WO2021249500A1 - Method and apparatus for clustering private data of multiple parties - Google Patents

Abstract

Embodiments of the present description provide a method and apparatus for clustering private data of multiple parties. The method comprises: a first party determines first center data currently corresponding to each class cluster in a first class cluster set, a second party having second center data currently corresponding to each class cluster in a second class cluster set, and the first class cluster set and the second class cluster set forming a total class cluster set; calculates a first plaintext distance between first private data and each first center data; on the basis of a homomorphic encryption method, calculates, together with the second party, a first ciphertext distance between the first private data and each second center data; decrypts the first ciphertext distances to obtain second plaintext distances between the first private data and the second center data; and according to the first plaintext distances and the second plaintext distances, selects a class cluster corresponding to the shortest plaintext distance as the class cluster to which the first private data currently belongs. Leakage of private data can be prevented.

Description

Method and device for clustering private data of multiple parties Technical field

One or more embodiments of this specification relate to the computer field, and in particular to a method and device for clustering private data from multiple parties.

Background technique

Clustering is a very common technique in machine learning. It is often applied to tasks such as community discovery and anomaly detection. The usual clustering algorithm is an unsupervised learning algorithm whose purpose is to group similar objects into the same cluster. The more similar the objects in the cluster, the better the clustering effect. The biggest difference between clustering and classification is that the target of classification is known in advance, while clustering is different. The result is the same as the classification, but the classification is not pre-defined.

In some scenarios, the data is distributed horizontally across multiple parties. The data possessed by the parties may be private data. In other words, the private data possessed by one party cannot be disclosed to other parties. In this case, it is hoped that there will be an improved solution that can prevent the leakage of private data when clustering private data from multiple parties.

Summary of the invention

One or more embodiments of this specification describe a method and device for clustering private data of multiple parties, which can prevent the leakage of private data when clustering private data of multiple parties.

In a first aspect, a method for clustering private data of multiple parties is provided, the multiple parties including a first party and a second party, the first party having a first set of private data, and the first set of private data The method includes multiple first private data, which is executed by the first party, and includes multiple rounds of iterative processes, where any round of iteration includes: determining the first center data corresponding to each cluster in the first type cluster set The second party has the second center data corresponding to each cluster in the second type cluster set; the first type cluster set and the second type cluster set constitute a total cluster set; calculate the first type cluster set 1. The first plaintext distance between private data and each first center data; based on homomorphic encryption, the first secret between the first private data and each second center data is jointly calculated with the second party Text distance; decrypt the first ciphertext distance to obtain the second plaintext distance between the first private data and the second central data; according to each first plaintext distance and each second plaintext distance, select The cluster corresponding to the shortest plaintext distance is used as the cluster to which the first private data currently belongs.

In a possible implementation manner, the method based on homomorphic encryption and the second party jointly calculate the first ciphertext distance between the first private data and each second central data, including: The first private data is encrypted with the public key of the first party to obtain the first encrypted data; the first encrypted data is sent to the second party, so that the second party can homomorphically calculate the The distance between the first encrypted data and the second central data to obtain the first ciphertext distance; receiving the first ciphertext distance from the second party.

Further, the decrypting the first ciphertext distance includes: using the private key of the first party to decrypt the first ciphertext distance to obtain the second plaintext distance; The private key and the public key of the first party form a set of public-private key pairs.

In a possible implementation manner, the arbitrary round of iteration is the first iteration, and the first center data corresponding to each cluster in the first type cluster set is randomly initialized data.

In a possible implementation manner, the second party has a second private data set, and the second private data set includes a plurality of second private data; and the cluster corresponding to the shortest plaintext distance is selected as the After the class cluster to which the first private data currently belongs, the method further includes: for the first private data set, determining a first average value of each first private data belonging to the first class cluster; the first class cluster Is any type of cluster in the first type cluster set; receiving from the second party the second average value of each second private data belonging to the first type cluster determined by the second party; The first average value and the second average value are updated, and the first center data corresponding to the first type cluster is updated.

Further, after the update of the first center data corresponding to the first type cluster, the method further includes: jointly determining with the second party that changes in the center data of each type of cluster in the general type cluster set Whether the amount meets the preset iterative stop condition; if the result of the judgment is that the change in the central data of the various clusters in the total cluster set does not meet the preset iterative stop condition, then the multi-round iteration process is performed The next iteration.

