exposing any information beyond the intersection. It is commonly used in applications such as joint recommendation, ad targeting, and data alignment before vertical secure multi-party learning. For example, PSI can be used before vertical secure multi-party learning to determine the shared ID value for each participant without exposing the intersection elements of the multiple participants.

[0004] Before vertical secure multi-party learning, the intersection of the data sets of multiple computing participants is calculated through the private set intersection (PSI) technology without exposing the intersection elements of multiple computing participants. The AI ​​model is then trained based on the data (eigenvalues) corresponding to the intersection of the data sets of multiple participants. In related technical solutions, when calculating the intersection of the IDs in the data held by multiple computing participants based on PSI technology, the plaintext IDs in the intersection are exposed. In this way, each computing party will know which IDs are in the intersection, thereby exposing the customer's ID privacy and poor security.

[0005] Therefore, how to improve the security of multi-party computing and protect the data privacy of multiple computing parties has become a technical problem that needs to be solved urgently.

[0006] Summary of the Invention

[0007] The present application provides a data processing method, system, and computing device, which can improve the security of multi-party computing and protect the data privacy of multiple computing parties.

[0008] In a first aspect, a method for data processing is provided, the method comprising: a first computing participant and a second computing participant respectively obtain first ciphertext data, the first ciphertext data being used to indicate whether the first ID of the first computing participant is an intersection ID, the intersection ID indicating that the first ID is the same as any one of the at least one ID of the second computing participant; the second computing participant obtains a share of a first eigenvalue corresponding to the first ID from the first computing participant, the share of the first eigenvalue being the first computing participant sharing the first eigenvalue between the first computing participant and the second computing participant in a secret sharing manner, the shares of the first eigenvalue each obtained by the first computing participant and the second computing participant being able to restore the first eigenvalue; the first computing participant obtains second data, wherein, if the first ID is an intersection ID, ID, the second data is the share of the second eigenvalue corresponding to the first ID in the second computing participant. If the first ID is not an intersection ID, the second data is a random number. The share of the second eigenvalue is the second eigenvalue shared by the second computing participant between the first computing participant and the second computing participant in a secret sharing manner. The share of the second eigenvalue obtained by the first computing participant and the second computing participant can restore the second eigenvalue; if the first ciphertext data indicates that the first ID is an intersection ID, the first computing participant uses the share of the first eigenvalue and the second data held by it as training data of the neural network; if the first ciphertext data indicates that the first ID is an intersection ID, the second computing participant uses the share of the first eigenvalue and the share of the second eigenvalue held by it as training data of the neural network.

[0009] In the above technical solution, subsequent AI model training can be performed based on the data (eigenvalues) corresponding to the ID in the intersection without exposing the customer's ID privacy, thereby protecting the customer's ID privacy and providing high security, meeting the functional and privacy requirements required in the data preprocessing stage of vertical multi-party secure learning.

[0010] In combination with the first aspect, in certain implementations of the first aspect, the method further includes: the first computing participant and the second computing participant respectively recover the first ciphertext data to obtain first data; and determine whether the first ID is the intersection ID based on the first data.

[0011] In combination with the first aspect, in certain implementations of the first aspect, the method also includes: if the first computing participant indicates based on the first ciphertext data obtained by it that the first ID is not an intersection ID, the first computing participant deletes its share of the first eigenvalue and the second data; if the second computing participant indicates based on the first ciphertext data obtained by it that the first ID is not an intersection ID, the second computing participant deletes its share of the first eigenvalue and the share of the second eigenvalue.

[0012] In the above technical solution, if the first ID is not the intersection ID, then the second data corresponding to it is a random number. The second data can be deleted here, so that only the characteristic values ​​corresponding to the intersection ID can participate in the training of the neural network model, avoiding additional overhead and improving the efficiency of model training.

[0013] In combination with the first aspect, in certain implementations of the first aspect, the method further includes: the first computing participant and the second computing participant determine whether the number of intersection IDs is greater than a preset threshold based on multiple ciphertext data corresponding to multiple hash buckets, and the multiple ciphertext data include the first ciphertext data; if the number of intersection IDs is greater than the preset threshold, the second computing participant obtains the share of the first eigenvalue corresponding to the first ID from the first computing participant; if the number of intersection IDs is greater than the preset threshold, the first computing participant obtains the second data.

[0014] The above technical solution provides a threshold judgment. If the number of elements in the intersection is small, the subsequent operation is terminated. This avoids insufficient data, resulting in suboptimal model training results and wasted training costs.

[0015] In combination with the first aspect, in certain implementations of the first aspect, the method further includes: the first computing participant and the second computing participant randomly scramble the order of the multiple ciphertext data respectively obtained.

[0016] In combination with the first aspect, in certain implementations of the first aspect, the first computing participant and the second computing participant respectively obtain the first ciphertext data through a programmable oblivious pseudo-random function OPPRF or a key exchange DH.

[0017] In combination with the first aspect, in certain implementations of the first aspect, the first computing participant obtains the second data through a programmable oblivious pseudo-random function OPPRF.

[0018] In combination with the first aspect, in certain implementations of the first aspect, the first ID of the first computing participant is the ID of the first computing participant in the first hash bucket, the at least one ID of the second computing participant is the at least one ID of the second computing participant in the first hash bucket, the first ciphertext data is the ciphertext data corresponding to the first hash bucket, and the second data is the second data corresponding to the first hash bucket.

[0019] In a second aspect, a data processing system is provided, which includes a first computing participant and a second computing participant, wherein the first computing participant and the second computing participant are respectively used to obtain first ciphertext data, and the first ciphertext data is used to indicate whether the first ID of the first computing participant is an intersection ID, and the intersection ID indicates that the first ID is the same as any one of the at least one ID of the second computing participant; the second computing participant is also used to obtain a share of a first eigenvalue corresponding to the first ID from the first computing participant, and the share of the first eigenvalue is the first computing participant sharing the first eigenvalue between the first computing participant and the second computing participant in a secret sharing manner, and the shares of the first eigenvalue obtained by each of the first computing participant and the second computing participant can restore the first eigenvalue; the first computing participant is also used to obtain second data, wherein, If the first ID is an intersection ID, the second data is the share of the second eigenvalue corresponding to the first ID in the second computing participant; if the first ID is not an intersection ID, the second data is a random number, and the share of the second eigenvalue is the second eigenvalue shared by the second computing participant between the first computing participant and the second computing participant in a secret sharing manner, and the share of the second eigenvalue obtained by the first computing participant and the second computing participant respectively can restore the second eigenvalue; if the first ciphertext data indicates that the first ID is an intersection ID, the first computing participant is also used to use the share of the first eigenvalue it holds and the second data as training data for the neural network; if the first ciphertext data indicates that the first ID is an intersection ID, the second computing participant is also used to use the share of the first eigenvalue it holds and the share of the second eigenvalue it holds as training data for the neural network.

[0020] In combination with the second aspect, in certain implementations of the second aspect, the first computing participant and the second computing participant are also respectively used to recover the first ciphertext data to obtain the first data; the first computing participant and the second computing participant are also respectively used to determine whether the first ID is the intersection ID based on the first data.

[0021] In combination with the second aspect, in certain implementations of the second aspect, if the first computing participant indicates based on the first ciphertext data obtained by it that the first ID is not an intersection ID, the first computing participant is also used to delete the share of the first eigenvalue and the second data held by it; if the second computing participant indicates based on the first ciphertext data obtained by it that the first ID is not an intersection ID, the second computing participant is also used to delete the share of the first eigenvalue and the share of the second eigenvalue held by it.