Further, the method further includes: if the result of the judgment is that the variation of the central data of the various clusters in the total cluster set meets a preset condition for stopping iteration, then assigning the first private data to the category to which the first private data currently belongs The cluster is determined as the class cluster to which the first private data ultimately belongs.

Further, the joint judgment includes: a local judgment, whether the amount of change in the center data of the various clusters in the first type cluster set meets a preset iterative stop condition, to obtain a first judgment result; The two parties receive a second judgment result, the second judgment result being used to indicate whether the amount of change in the center data of each type of cluster in the second type cluster set satisfies a preset condition for stopping iteration; according to the first judgment A comprehensive judgment is made between the result and the second judgment result, and whether the change amount of the center data of the various clusters in the general cluster set meets the preset iterative stop condition.

In a possible implementation manner, the second party has a second private data set, and the second private data set includes a plurality of second private data; the method further includes a method based on homomorphic encryption, Calculate the second ciphertext distance between the second privacy data and each first central data jointly with the second party, so that the second party obtains the first ciphertext distance after decryption according to the second ciphertext distance Second, the second plaintext distance between the private data and the first central data.

Further, the method based on homomorphic encryption and the second party to jointly calculate the second ciphertext distance between the second private data and each first central data includes: receiving from the second party Second encrypted data; the second encrypted data is obtained by the second party encrypting the second private data with the public key of the second party; homomorphic calculation of the second encrypted data and the first A distance between central data to obtain the second ciphertext distance; sending the second ciphertext distance to the second party.

In a second aspect, there is provided an apparatus for clustering private data of multiple parties, the multiple parties including a first party and a second party, the first party having a first set of private data, and the first set of private data The device includes a plurality of first privacy data, and the device is set at the first party and is used to perform multiple rounds of iterative processes, including the following units for performing any round of iteration: a center determining unit, used to determine the first type of cluster Each cluster in the set currently corresponds to the first center data; the second party has the second center data currently corresponding to each cluster in the second cluster set; the first cluster set and the second cluster set The cluster set constitutes a total cluster set; an independent calculation unit is used to calculate the first plaintext distance between the first privacy data and each first center data determined by the center determination unit; the first joint calculation unit uses Based on the homomorphic encryption method, the first ciphertext distance between the first private data and each second central data is jointly calculated with the second party; the decryption unit is used for the first joint calculation unit The obtained first ciphertext distance is decrypted to obtain the second plaintext distance between the first private data and the second central data; the cluster determining unit is used for each first plaintext obtained by the independent computing unit The distance, and each second plaintext distance obtained by the decryption unit, select the cluster corresponding to the shortest plaintext distance as the cluster to which the first private data currently belongs.

In a third aspect, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed in a computer, the computer is caused to execute the method of the first aspect.

In a fourth aspect, a computing device is provided, including a memory and a processor, the memory stores executable code, and the processor implements the method of the first aspect when the executable code is executed by the processor.

With the method and device provided in the embodiments of this specification, not one party alone determines the center data of each cluster, but the first party determines the first center data corresponding to each cluster in the first type cluster set. The two parties determine the second central data corresponding to each cluster in the second type cluster set, the first type cluster set and the second type cluster set constitute a general cluster set; and the first party subsequently calculates the first private data The first plaintext distance between each first center data determined by the party; based on homomorphic encryption, the first private data and each second center data determined by the second party are jointly calculated with the second party The first ciphertext distance between, and decrypt the first ciphertext distance to obtain the second plaintext distance between the first private data and the second central data; finally, according to each first plaintext distance, and For each second plaintext distance, the cluster corresponding to the shortest plaintext distance is selected as the cluster to which the first private data currently belongs. The whole process is based on homomorphic encryption, which can prevent the leakage of private data when clustering private data from multiple parties.

Description of the drawings

In order to explain the technical solutions of the embodiments of the present invention more clearly, the following will briefly introduce the drawings used in the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative work, other drawings can be obtained from these drawings.

Figure 1 is a schematic diagram of an implementation scenario of an embodiment disclosed in this specification;

Fig. 2 shows a flowchart of a method for clustering private data of multiple parties according to an embodiment;

Fig. 3 shows a schematic block diagram of an apparatus for clustering private data of multiple parties according to an embodiment.

detailed description

The following describes the solutions provided in this specification with reference to the accompanying drawings.