[0022] In combination with the second aspect, in certain implementations of the second aspect, the first computing participant and the second computing participant are further used to determine whether the number of intersection IDs is greater than a preset threshold based on multiple ciphertext data corresponding to multiple hash buckets, and the multiple ciphertext data include the first ciphertext data; the second computing participant is specifically used to obtain the share of the first eigenvalue corresponding to the first ID from the first computing participant when the number of intersection IDs is greater than the preset threshold; the first computing participant is specifically used to obtain the second data when the number of intersection IDs is greater than the preset threshold.

[0023] In combination with the second aspect, in certain implementations of the second aspect, the first computing participant and the second computing participant are further configured to randomly disrupt the order of the multiple ciphertext data respectively obtained.

[0024] In combination with the second aspect, in certain implementations of the second aspect, the first computing participant and the second computing participant are specifically configured to obtain the first ciphertext data respectively through a programmable oblivious pseudo-random function OPPRF or a key exchange DH.

[0025] In combination with the second aspect, in certain implementations of the second aspect, the first computing participant is specifically configured to obtain the second data through a programmable oblivious pseudo-random function OPPRF.

[0026] In combination with the second aspect, in certain implementations of the second aspect, the first ID of the first computing participant is the ID of the first computing participant in the first hash bucket, the at least one ID of the second computing participant is the at least one ID of the second computing participant in the first hash bucket, the first ciphertext data is the ciphertext data corresponding to the first hash bucket, and the second data is the second data corresponding to the first hash bucket.

[0027] In a third aspect, a computing device cluster is provided, comprising at least one computing device, each computing device comprising a processor and a memory; the processor of the at least one computing device is used to execute instructions stored in the memory of the at least one computing device, so that the computing device cluster executes the method in the first aspect or any possible implementation of the first aspect.

[0028] Optionally, the processor may be a general-purpose processor, which may be implemented in hardware or software. When implemented in hardware, the processor may be a logic circuit, an integrated circuit, or the like; when implemented in software, the processor may be a general-purpose processor implemented by reading software code stored in a memory, which may be integrated into the processor or located independently of the processor.

[0029] In a fourth aspect, a chip is provided, which obtains instructions and executes the instructions to implement the method in the above-mentioned first aspect and any implementation manner of the first aspect.

[0030] Optionally, as an implementation, the chip includes a processor and a data interface, and the processor reads instructions stored in the memory through the data interface to execute the method in the above-mentioned first aspect and any implementation of the first aspect.

[0031] Optionally, as an implementation method, the chip may also include a memory, in which instructions are stored, and the processor is used to execute the instructions stored on the memory. When the instructions are executed, the processor is used to execute the method in the first aspect and any one of the implementation methods of the first aspect.

[0032] In a fifth aspect, a computer program product comprising instructions is provided. When the instructions are executed by a computing device cluster, the computing device cluster executes the method in the first aspect and any one of the implementations of the first aspect.

[0033] In a sixth aspect, a computer-readable storage medium is provided, comprising computer program instructions. When the computer program instructions are executed by a computing device cluster, the computing device cluster executes the method according to the first aspect and any one of the implementations of the first aspect.

[0034] By way of example, these computer-readable storages include, but are not limited to, one or more of the following: read-only memory (ROM), programmable ROM (PROM), erasable PROM (EPROM), Flash memory, electrically EPROM (EEPROM), and a hard drive.

[0035] Optionally, as an implementation manner, the above-mentioned storage medium may specifically be a non-volatile storage medium. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] FIG1 is a schematic diagram of vertically distributed data held by computing participants in an embodiment of the present application.

[0037] FIG2 is a schematic flow chart of a data processing method provided in an embodiment of the present application.

[0038] FIG3 is a schematic diagram of the results of a first computing participant and a second computing participant performing hashing on an ID set in the data they hold, respectively, according to an embodiment of the present application.

[0039] FIG4 is a diagram of a first computing participant and a second computing participant obtaining a secret shared value of y through OPPRF provided in an embodiment of the present application. <y>Schematic block diagram of .

[0040] FIG5 is a diagram of a first computing participant and a second computing participant obtaining a value through OPPRF provided in an embodiment of the present application. <fs>Schematic block diagram of .

[0041] FIG6 is a schematic diagram of data selection provided in an embodiment of the present application.

[0042] FIG7 is a schematic diagram of another data selection method provided in an embodiment of the present application.

[0043] FIG8 is a schematic block diagram of a data processing system 800 provided in an embodiment of the present application.

[0044] FIG9 is a schematic diagram of the architecture of a computing device 1500 provided in an embodiment of the present application.

[0045] FIG10 is a schematic diagram of the architecture of a computing device cluster provided in an embodiment of the present application.

[0046] FIG11 is a schematic diagram of a network connection between computing devices 1500A and 1500B provided in an embodiment of the present application.

[0047] FIG12 is a schematic block diagram of a cloud scenario applicable to an embodiment of the present application. DETAILED DESCRIPTION

[0048] The technical solution in this application will be described below with reference to the accompanying drawings.

[0049] This application will present various aspects, embodiments, or features around systems including multiple devices, components, modules, etc. It should be understood and appreciated that each system may include additional devices, components, modules, etc., and / or may not include all of the devices, components, modules, etc. discussed in conjunction with the figures. Furthermore, combinations of these aspects may also be used.

[0050] Additionally, in the embodiments of this application, words such as "exemplary" and "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described in this application as "exemplary" should not be construed as preferred or advantageous over other embodiments or designs. Rather, the use of the word "exemplary" is intended to present concepts in a concrete manner.

[0051] In the embodiments of the present application, “corresponding” and “relevant” may sometimes be used interchangeably. It should be noted that when the distinction between them is not emphasized, the meanings they intend to express are consistent.

[0052] The business scenarios described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. A person skilled in the art will appreciate that, with the evolution of network architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of the present application are equally applicable to similar technical problems.

[0053] References to "one embodiment" or "some embodiments" throughout this specification mean that a particular feature, structure, or characteristic described in connection with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in yet other embodiments" appearing in various places throughout this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically stated. The terms "including," "comprising," "having," and variations thereof mean "including, but not limited to," unless otherwise specifically stated.

[0054] In this application, "at least one" means one or more, and "more" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: including the existence of A alone, the existence of A and B at the same time, and the existence of B alone, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or plural.

[0055] For the convenience of description, the relevant concepts involved in the embodiments of the present application are first described in detail below.

[0056] 1. Neural Networks

[0057] A neural network is a computer system that mimics the functionality of biological neural networks, learning and training from data to accomplish various tasks. It is a computational model composed of multiple nodes, each of which mimics the behavior of a biological neuron, receiving and processing input signals and generating output signals by learning and adjusting weights.

[0058] Neural networks typically consist of multiple layers of nodes, with the input layer receiving external data input, the output layer generating the final output, and the intermediate hidden layers processing and transforming the input. In a neural network, the connections between nodes have different weights, which can be adjusted through training and learning to improve the network's accuracy and performance. Neural networks are widely used in various fields, including image recognition, speech recognition, natural language processing, autonomous driving, medical diagnosis, and financial forecasting. Common neural network architectures include feedforward neural networks, convolutional neural networks, recurrent neural networks, and deep neural networks.

[0059] 2. Secure multi-party learning (SMPL)

[0060] Secure multi-party learning is a distributed machine learning paradigm that allows multiple parties to collaboratively train AI models using their own data, without aggregating data from multiple parties (organizations or users). Traditional machine learning paradigms require the aggregation of large amounts of data for model training, and this training data is likely to come from multiple different organizations or users. Aggregating data from multiple different organizations or users poses a significant risk of data leakage, exposing information assets for organizations and potentially compromising personal privacy for individual users.