Figure 1 is a schematic diagram of an implementation scenario of an embodiment disclosed in this specification. This implementation scenario involves clustering private data from multiple parties. It is understandable that the above-mentioned multiple parties may be two parties or more than two parties, for example, three parties, four parties, and so on. In the embodiments of this specification, clustering of private data of two parties is taken as an example for description. 1, the first party 11 has privacy data 1, privacy data 2, privacy data 3, privacy data 4, and privacy data 5; the second party 12 has privacy data 6, privacy data 7, privacy data 8, and privacy data 9. Among them, the first party and the second party are only a distinction between the two parties. The first party may also be referred to as party A, the second party may be referred to as party B, and so on.

In the embodiments of this specification, the information covered by the private data is not limited, and it can be any information that cannot be communicated, for example, the user's personal information or business secrets. For example, the private data is the user's personal information, including the user's name, age, income, etc., for details, please refer to the correspondence table of the information contained in each private data shown in Table 1.

Table 1: Correspondence table of information contained in each private data

To	Name	age)	Income (ten thousand yuan)
Privacy data 1	Zhang Yi	25	1.5
Privacy data 2	Zhang Er	26	2.2
Privacy data 3	Zhang San	35	0.8
Privacy data 4	Zhao Yi	41	1.8
Privacy data 5	Zhao Er	19	0.6
Privacy data 6	Zhao San	28	3.5
Privacy data 7	Zhao Si	36	1.2
Privacy data 8	Yi Wang	29	1.3
Privacy data 9	King Two	30	2.2

It can be seen from Table 1 that the data of different rows in Table 1 may be distributed in different parties. For example, private data 1 is distributed in the first party, and private data 8 is distributed in the second party. This kind of data distribution is distributed among multiple parties. It can be called horizontal segmentation.

In the embodiment of this specification, it is necessary to cluster private data of multiple parties. Taking Figure 1 as an example, it is for private data 1, private data 2, private data 3, private data 4, private data 5, private data 6, and private data 7. ,Privacy data8,Privacy data9 are clustered. Private data distributed in different parties may be classified into the same cluster. For example, private data 1, private data 3, private data 6, and private data 7 are classified into categories Cluster 1, private data 2, private data 4, private data 5, private data 8, and private data 9 are divided into cluster 2. In the embodiment of this specification, the method of homomorphic encryption is used to realize the clustering of the private data of multiple parties under the premise of not leaking the private data.

FIG. 2 shows a flowchart of a method for clustering private data of multiple parties according to an embodiment, and the method may be based on the implementation scenario shown in FIG. 1. The multiple parties include a first party and a second party, the first party has a first private data set, the first private data set includes a plurality of first private data, and the method is executed by the first party , Including multiple rounds of iterative process, as shown in Figure 2, where any round of iteration includes the following steps: step 21, determine the first center data corresponding to each cluster in the first cluster set; the second party has The second center data corresponding to each cluster in the second type cluster set; the first type cluster set and the second type cluster set constitute a total cluster set; step 22, calculate the first privacy data and The first plaintext distance between each first center data; step 23, based on homomorphic encryption, jointly calculate the first ciphertext between the first private data and each second center data with the second party Distance; Step 24, decrypt the first ciphertext distance to obtain the second plaintext distance between the first private data and the second central data; step 25, according to each first plaintext distance, and each first Two plaintext distance, select the cluster corresponding to the shortest plaintext distance as the cluster to which the first private data currently belongs. The following describes the specific implementation of each of the above steps.

First, in step 21, determine the first center data corresponding to each cluster in the first type cluster set; the second party has the second center data currently corresponding to each cluster in the second type cluster set; The first-type cluster set and the second-type cluster set constitute a general-type cluster set. It is understandable that the first number of clusters included in the first type cluster set can be preset, or the second number of clusters included in the second type cluster set can be preset, or Determine the third number of each cluster included in the above-mentioned total cluster set. For example, preset the first number to 3, the second number to 3, and the third number to 6; or, to preset the first number to 1, the second number to 2, and the third number to 3; or, to preset Let the first number be 2, the second number be 5, and the third number be 7.

In the embodiment of this specification, each center data in the total cluster set is jointly determined by the first party and the second party, and the first party can only determine the first center data corresponding to each cluster in the first cluster set. , The second party determines the corresponding second center data of each cluster in the second type cluster set. Either the first party and the second party only have part of the center data in the total cluster set, and not It has all the central data in the total cluster set.

It should be noted that, on the premise that the third number has been determined, the first number and the second number may be equal or unequal, and the first and second numbers may be adjusted according to the distribution of the amount of data in the first and second parties. Number and second number to speed up the clustering process.

In an example, the arbitrary round of iteration is the first iteration, and the first center data corresponding to each cluster in the first type cluster set is randomly initialized data.

For example, assuming that the third number is 6, the first number is 3, and the second number is 3, the first party randomly initializes 3 first center data, denoted as (c_1, c_2, c_3); Ground, the second party randomly initializes 3 second center data, denoted as (c_4, c_5, c_6).

Further, the first party may initialize a K-dimensional cluster vector for each first private data to mark the cluster to which the first private data belongs, where K is the number of clusters, when K=6 When, initialize a 6-dimensional cluster vector, for example, the initial vector is all 0, that is, [0,0,0,0,0,0].

Then in step 22, the first plaintext distance between the first privacy data and each first central data is calculated. It is understandable that since the first party has both the first private data and the first central data, the above-mentioned first plaintext distance can be calculated independently.

In the embodiment of this specification, it is assumed that x1 represents the first privacy data, c_1, c_2, and c_3 represent the respective first central data, and the first plaintext distance between x1 and c_1 can be obtained by calculating (c_1-x1)^2, denoted Is dx1; the first plaintext distance between x1 and c_2 can be obtained by calculating (c_2-x1)^2, denoted as dx2; the first plaintext distance between x1 and c_3 can be obtained by calculating (c_3-x1)^2 , Denoted as dx3.

It can be understood that, generally, the first privacy data and the first center data are both in the form of vectors, and the process of calculating the first plaintext distance is a vector operation process.

Next, in step 23, based on the homomorphic encryption method, the first ciphertext distance between the first private data and each second central data is jointly calculated with the second party. It is understandable that the first party has first private data, and the second party has second central data. Therefore, the first party needs to jointly calculate the first ciphertext distance with the second party.

In an example, the method based on homomorphic encryption and the second party jointly calculate the first ciphertext distance between the first private data and each second central data, including: The private data is encrypted with the public key of the first party to obtain the first encrypted data; the first encrypted data is sent to the second party, so that the second party homomorphically calculates the first encrypted data The distance from the second central data to obtain the first ciphertext distance; receiving the first ciphertext distance from the second party.

Assuming that [] is used for encryption and []a is used for encryption with the public key of the first party, the first party can encrypt the first private data with Pka and give it to the second party. Assuming that the first privacy data is represented by x1, the first encrypted data can be represented as [x1]a.

The second party homomorphically calculates the distance between the first encrypted data [x1]a and each of the second central data c_4, c_5, c_6, where [x1]a and c_4, c_5, c_6 are all vectors, and the distance calculation can be through The vector elements are calculated, and the ciphertext of the distance can be directly homomorphically calculated. The ciphertext is the aforementioned first ciphertext distance, which is recorded as [dx4]a,[dx5]a,[dx6]a, and the ciphertext is returned To the first party.

In step 24, the first ciphertext distance is decrypted to obtain the second plaintext distance between the first private data and the second central data. It is understandable that in order to facilitate the comparison of the distances calculated in different ways, the first ciphertext distance needs to be decrypted.

In an example, the decrypting the first ciphertext distance includes: using the private key of the first party to decrypt the first ciphertext distance to obtain the second plaintext distance; The private key of the party and the public key of the first party form a set of public-private key pairs.

For example, the first party (also referred to as party A) has a public-private key pair (Pka, Ska), where Pka represents the public key of the first party, and Ska represents the private key of the first party.

The first party decrypts each first ciphertext distance [dx4]a, [dx5]a, [dx6]a, and obtains each second plaintext distance dx4, dx5, dx6.

Finally, in step 25, according to each first plaintext distance and each second plaintext distance, the cluster corresponding to the shortest plaintext distance is selected as the cluster to which the first private data currently belongs. It is understandable that in different rounds of iterative processes, the clusters to which the first private data belongs may be different.

For example, each first plaintext distance is dx1, dx2, dx3, and each second plaintext distance is dx4, dx5, dx6, where the shortest plaintext distance is dx2, that is, the first private data x1 is away from the first center The data c_2 is the most recent, and the cluster to which the first private data x1 currently belongs is the cluster corresponding to c_2, and the cluster vector corresponding to x1 is updated to [0,1,0,0,0,0].