[0061] 3. Vertical secure multi-party learning (VSMPL)

[0062] Vertical secure multi-party learning (SML) involves combining the data held by multiple participants using secure multi-party computing (SMC) to train AI models, using vertically distributed data to protect data privacy while maximizing its value. In SML, participants may be different organizations, such as hospitals, banks, or government agencies. These organizations typically hold user IDs and their corresponding features, with different participants holding different features. This data is sensitive and private, so aggregating SML data from different organizations to train AI models can lead to data privacy leaks. To address the privacy concerns of SML, SML technology has emerged. SML uses private set intersection technology to align the data of multiple participants, then employs secure multi-party computing techniques such as secret sharing and obfuscation circuits to jointly train the AI ​​model.

[0063] 4. Oblivious programmable pseudorandom function (OPPRF)

[0064] A programmable oblivious pseudorandom function (OPRF) is a cryptographic protocol that implements the following functionality: Alice and Bob input an input x into the OPPRF, and Bob inputs a number of (xi, yi) pairs. After the OPPRF, Alice receives a value y. If Alice's x equals one of Bob's xi, then y is yi; otherwise, y is a random number r. During the entire OPPRF process, neither Alice nor Bob receives any additional information. Bob does not know Alice's input x or the value y she receives. Alice only receives a string y. Based on y, she cannot determine whether she received a random number r or yi, and therefore does not know whether her x equals one of Bob's xi. Alice is generally referred to as the receiver, and Bob as the sender.

[0065] 5. Cuckoo hashing

[0066] Cuckoo hashing is an implementation of a hash table that solves the problem of hash collisions. In cuckoo hashing, each element has two hash functions and can be assigned to a position in two hash tables. If the position in the first hash table is already occupied, the element is placed in the corresponding position in the other hash table. If the corresponding position in the other hash table is also occupied, the element is squeezed out of its original position and placed in the new position. This process may be repeated many times until all elements can be successfully placed in the hash table. Therefore, when using cuckoo hashing for a set, each bucket contains only one element from the set, and some elements may be placed in the stash if they cannot find a position. The average time complexity of cuckoo hashing is constant, making it a very efficient hash table implementation.

[0067] 6. Private set intersection (PSI)

[0068] Private Set Intersection (PSI) is a secure multi-party computation protocol that can calculate the intersection of multiple data sets without exposing any other data set information. It is commonly used in joint recommendations, ad targeting, and data alignment before vertical secure multi-party learning. For example, before vertical secure multi-party learning, PSI can be used to calculate the ID value shared by all participants without exposing the intersection elements of multiple participants.

[0069] PSI has been developed for nearly 20 years. Technologies implementing PSI include, but are not limited to, key exchange (Diffie-Hellman, DH), oblivious transfer (OT), oblivious pseudorandom function (OPRF), programmable oblivious pseudorandom function (OPPRF), and homomorphic encryption (HE). Initial PSI schemes could calculate the intersection of two parties without exposing any information beyond the intersection. Subsequent PSI schemes have focused on improvements in scenarios, functionality, privacy, and efficiency. In terms of scenarios, there are PSI schemes designed for imbalanced datasets. In terms of functionality, there are multi-party PSI schemes that can support more than two parties, as well as circuit-PSI schemes that support calculations using the associated values ​​corresponding to key elements in the intersection. In terms of privacy, there are PSI schemes that do not expose the intersection results. In terms of efficiency, methods using hashing to reduce computational and communication overhead have been proposed.

[0070] 7. Secret Sharing

[0071] Secret sharing, also known as secret splitting, is a cryptographic concept used to split a secret value into multiple shares so that it can be shared among multiple entities. Each entity receives a portion of the secret, called a share or shard. Depending on how the secret is split, the original secret value can be recovered after collecting enough shares.

[0072] It should be understood that different segmentation methods form different secret sharing technologies, common ones include additive secret sharing (ASS) and Boolean secret sharing (BSS).

[0073] 8. Additive secret sharing (ASS)

[0074] Additive secret sharing is an instance of secret sharing that splits a secret value into multiple shares so that it can be shared among multiple parties. In additive secret sharing, the original secret is represented as an integer and can be shared by splitting it into multiple parts. The original secret value can only be recovered when all the shares are reassembled. This method can be used to implement secure multi-party computation, such as data analysis while protecting individual privacy, or secure computation in distributed computing.

[0075] One method for splitting additive secret sharing is to use domain addition. For example, there are three participants. One participant holds the original secret value of 10. This is split into three parts on a ring of size 16, generating two random numbers, 13 and 7, respectively. The final shared values ​​are 13, 7, and 6 (= (10-13-6) mod 16). The participant holding the secret value 10 sends two of these three shares to the other two participants. The other two participants only hold one share and cannot recover the original share value. Only by adding the three shared values ​​together can the original secret value 10 be recovered (= (13+7+6) mod 16).

[0076] 9. Boolean secret sharing (BSS)

[0077] Boolean secret sharing uses XOR to generate a secret value using multiple random numbers. The XOR value of these random numbers is the original secret value. For example, there are three participants. One participant holds the original secret value 0b010. This is split into three parts on a ring of size 8, generating two random numbers: 0b110 and 0b011. The final shared values ​​are 0b110, 0b011, and 0b111 (= 0b010^0b110^0b011). The participant holding the secret value 0b010 sends two of the three shared values ​​to the other two participants. The other two participants only hold one share and cannot recover the original shared value. Only by XORing all three shared values ​​together can the original secret value 0b010 (= 0b111^0b110^0b011) be recovered.

[0078] Before vertical secure multi-party learning, the intersection of the data sets of multiple computing participants is calculated through the private set intersection (PSI) technology without exposing the intersection elements of multiple computing participants. The AI ​​model is then trained based on the data (eigenvalues) corresponding to the intersection of the data sets of multiple participants. In related technical solutions, when calculating the intersection of the IDs in the data held by multiple computing participants based on PSI technology, the plaintext IDs in the intersection are exposed. In this way, each computing party will know which IDs are in the intersection, thereby exposing the customer's ID privacy and poor security.

[0079] In view of this, an embodiment of the present application provides a data processing method that can perform subsequent AI model training based on the data (feature values) corresponding to the ID in the intersection without exposing the customer's ID privacy, thereby protecting the customer's ID privacy and providing higher security.

[0080] It should be understood that the methods provided in the embodiments of this application are applied in scenarios where multiple organizations jointly train a secure multi-party learning model, and the preprocessing process for AI model training is completed when the data is vertically distributed. Specifically, this scenario may include, but is not limited to, cross-company or cross-industry data joint modeling scenarios, and there are no requirements for the specific data distribution method or the number of participants.

[0081] As an example, let's take a scenario involving joint modeling between the financial industry and carriers. For example, multiple financial industry users (such as banks and securities firms) and carriers (such as China Mobile, China Telecom, and China Unicom) use their own user data for joint modeling to improve the accuracy of financial industry models (such as credit score models and credit card default models). As shown in Figure 1, the data held by these different organizations is distributed vertically. Each organization has user ID values ​​and feature values, and the ID values ​​and feature values ​​of different organizations are not exactly the same. These different organizations want to use the feature data (feature values) of common users to train an assessment model to predict the user's credit level.

[0082] Figure 2 is a schematic flowchart of a data processing method provided in an embodiment of the present application. The method is applied to the preprocessing stage of data before AI model training. The method may include steps 210-240, and steps 210-240 are described in detail below.