In an example, the second party has a second private data set, and the second private data set includes a plurality of second private data; the selecting the cluster corresponding to the shortest plaintext distance is the first private data After the current category cluster belongs, the method further includes: for the first private data set, determining a first average value of each first private data belonging to the first category cluster; the first category cluster is the first category cluster. Any type of cluster in a type of cluster set; receiving from the second party the second average value of each second private data belonging to the first type of cluster determined by the second party; according to the first average value And the second average value, and update the first center data corresponding to the first type cluster.

For example, the first party and the second party update the central data c_1 to c_6 according to the cluster vector of all private data. Taking c_1 as an example, the update process is as follows:

The first party calculates the mean value of all the first private data whose cluster vector is [1,0,0,0,0,0], denoted as e1;

The second party calculates the mean value of all the second private data whose cluster vector is [1,0,0,0,0,0], denote it as e2, and sends e2 to the first party;

The first party calculates the average value of e1 and e2, and the average value is the new c_1.

Further, after the update of the first center data corresponding to the first type cluster, the method further includes: jointly determining with the second party that changes in the center data of each type of cluster in the general type cluster set Whether the amount meets the preset iterative stop condition; if the result of the judgment is that the change in the central data of the various clusters in the total cluster set does not meet the preset iterative stop condition, then the multi-round iteration process is performed The next iteration.

Further, the method further includes: if the result of the judgment is that the variation of the central data of the various clusters in the total cluster set meets a preset condition for stopping iteration, then assigning the first private data to the category to which the first private data currently belongs The cluster is determined as the class cluster to which the first private data ultimately belongs.

Further, the joint judgment includes: a local judgment, whether the amount of change in the center data of the various clusters in the first type cluster set meets a preset iterative stop condition, to obtain a first judgment result; The two parties receive a second judgment result, the second judgment result being used to indicate whether the amount of change in the center data of each type of cluster in the second type cluster set satisfies a preset condition for stopping iteration; according to the first judgment A comprehensive judgment is made between the result and the second judgment result, and whether the change amount of the center data of the various clusters in the general cluster set meets the preset iterative stop condition.

For example, the above stop iteration condition is |C(t)-C(t+1)|^2<delta, where delta can be a preset value, C(t) represents the central data before update, C(t+ 1) Represents the updated central data.

In the embodiment of this specification, the processing procedure from step 21 to step 25 is mainly to describe the first party's first private data for its own party, and determine the cluster to which the first private data belongs. In addition, the first party's first private data for the second party The second private data also needs to cooperate with the second party in the homomorphic encryption method to determine the cluster to which the second private data belongs.

In an example, the second party has a second private data set, and the second private data set includes a plurality of second private data; the method further includes: based on a homomorphic encryption method, and the first private data set The two parties jointly calculate the second ciphertext distance between the second privacy data and each first central data, so that the second party obtains the second privacy data and the second ciphertext distance after decryption according to the second ciphertext distance The second plaintext distance between the first center data.

Further, the method based on homomorphic encryption and the second party to jointly calculate the second ciphertext distance between the second private data and each first central data includes: receiving from the second party Second encrypted data; the second encrypted data is obtained by the second party encrypting the second private data with the public key of the second party; homomorphic calculation of the second encrypted data and the first A distance between central data to obtain the second ciphertext distance; sending the second ciphertext distance to the second party.

For example, the second party (also called party B) may have a public-private key pair (Pkb, Skb), where Pkb represents the public key of the second party, and Skb represents the private key of the second party.

Assuming that [] is used for encryption and []b is used for encryption with the public key of the second party, the second party can encrypt the second private data with Pkb to obtain the second encrypted data, and send the second encrypted data to the first party .

It is understandable that in the method for clustering private data of multiple parties, the status of the first party and the second party are equal, and the processing procedures of the first party and the second party are not substantially different. In the embodiment of this specification, Mainly take the first party as the executive body to describe the corresponding processing process.

Through the method provided by the embodiment of this specification, neither party determines the center data of each cluster individually, but the first party determines the current corresponding first center data of each cluster in the first cluster set, and the second party Determine the second center data corresponding to each cluster in the second type of cluster set, the first type of cluster set and the second type of cluster set constitute the total type of cluster set; and the subsequent first party calculates the first private data and the original The first plaintext distance between each first center data determined by the party; based on the homomorphic encryption method, and the second party jointly calculate the distance between the first privacy data and each second center data determined by the second party The first ciphertext distance; decrypt the first ciphertext distance to obtain the second plaintext distance between the first private data and the second central data; finally, according to each first plaintext distance, and each first ciphertext distance Two plaintext distance, select the cluster corresponding to the shortest plaintext distance as the cluster to which the first private data currently belongs. The whole process is based on homomorphic encryption, which can prevent the leakage of private data when clustering private data from multiple parties.