[0083] Step 210: Determine whether there is an intersection ID between the ID elements in the hash bucket with the same number of the first computing participant and the second computing participant.

[0084] In the embodiment of the present application, the first computing participant and the second computing participant respectively map the IDs in the user data they hold to hash buckets.

[0085] The present application does not impose any specific restrictions on the number of computing participants, as long as it is greater than one computing participant. FIG2 illustrates the example of two computing participants (a first computing participant and a second computing participant).

[0086] In an embodiment of the present application, the user data held by each computing participant may include multiple pairs of (x, Feature), where x is the ID and Feature is the feature value corresponding to the ID. The feature value may be a feature value corresponding to the ID, or may be multiple feature values ​​corresponding to the ID. The embodiment of the present application does not make specific limitations on this.

[0087] For example, the ID in a bank's user data is the account's mobile phone number, and the characteristic value is the deposit and withdrawal information corresponding to the ID; the ID in a brokerage firm's user data is the account's mobile phone number, and the characteristic value is the securities trading information corresponding to the ID; the ID in an operator's user data is the account's mobile phone number, and the characteristic value is the call records corresponding to the ID, etc.

[0088] In an embodiment of the present application, it is assumed that the first computing participant uses the cuckoo hash algorithm to map the ID in its own user data to the hash bucket, and the second computing participant uses the ordinary hash algorithm to map the ID in its own user data to the hash bucket.

[0089] It should be noted that the first computing participant using the Cuckoo Hash algorithm can also be called the leader participant, that is, the leader participant uses the Cuckoo Hash algorithm to map the ID in its own user data to the hash bucket, while other participants who are not leader participants use the ordinary hash algorithm to map the ID in their own user data to the hash bucket.

[0090] It should be understood that the above-mentioned leader participant is randomly selected from at least two computing participants. For ease of description, FIG2 takes the first computing participant as the leader participant as an example for description.

[0091] For example, as shown in Figure 3, for the first computing participant, since it performs cuckoo hashing on its own ID set, each bucket contains only one ID element. For the second computing participant, since it performs ordinary hashing on its own ID set, each bucket contains multiple ID elements. Because the first and second computing participants use the same hash function, if the user data of the two computing participants contains the same ID elements, these same ID elements will be hashed into the same numbered hash bucket.

[0092] It should be understood that the present embodiment does not specifically limit the number of hash buckets. Figure 3 illustrates an example where the amount of data in each computing participant's ID set is n and there are n hash buckets. If the amount of data in a computing participant's ID set is less than n, random numbers can be generated to make up the difference.

[0093] In an embodiment of the present application, it is possible to determine whether there is an intersection between the ID elements in the hash bucket of the same number of the first computing participant and the second computing participant and the ID elements in the hash bucket of the same number.

[0094] For ease of description, the following uses hash bucket numbered 1 (hash bucket #1) as an example to determine whether there is an intersection between the ID element of the first computation participant in hash bucket #1 and the ID element of the second computation participant in hash bucket #1. The execution process for other numbered hash buckets is the same as that for hash bucket #1 and is not repeated here.

[0095] As an example, the first computing participant and the second computing participant can respectively obtain the secret shared value of y ( <y>), where y is used to indicate whether the ID element of the first computing participant in hash bucket No. 1 is an intersection ID, that is, whether there is an ID element in the ID element of the second computing participant in hash bucket No. 1 that is equal to the ID element of the first computing participant in hash bucket No. 1. For example, if y is 1, it can be said that there is an intersection between the ID element of the first computing participant in hash bucket No. 1 and the ID element of the second computing participant in hash bucket No. 1, and the ID element of the first computing participant in hash bucket No. 1 is the intersection ID. If y is 0, it can be said that there is no intersection between the ID element of the first computing participant in hash bucket No. 1 and the ID element of the second computing participant in hash bucket No. 1, and the ID element of the first computing participant in hash bucket No. 1 is not an intersection ID.

[0096] It should be understood that the secret sharing value of y mentioned above can be a share or a slice of y, and each computing participant holds a share of y. The shares of y held by each computing participant can be combined to obtain y.

[0097] In a possible implementation, the OPPRF protocol can be used to enable the first computing participant and the second computing participant to obtain the secret shared value of y respectively ( <y>As an example, the first computing participant and the second computing participant can obtain the secret sharing value of y respectively based on the forwarding of the cloud server through the OPPRF protocol ( <y>For example, FIG12 is a schematic block diagram of a cloud scenario, and the cloud server is a server of the cloud data center shown in FIG12 .

[0098] The cloud scenario in Figure 12 may include a cloud management platform 110, the internet 120, and a client 130. Cloud management platform 110 manages the infrastructure that provides multiple cloud services. This infrastructure includes multiple cloud data centers, each of which contains multiple servers. Each server includes cloud service resources and provides corresponding cloud services to tenants. Cloud management platform 110 may be located in the cloud data center and may provide an access interface (such as a user interface or application program interface (API)). Tenants can use client 130 to remotely access the access interface, register a cloud account and password on cloud management platform 110, and log in to cloud management platform 110. After cloud management platform 110 successfully authenticates the cloud account and password, the tenant can then pay to select and purchase a virtual machine with specific specifications (processor, memory, and disk) on cloud management platform 110. After the purchase is successful, cloud management platform 110 provides the remote login account and password for the purchased virtual machine. Client 130 can then remotely log in to the virtual machine and install and run the tenant's applications within it. Therefore, tenants can create, manage, log in to, and operate virtual machines in the cloud data center through cloud management platform 110. Among them, virtual machines can also be called cloud servers (elastic compute services, ECS) or elastic instances (different cloud service providers have different names).

[0099] It should be understood that tenants of cloud services can be individuals, enterprises, schools, hospitals, administrative agencies, etc.

[0100] The functions of the cloud management platform 110 include, but are not limited to, a user console, computing management services, network management services, storage management services, authentication services, and image management services. The user console provides an interface or API for interacting with tenants. The computing management service is used to manage servers running virtual machines and containers, as well as bare metal servers. The network management service is used to manage network services (such as gateways and firewalls). The storage management service is used to manage storage services (such as data bucket services). The authentication service is used to manage tenant accounts and passwords. The image management service is used to manage virtual machine images. Tenants can use the client 130 to log in to the cloud management platform 110 via the Internet 120 to manage the rented cloud services.

[0101] The following is a case where the first computing participant and the second computing participant obtain the secret shared value of y through the OPPRF protocol ( <y>) is described in detail.

[0102] For example, assume that the ID element of the first computing participant in hash bucket No. 1 is x0, and the ID elements of the second computing participant in hash bucket No. 1 are x1, x2...xm. The first computing participant inputs x0 in hash bucket No. 1 as input information to the OPPRF, and the second computing participant inputs multiple pairs (xi, yi) in hash bucket No. 1 as input information to the OPPRF, where the yi corresponding to the xi in the same bucket are all set to the same random number y*, that is, y1=y2=...ym=y*. The second computing participant uses the y* set by itself as the secret shared value of y held by itself. <y>If x0 is an intersection element, that is, the second computing participant has an xi in x1, x2···xm in hash bucket 1 that is equal to x0, the secret shared value of y output by OPPRF to the first computing participant is <y>The secret sharing value is 1. If x0 is not an intersection element, that is, the second computing participant does not have a xi in x1, x2...xm in hash bucket No. 1 that is equal to x0, the secret sharing value of y output by OPPRF to the first computing participant is <y>A secret sharing value of 0.