According to another embodiment, a device for clustering private data of multiple parties is also provided, and the device is used to implement the method for clustering private data of multiple parties provided in the embodiment of this specification. The multiple parties include a first party and a second party, the first party has a first private data set, the first private data set includes a plurality of first private data, and the device is set on the first party , Used to perform multiple rounds of iterative process. Fig. 3 shows a schematic block diagram of an apparatus for clustering private data of multiple parties according to an embodiment. As shown in FIG. 3, the device 300 includes the following units for performing any round of iteration: a center determining unit 31, configured to determine the first center data corresponding to each cluster in the first type cluster set; The two parties have the second center data corresponding to each cluster in the second type cluster set; the first type cluster set and the second type cluster set constitute a total cluster set; the independent calculation unit 32 is used for calculation The first plaintext distance between the first privacy data and the respective first center data determined by the center determination unit 31; the first joint calculation unit 33 is configured to communicate with the second party based on a homomorphic encryption method Jointly calculate the first ciphertext distance between the first private data and each second central data; the decryption unit 34 is configured to decrypt the first ciphertext distance obtained by the first joint calculation unit 33 to obtain the The second plaintext distance between the first privacy data and the second central data; the cluster determining unit 35 is used to determine the first plaintext distances obtained by the independent computing unit 32 and the decryption unit 34 For each second plaintext distance, the cluster corresponding to the shortest plaintext distance is selected as the cluster to which the first private data currently belongs.

Optionally, as an embodiment, the first joint computing unit 33 includes: an encryption subunit, configured to encrypt the first private data with the public key of the first party to obtain the first encrypted data; and send; A subunit, configured to send the first encrypted data obtained by the encryption subunit to the second party, so that the second party homomorphically calculates the relationship between the first encrypted data and the second central data To obtain the first ciphertext distance; and a receiving subunit for receiving the first ciphertext distance from the second party.

Further, the decryption unit 34 is specifically configured to use the private key of the first party to decrypt the first ciphertext distance to obtain the second plaintext distance; the private key of the first party and the The public key of the first party forms a set of public-private key pairs.

Optionally, as an embodiment, the arbitrary round of iteration is the first iteration, and the center determining unit 31 is specifically configured to determine the first center corresponding to each cluster in the first cluster set. The data is randomly initialized data.

Optionally, as an embodiment, the second party has a second private data set, and the second private data set includes a plurality of second private data; the device further includes: an average value determining unit, configured to: After the cluster determining unit 35 selects the cluster corresponding to the shortest plaintext distance as the cluster to which the first private data currently belongs, for the first private data set, determine each first privacy attribute belonging to the first cluster The first average value of the data; the first type cluster is any type of cluster in the first type cluster set; the receiving unit is configured to receive from the second party that the second party determines that it belongs to the The second average value of each second privacy data of the first type cluster; an update unit, configured to update the corresponding first type cluster according to the first average value determined by the average value determining unit and the second average value received by the receiving unit The first central data.

Further, the device further includes: a joint judging unit, configured to jointly judge with the second party after the update unit updates the first center data corresponding to the first type of cluster, the set of total type clusters Whether the variation of the central data of each type of cluster in the cluster meets the preset iterative stop condition; an iteration trigger unit is used to determine if the judgment result of the joint judging unit is the central data of each type of cluster in the general cluster set If the amount of change does not meet the preset iterative stop condition, then the next iteration of the multiple iterations is performed.

Further, the device further includes: a final determination unit, configured to, if the judgment result of the joint judgment unit is that the change amount of the center data of the various clusters in the general cluster set satisfies a preset stop iteration condition, Then the cluster to which the first private data currently belongs is determined as the cluster to which the first private data ultimately belongs.