[0103] The secret shared value of y held by the first computing participant and the second computing participant <y>Combining them together we get y, which has a value of either 0 or 1.

[0104] In some embodiments, if the multi-party secure computing scenario includes three or more computing participants, for example, including the first computing participant, the second computing participant, the third computing participant, and the fourth computing participant, the first computing participant and the second computing participant execute the process shown in FIG4 so that the first computing participant obtains the secret sharing value of y01. <y01>Similarly, the first computing participant and the third computing participant execute the process shown in Figure 4 so that the first computing participant obtains the secret shared value of y02 <y02>The first computing participant and the fourth computing participant execute the process shown in FIG4 so that the first computing participant obtains the secret shared value of y03. <y03>The first computing participant can obtain <y01> 、 <y02> 、 <y02>Perform XOR and use the result of XOR as the secret shared value of y held by oneself <y>.

[0105] Step 220: Perform a threshold determination on the number of IDs in the intersection ID.

[0106] In the embodiment of the present application, after the ID intersection in step 220, each computing participant will obtain the secret shared value of y <y>, which indicates whether x in the first calculation participant is an intersection element. If y is 1, then x is an intersection element; if y is 0, then x is not an intersection element.

[0107] In step 220, assuming there are n hash buckets, the first computing participant obtains the n secret shared values ​​of y corresponding to the n hash buckets. <y>The sum is then added, and the result of the addition is the number of IDs in the intersection ID. A threshold is then applied to the sum. If the sum is greater than the preset threshold, step 230 is executed. If the sum is less than the preset threshold, the process ends.

[0108] The embodiment of the present application can provide a threshold judgment. If the number of elements in the intersection is small, the subsequent operation is terminated to avoid insufficient data, unsatisfactory training model results, and waste of training costs.

[0109] Step 230: The first computing participant and the second computing participant each obtain the secret shared value of the feature value feature corresponding to the ID in the intersection <fs>.

[0110] For ease of description, a bucket, such as a hash bucket numbered 1 (hash bucket No. 1), is used as an example for explanation below.

[0111] In the embodiment of the present application, for the first computing participant, there is only one ID element x0 in the hash bucket No. 1, so the first computing participant directly shares the secret feature value corresponding to x0 with the second computing participant. That is, the first computing participant and the second computing participant each hold the secret sharing value of the feature value feature corresponding to x0. <fs>.

[0112] In the embodiment of the present application, for the second computing participant, there are multiple ID elements (x1, x2...xm) in hash bucket No. 1. After the ID intersection is calculated, the second computing participant does not know whether there is an intersection ID in hash bucket No. 1, nor does it know which ID element is the intersection ID. Therefore, it does not know which ID element's characteristic value it should share.

[0113] As an example, as shown in Figure 5, the second computing participant and the first computing participant execute OPPRF. The first computing participant inputs x0 in hash bucket No. 1 as input information to OPPRF, and the second computing participant inputs multiple pairs (xi, yi) in hash bucket No. 1 as input information to OPPRF, where the yi corresponding to xi in the same bucket are all set to feature-r, where feature is the eigenvalue corresponding to xi, and the r values ​​of different elements in the same bucket are the same. If x0 is an intersection element, that is, the second computing participant has an xi in x1, x2...xm in hash bucket No. 1 that is equal to x0, OPPRF outputs <fs>f10=fi-r, where fi is the eigenvalue corresponding to xi (xi=x0). The second computing participant holds r. If x0 is not an intersection element, that is, the second computing participant does not have any xi equal to x0 in the x1, x2...xm of the hash bucket No. 1, OPPRF outputs <fs>A meaningless random number.

[0114] In some embodiments, if the multi-party secure computing scenario includes three or more computing participants, for example, including the first computing participant, the second computing participant, the third computing participant, and the fourth computing participant. The first computing participant can also obtain f20 and f30 through the above, where f20 is the amount shared by the third computing participant to the first computing participant. <fs>, f30 is the amount shared by the fourth computing participant to the first computing participant <fs>The first computing participant can also share the f10, f20, and f30 it obtains with the computing participants that do not have the same share, and secretly share the eigenvalue corresponding to its own x0 with each computing participant that does not have the same share.

[0115] Specifically, the first computing participant secretly shares the eigenvalue corresponding to x0 with the third computing participant and the fourth computing participant, shares f10 with the third computing participant and the fourth computing participant, shares f20 with the second computing participant and the fourth computing participant, and shares f30 with the second computing participant and the third computing participant.

[0116] Step 240: According to the obtained values ​​of multiple y, multiple <fs>Perform data screening to obtain training data for model training.

[0117] In the embodiment of the present application, taking n hash buckets as an example, each computing participant can obtain n <y>, and with n <y>The corresponding n <fs>By <y>Recover n y and make n calculations based on the value of y <fs>For example, since the value of y is 1, it is used to represent that the first computing participant's x in a certain hash bucket is an intersection element, and the corresponding <fs>The secret value includes the characteristic value corresponding to x of each participant. The value y is 0, which is used to indicate that the x of the first computing participant in a certain hash bucket is not an intersection element. <fs>is a meaningless random number. Therefore, we can keep the value of 1 corresponding to y <fs>, delete the y value of 0 <fs>and will retain <fs>As training data for subsequent model training.

[0118] Through the above data selection, only the secret shared data in the intersection is used for subsequent model training, avoiding additional overhead.

[0119] Optionally, in some embodiments, before step 240, the obtained n <y>For example, n numbers can be obtained by secure shuffle. <y>order.

[0120] For example, as shown in Figure 6, there are two steps: the first step is to generate a selection pool, and the second step is to obtain training data. The selection pool in the first step consists of multiple one-hot vectors with secret shared values. The one-hot vectors have the same subscript position corresponding to the intersection ID of the first computation participant as 1, and all other positions as 0. Therefore, the number of one-hot vectors is the same as the number of intersection IDs, and each one-hot vector represents a selected piece of merged data. The second step is to obtain training data by performing a secure multi-party dot product operation using the one-hot vectors and the merged data. The number of one-hot vectors corresponding to the number of training data is randomly selected from the selection pool.

[0121] Specifically, as shown in Figure 7, first, the y list obtained by ID intersection is regarded as a vector, and the record <ys>.Will <ys>After secure multi-party scrambling, the order is<ys_shuffle> , that is, each participant does not know the mapping relationship before and after the scrambling. Then the scrambled vector<ys_shuffle> Restore to plaintext. Generate a pool of plaintext one-hot vectors based on the positions of the 1s. However, the positions of the 1s in this pool are not the actual positions of the intersection elements, but rather the scrambled positions. Therefore, secure multi-party scrambling is then used to restore the secret shared value of the intersection's true one-hot vectors.

[0122] In the embodiments of the present application, the functional and privacy requirements required for the data preprocessing stage of vertical multi-party secure learning are met, and relevant cryptographic protocols are designed to complete the functional and privacy requirements required for the data preprocessing stage.

[0123] The method provided by the embodiment of the present application is described in detail above in conjunction with Figures 1 to 7 . The system embodiment of the present application will be described in detail below in conjunction with Figures 8 to 11 . It should be understood that the description of the method embodiment corresponds to the description of the system embodiment. Therefore, for portions not described in detail, reference can be made to the above method embodiment.

[0124] 8 is a schematic block diagram of a data processing system 800 provided in an embodiment of the present application. The system 800 can be implemented through software, hardware, or a combination of both. The system 800 provided in the embodiment of the present application can implement the method flow shown in the embodiment of the present application, and the system 800 includes: a first computing participant 810 and a second computing participant 820, wherein the first computing participant 810 and the second computing participant 820 are respectively used to obtain first ciphertext data, and the first ciphertext data is used to indicate whether the first ID of the first computing participant 810 is an intersection ID, and the intersection ID indicates that the first ID is the same as any one of the at least one ID of the second computing participant 820; the second computing participant 820 is also used to obtain a share of the first eigenvalue corresponding to the first ID from the first computing participant 810, and the share of the first eigenvalue is the first computing participant sharing the first eigenvalue between the first computing participant 810 and the second computing participant 820 in a secret sharing manner, and the shares of the first eigenvalue obtained by each of the first computing participant 810 and the second computing participant 820 can restore the first eigenvalue; the first computing participant 810 is also used to obtain the second data, wherein, if the first ID is an intersection ID, the second data is the share of the second eigenvalue corresponding to the first ID in the second computing participant 820; if the first ID is not an intersection ID, the second data is a random number, and the share of the second eigenvalue is the second computing participant 820 sharing the second eigenvalue between the first computing participant 810 and the second computing participant 820 in a secret sharing manner, and the shares of the second eigenvalue obtained by the first computing participant 810 and the second computing participant 820 respectively can restore the second eigenvalue; if the first ciphertext data indicates that the first ID is an intersection ID, the first computing participant 810 is also used to use the share of the first eigenvalue it holds and the second data as training data for the neural network; if the first ciphertext data indicates that the first ID is an intersection ID, the second computing participant 820 is also used to use the share of the first eigenvalue it holds and the share of the second eigenvalue it holds as training data for the neural network.

[0125] Optionally, the first computing participant 810 and the second computing participant 820 are also used to recover the first ciphertext data to obtain first data; the first computing participant 810 and the second computing participant 820 are also used to determine whether the first ID is the intersection ID based on the first data.

[0126] Optionally, if the first computing participant 810 indicates that the first ID is not an intersection ID based on the first ciphertext data obtained by it, the first computing participant 810 is also used to delete the share of the first eigenvalue and the second data held by it; if the second computing participant 820 indicates that the first ID is not an intersection ID based on the first ciphertext data obtained by it, the second computing participant 820 is also used to delete the share of the first eigenvalue and the share of the second eigenvalue held by it.

[0127] Optionally, the first computing participant 810 and the second computing participant 820 are further used to determine whether the number of intersection IDs is greater than a preset threshold based on multiple ciphertext data corresponding to multiple hash buckets, and the multiple ciphertext data include the first ciphertext data; the second computing participant 820 is specifically used to obtain the share of the first eigenvalue corresponding to the first ID from the first computing participant when the number of intersection IDs is greater than the preset threshold; the first computing participant 810 is specifically used to obtain the second data when the number of intersection IDs is greater than the preset threshold.

[0128] Optionally, the first computing participant 810 and the second computing participant 820 are further configured to randomly disrupt the order of the multiple ciphertext data respectively obtained.

[0129] Optionally, the first computing participant 810 and the second computing participant 820 are specifically configured to respectively obtain the first ciphertext data through a programmable oblivious pseudo-random function OPPRF or a key exchange DH.

[0130] Optionally, the first computing participant 810 is specifically configured to obtain the second data through a programmable oblivious pseudo-random function OPPRF.

[0131] The system 800 herein can be embodied in the form of functional modules. The term "module" herein can be implemented in the form of software and / or hardware, and is not specifically limited thereto.

[0132] For example, the "first computing participant" can be a software program, hardware circuit, or a combination of the two that implements the aforementioned functions. For example, the implementation of the first computing participant will be described below using the first computing participant as an example. Similarly, the implementation of other modules, such as the second computing participant, can refer to the implementation of the first computing participant.

[0133] As an example of a software functional unit, the first computing participant may include code running on a computing instance. The computing instance may include at least one of a physical host (computing device), a virtual machine, and a container. Furthermore, the computing instance may be one or more. For example, the first computing participant may include code running on multiple hosts / virtual machines / containers. It should be noted that the multiple hosts / virtual machines / containers used to run the code may be distributed in the same region or in different regions. Furthermore, the multiple hosts / virtual machines / containers used to run the code may be distributed in the same availability zone (AZ) or in different AZs, each AZ including one data center or multiple geographically close data centers. Typically, a region may include multiple AZs.

[0134] Similarly, multiple hosts / virtual machines / containers running the code can be distributed within the same virtual private cloud (VPC) or across multiple VPCs. Typically, a VPC is set up within a region. Cross-region communication between two VPCs within the same region, or between VPCs in different regions, requires a communication gateway within each VPC to interconnect the VPCs.

[0135] As an example of a hardware functional unit, the first computing participant may include at least one computing device, such as a server. Alternatively, the first computing participant may be a device implemented using an application-specific integrated circuit (ASIC) or a programmable logic device (PLD). The PLD may be a complex programmable logical device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0136] The multiple computing devices included in the first computing participant can be distributed in the same region or in different regions. The multiple computing devices included in the first computing participant can be distributed in the same AZ or in different AZs. Similarly, the multiple computing devices included in the first computing participant can be distributed in the same VPC or in multiple VPCs. The multiple computing devices can be any combination of computing devices such as servers, ASICs, PLDs, CPLDs, FPGAs, and GALs.

[0137] Therefore, the modules of each example described in the embodiments of this application can be implemented with electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0138] In addition, the system embodiment and method embodiment provided in the above embodiments belong to the same concept, and their specific implementation process is detailed in the method embodiment above, which will not be repeated here.

[0139] The method provided in the embodiment of the present application can be performed by a computing device, which can also be referred to as a computer system. It includes a hardware layer, an operating system layer running on the hardware layer, and an application layer running on the operating system layer. The hardware layer includes hardware such as a processing unit, a memory and a memory control unit, and the function and structure of the hardware are subsequently described in detail. The operating system is any one or more computer operating systems that implement business processing through a process, for example, a Linux operating system, a Unix operating system, an Android operating system, an iOS operating system or a Windows operating system. The application layer includes applications such as a browser, an address book, a word processing software, and an instant messaging software. Furthermore, optionally, the computer system is a handheld device such as a smart phone, or a terminal device such as a personal computer, and this application is not particularly limited, as long as the method provided in the embodiment of the present application can be used. The execution subject of the method provided in the embodiment of the present application can be a computing device, or a functional module in a computing device that can call a program and execute a program.

[0140] 9 , a computing device provided in an embodiment of the present application is described in detail below.

[0141] FIG9 is a schematic diagram of the architecture of a computing device 1500 provided in an embodiment of the present application. The computing device 1500 may be a server, a computer, or other device with computing capabilities. The computing device 1500 shown in FIG9 includes: at least one processor 1510 and a memory 1520.

[0142] It should be understood that this application does not limit the number of processors and memories in the computing device 1500.

[0143] The processor 1510 executes the instructions in the memory 1520 so that the computing device 1500 implements the method provided in this application. Alternatively, the processor 1510 executes the instructions in the memory 1520 so that the computing device 1500 implements the functional modules provided in this application, thereby implementing the method provided in this application.

[0144] Optionally, the computing device 1500 further includes a communication interface 1530. The communication interface 1530 uses a transceiver module such as, but not limited to, a network interface card or a transceiver to implement communication between the computing device 1500 and other devices or a communication network.

[0145] Optionally, the computing device 1500 further includes a system bus 1540, wherein the processor 1510, the memory 1520, and the communication interface 1530 are respectively connected to the system bus 1540. The processor 1510 can access the memory 1520 through the system bus 1540. For example, the processor 1510 can read and write data or execute code in the memory 1520 through the system bus 1540. The system bus 1540 is a peripheral component interconnect express (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The system bus 1540 is divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in Figure 9, but this does not mean that there is only one bus or one type of bus.

[0146] In one possible implementation, the processor 1510 is primarily responsible for interpreting computer program instructions (or codes) and processing data in the computer software. The computer program instructions and the data in the computer software may be stored in the memory 1520 or the cache 1516.

[0147] Optionally, the processor 1510 may be an integrated circuit chip having signal processing capabilities. By way of example and not limitation, the processor 1510 is a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component. The general-purpose processor is a microprocessor, etc. For example, the processor 1510 is a central processing unit (CPU).

[0148] Optionally, each processor 1510 includes at least one processing unit 1512 and a memory control unit 1514 .

[0149] Processing unit 1512, also known as a core, is the most important component of a processor. Processing unit 1512 is manufactured from single-crystal silicon using a specific production process. All processor calculations, command reception, command storage, and data processing are performed by the core. Each processing unit independently executes program instructions, leveraging parallel computing capabilities to accelerate program execution. Each processing unit has a fixed logical structure. For example, a processing unit includes logical units such as a level 1 cache, a level 2 cache, an execution unit, an instruction-level unit, and a bus interface.

[0150] In one implementation example, the memory control unit 1514 is configured to control data exchange between the memory 1520 and the processing unit 1512. Specifically, the memory control unit 1514 receives memory access requests from the processing unit 1512 and controls access to the memory based on the memory access requests. By way of example and not limitation, the memory control unit is a device such as a memory management unit (MMU).

[0151] In one implementation example, each memory control unit 1514 addresses the memory 1520 via the system bus. An arbiter (not shown in FIG. 9 ) is configured in the system bus to handle and coordinate competing accesses by multiple processing units 1512 .

[0152] In an implementation example, the processing unit 1512 and the memory control unit 1514 are connected to each other via connection lines inside the chip, such as address lines, so as to achieve communication between the processing unit 1512 and the memory control unit 1514.

[0153] Optionally, each processor 1510 also includes a cache 1516, which is a buffer for data exchange (called a cache). When a processing unit 1512 needs to read data, it first searches the cache for the required data. If the data is found, it executes the request directly; if not, it searches the memory. Because the cache runs much faster than the memory, the cache helps the processing unit 1512 run faster.

[0154] Memory 1520 can provide runtime space for processes in computing device 1500. For example, memory 1520 stores computer programs (specifically, program code) used to generate processes. After the computer program is executed by the processor to generate a process, the processor allocates corresponding storage space for the process in memory 1520. Furthermore, the aforementioned storage space further includes a text segment, an initialized data segment, a bit-initialized data segment, a stack segment, a heap segment, and the like. Memory 1520 stores data generated during the execution of the process, such as intermediate data or process data, in the storage space corresponding to the aforementioned process.

[0155] Optionally, the memory is also called the internal memory. Its function is to temporarily store the data processed by the processor 1510 and the data exchanged with external storage such as a hard disk. As long as the computer is running, the processor 1510 will load the data to be calculated into the internal memory for calculation. When the calculation is completed, the processing unit 1512 will transmit the result.

[0156] By way of example and not limitation, memory 1520 is a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Non-volatile memory is read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory is random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus RAM (DR RAM). It should be noted that the memory 1520 of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0157] The structure of the computing device 1500 listed above is only for illustrative purposes, and the present application is not limited thereto. The computing device 1500 of the embodiment of the present application includes various hardware in the computer system in the prior art. For example, the computing device 1500 also includes other memories in addition to the memory 1520, such as disk storage, etc. Those skilled in the art should understand that the computing device 1500 may also include other devices necessary to achieve normal operation. At the same time, according to specific needs, those skilled in the art should understand that the above-mentioned computing device 1500 may also include hardware devices that implement other additional functions. In addition, those skilled in the art should understand that the above-mentioned computing device 1500 may also include only the devices necessary to implement the embodiment of the present application, and does not necessarily include all the devices shown in Figure 9.

[0158] The present application also provides a computing device cluster. The computing device cluster includes at least one computing device. The computing device may be a server. In some embodiments, the computing device may also be a terminal device such as a desktop computer, a laptop computer, or a smartphone.

[0159] As shown in Figure 10, the computing device cluster includes at least one computing device 1500. The memory 1520 in one or more computing devices 1500 in the computing device cluster may store the same instructions for executing the above method.

[0160] In some possible implementations, the memory 1520 of one or more computing devices 1500 in the computing device cluster may also store some instructions for executing the above method. In other words, the combination of one or more computing devices 1500 can jointly execute the instructions of the above method.

[0161] It should be noted that the memory 1520 in different computing devices 1500 in the computing device cluster can store different instructions, each for executing part of the functions of the above-mentioned computing participants. In other words, the instructions stored in the memory 1520 in different computing devices 1500 can implement one or more functions of the above-mentioned computing participants.

[0162] In some possible implementations, one or more computing devices in a computing device cluster can be connected via a network. The network can be a wide area network (WAN) or a local area network (LAN), among others. FIG. 10 illustrates one possible implementation. As shown in FIG. 10 , two computing devices 1500A and 1500B are connected via a network. Specifically, each computing device is connected to the network via a communication interface within the computing device.

[0163] It should be understood that the functionality of the computing device 1500A shown in FIG10 may also be implemented by multiple computing devices 1500. Similarly, the functionality of the computing device 1500B may also be implemented by multiple computing devices 1500.

[0164] This embodiment also provides a computer program product including instructions. The computer program product may be software or a program product including instructions that can be run on a computing device or stored in any available medium. When the computer program product is run on a computing device, it causes the computing device to perform the method provided above, or causes the computing device to implement the functions of the system provided above.

[0165] This embodiment also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium capable of storing data on a computing device, or a data storage device such as a data center that contains one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, or a magnetic tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive). The computer-readable storage medium includes instructions that, when executed on a computing device, cause the computing device to perform the method provided above.

[0166] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0167] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0168] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0169] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0170] The units described as separate components may or may not be physically separate, and 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 these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0171] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0172] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0173] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.< / ys> < / ys> < / y> < / y> < / fs> < / fs> < / fs> < / fs> < / fs> < / fs> < / y> < / fs> < / y> < / y> < / fs> < / fs> < / fs> < / fs> < / fs> < / fs> < / fs> < / y> < / y> < / y> < / y02> < / y01> < / y> < / y> < / y> < / y> < / y> < / y> < / y> < / y> < / fs> < / y>

Claims

1. A method for data processing, characterized in that: The method comprises: The first computing participant and the second computing participant respectively obtain first ciphertext data, where the first ciphertext data is used to indicate whether the first ID of the first computing participant is an intersection ID, where the intersection ID indicates that the first ID is the same as any one of the at least one ID of the second computing participant; The second computing participant obtains a share of the first characteristic value corresponding to the first ID from the first computing participant, where the share of the first characteristic value is the first computing participant sharing the first characteristic value between the first computing participant and the second computing participant in a secret sharing manner, and the shares of the first characteristic value obtained by the first computing participant and the second computing participant respectively can restore the first characteristic value; The first computing participant obtains second data, wherein if the first ID is an intersection ID, the second data is a share of the second characteristic value corresponding to the first ID in the second computing participant, and if the first ID is not an intersection ID, the second data is a random number, and the share of the second characteristic value is the second computing participant sharing the second characteristic value between the first computing participant and the second computing participant in a secret sharing manner, and the shares of the second characteristic value obtained by the first computing participant and the second computing participant respectively can restore the second characteristic value; If the first ciphertext data indicates that the first ID is an intersection ID, the first computing participant uses the share of the first characteristic value held by it and the second data as training data for the neural network; If the first ciphertext data indicates that the first ID is an intersection ID, the second computing participant uses its share of the first eigenvalue and its share of the second eigenvalue as training data for the neural network.

2. The method according to claim 1, characterized in that The method further comprises: The first computing participant and the second computing participant respectively recover the first ciphertext data to obtain first data; Determine whether the first ID is the intersection ID according to the first data.

3. The method according to claim 1 or 2, characterized in that: The method further comprises: If the first computing participant indicates that the first ID is not an intersection ID according to the first ciphertext data obtained by the first computing participant, the first computing participant deletes the share of the first characteristic value and the second data held by the first computing participant; If the second computing participant indicates, based on the first ciphertext data obtained by it, that the first ID is not an intersection ID, the second computing participant deletes its share of the first eigenvalue and its share of the second eigenvalue.

4. The method according to any one of claims 1 to 3, characterized in that The method further comprises: The first computing participant and the second computing participant determine whether the number of intersection IDs is greater than a preset threshold according to multiple ciphertext data corresponding to multiple hash buckets, the multiple ciphertext data including the first ciphertext data; The second computing participant obtains a share of the first characteristic value corresponding to the first ID from the first computing participant, including: When the number of the intersection IDs is greater than the preset threshold, the second computing participant obtains a share of the first characteristic value corresponding to the first ID from the first computing participant; The first computing participant obtains the second data, including: The first calculation participant obtains the second data when the number of the intersection IDs is greater than the preset threshold.

5. The method according to claim 4, characterized in that The method further comprises: The first computing participant and the second computing participant respectively randomly scramble the order of the multiple ciphertext data they have obtained.

6. The method according to any one of claims 1 to 5, characterized in that The first computing participant and the second computing participant respectively obtain first ciphertext data, including: The first computing participant and the second computing participant respectively obtain the first ciphertext data through a programmable oblivious pseudo-random function OPPRF or a key exchange DH.

7. The method according to any one of claims 1 to 6, characterized in that The first computing participant obtains the second data, including: The first computing participant obtains the second data through a programmable oblivious pseudo-random function OPPRF.

8. The method according to any one of claims 1 to 7, characterized in that The first ID of the first computing participant is the ID of the first computing participant in the first hash bucket, the at least one ID of the second computing participant is the at least one ID of the second computing participant in the first hash bucket, the first ciphertext data is the ciphertext data corresponding to the first hash bucket, and the second data is the second data corresponding to the first hash bucket.

9. A data processing system, characterized in that: The system comprises: The first computing participant and the second computing participant are respectively used to obtain first ciphertext data, where the first ciphertext data is used to indicate whether the first ID of the first computing participant is an intersection ID, where the intersection ID indicates that the first ID is the same as any one of the at least one ID of the second computing participant; The second computing participant is further configured to obtain a share of the first characteristic value corresponding to the first ID from the first computing participant, where the share of the first characteristic value is the first computing participant sharing the first characteristic value between the first computing participant and the second computing participant in a secret sharing manner, and the shares of the first characteristic value obtained by the first computing participant and the second computing participant respectively can restore the first characteristic value; The first computing participant is further used to obtain second data, wherein if the first ID is an intersection ID, the second data is a share of the second characteristic value corresponding to the first ID in the second computing participant, and if the first ID is not an intersection ID, the second data is a random number, and the share of the second characteristic value is the second computing participant sharing the second characteristic value between the first computing participant and the second computing participant in a secret sharing manner, and the shares of the second characteristic value obtained by the first computing participant and the second computing participant respectively can restore the second characteristic value; If the first ciphertext data indicates that the first ID is an intersection ID, the first computing participant is further configured to use the share of the first characteristic value held by it and the second data as training data for the neural network; If the first ciphertext data indicates that the first ID is an intersection ID, the second computing participant is further used to use its share of the first eigenvalue and its share of the second eigenvalue as training data for the neural network.

10. The system according to claim 9, characterized in that The first computing participant and the second computing participant are further used to respectively recover the first ciphertext data to obtain first data; The first computing participant and the second computing participant are also respectively used to determine whether the first ID is the intersection ID based on the first data.

11. The system according to claim 9 or 10, characterized in that: If the first computing participant indicates, according to the first ciphertext data obtained by the first computing participant, that the first ID is not an intersection ID, the first computing participant is further configured to delete the share of the first characteristic value and the second data held by the first computing participant; If the second computing participant indicates, based on the first ciphertext data obtained by it, that the first ID is not an intersection ID, the second computing participant is further used to delete the share of the first eigenvalue and the share of the second eigenvalue held by it.

12. The system according to any one of claims 9 to 11, characterized in that The first computing participant and the second computing participant are further configured to determine whether the number of intersection IDs is greater than a preset threshold value based on a plurality of ciphertext data, wherein the plurality of ciphertext data includes the first ciphertext data; The second calculation participant is specifically configured to obtain, from the first calculation participant, a share of the first characteristic value corresponding to the first ID when the number of the intersection IDs is greater than the preset threshold; The first calculation participant is specifically used to obtain the second data when the number of the intersection IDs is greater than the preset threshold.

13. The system according to claim 12, characterized in that The first computing participant and the second computing participant are further used to randomly scramble the order of the multiple ciphertext data respectively obtained.

14. The system according to any one of claims 9 to 13, characterized in that The first computing participant and the second computing participant are specifically used to obtain the first ciphertext data respectively through a programmable oblivious pseudo-random function OPPRF or a key exchange DH.

15. The system according to any one of claims 9 to 14, characterized in that The first computing participant is specifically used for: The second data is obtained by a programmable oblivious pseudo-random function OPPRF.

16. The system according to any one of claims 9 to 15, characterized in that The first ID of the first computing participant is the ID of the first computing participant in the first hash bucket, the at least one ID of the second computing participant is the at least one ID of the second computing participant in the first hash bucket, the first ciphertext data is the ciphertext data corresponding to the first hash bucket, and the second data is the second data corresponding to the first hash bucket.

17. A computing device cluster, characterized in that: comprising at least one computing device, each computing device comprising a processor and a memory; The processor of the at least one computing device is used to execute instructions stored in the memory of the at least one computing device so that The computing device cluster executes the method as claimed in any one of claims 1 to 8.

18. A computer program product comprising instructions, characterized in that When the instructions are executed by a computing device cluster, the computing device cluster is caused to perform the method according to any one of claims 1 to 8.

19. A computer-readable storage medium, characterized in that: The method comprises computer program instructions, and when the computer program instructions are executed by a computing device cluster, the computing device cluster performs the method according to any one of claims 1 to 8.

Citation Information

Patent Citations

  • Private data hiding and sharing method

    CN114661680A

  • Data intersection method and system based on privacy protection and related equipment

    CN115913537A

  • Data processing method and system and computing device

    CN117294425A

  • A privacy preserving data storing method and a privacy preserving data storing system for analyzing a travel behavior of one or more users of mobility-as-a-service (MAAS) transportation services

    US20230091700A1

  • Sample alignment method and apparatus in federated learning, and device and storage medium

    WO2023050687A1

Cited By

  • Privacy query method and system based on unbalanced privacy set intersection

    CN119293298A

  • Multi-party privacy set intersection method, device, equipment and medium

    CN119519968A

  • Financial data privacy protection method and system based on secure multi-party computing

    CN120524511A

  • Privacy information intersection method and computing device

    CN120729627A