Further, the joint judging unit includes: a local judging subunit for locally judging whether the change amount of the center data of the various clusters in the first type cluster set meets the preset stop iteration condition, and obtain the first Judgment result; a receiving subunit for receiving a second judgment result from the second party, where the second judgment result is used to indicate whether the variation of the central data of each type of cluster in the second type cluster set satisfies a predetermined Set iterative stop condition; a comprehensive judgment subunit for comprehensive judgment based on the first judgment result obtained by the local judgment subunit and the second judgment result received by the receiving subunit, in the general cluster set Whether the variation of the center data of various clusters meets the preset conditions for stopping iteration.

Optionally, as an embodiment, the second party has a second private data set, and the second private data set includes a plurality of second private data; the device further includes: a second joint computing unit, In a homomorphic encryption-based manner, the second ciphertext distance between the second private data and each first central data is jointly calculated with the second party, so that the second party can be based on the second ciphertext After the text distance is decrypted, the second plain text distance between the second private data and the first central data is obtained.

Further, the second joint computing unit includes: a receiving subunit for receiving second encrypted data from the second party; the second encrypted data is used by the second party to use the second private data Obtained by encrypting the public key of the second party; a calculation subunit for homomorphically calculating the distance between the second encrypted data received by the receiving subunit and the first central data to obtain the second Ciphertext distance; a sending subunit for sending the second ciphertext distance obtained by the calculation subunit to the second party.

With the device provided by the embodiment of this specification, no one party alone determines the center data of each cluster, but the center determining unit 31 of the first party determines the first center corresponding to each cluster in the first cluster set. Data, the second party determines the second center data corresponding to each cluster in the second type cluster set, the first type cluster set and the second type cluster set constitute the general cluster set; and the subsequent independent computing unit of the first party 32 Calculate the first plaintext distance between the first private data and each first center data determined by the party; the first joint calculation unit 33 jointly calculates the second party with the second party based on homomorphic encryption A first ciphertext distance between private data and each second central data determined by the second party; the decryption unit 34 decrypts the first ciphertext distance to obtain the first private data and the second central data According to each first plaintext distance and each second plaintext distance, the final cluster determining unit 35 selects the cluster corresponding to the shortest plaintext distance as the cluster to which the first private data currently belongs. The whole process is based on homomorphic encryption, which can prevent the leakage of private data when clustering private data from multiple parties.

According to another embodiment, there is also provided a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed in a computer, the computer is caused to execute the method described in conjunction with FIG. 2.

According to an embodiment of still another aspect, there is also provided a computing device, including a memory and a processor, the memory is stored with executable code, and when the processor executes the executable code, the implementation described in conjunction with FIG. 2 method.

Those skilled in the art should be aware that, in one or more of the foregoing examples, the functions described in the present invention can be implemented by hardware, software, firmware, or any combination thereof. When implemented by software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on the computer-readable medium.

The specific embodiments described above further describe the objectives, technical solutions and beneficial effects of the present invention in further detail. It should be understood that the above descriptions are only specific embodiments of the present invention and are not intended to limit the scope of the present invention. The protection scope, any modification, equivalent replacement, improvement, etc. made on the basis of the technical solution of the present invention shall be included in the protection scope of the present invention.

Claims (22)

1. A method for clustering private data of multiple parties. The multiple parties include a first party and a second party. The first party has a first set of private data. The first set of private data includes multiple first parties. For private data, the method is executed by the first party and includes multiple rounds of iterative processes, where any round of iteration includes:

   Determine the first center data corresponding to each cluster in the first type cluster set; the second party has the second center data currently corresponding to each cluster in the second type cluster set; the first type cluster set And the second-type cluster set constitute a general-type cluster set;

   Calculating the first plaintext distance between the first privacy data and each first central data;

   Based on a homomorphic encryption method, jointly calculate the first ciphertext distance between the first private data and each second central data with the second party;

   Decrypt the first ciphertext distance to obtain the second plaintext distance between the first private data and the second central data;

   According to each first plaintext distance and each second plaintext distance, the cluster corresponding to the shortest plaintext distance is selected as the cluster to which the first private data currently belongs.
2. The method of claim 1, wherein the method based on homomorphic encryption and the second party jointly calculate the first ciphertext distance between the first private data and each second central data, comprising :

   Encrypting the first private data with the public key of the first party to obtain first encrypted data;

   Send the first encrypted data to the second party, so that the second party homomorphically calculates the distance between the first encrypted data and the second central data to obtain the first encrypted data Text distance

   Receiving the first ciphertext distance from the second party.
3. The method according to claim 2, wherein said decrypting said first ciphertext distance comprises:

   Use the private key of the first party to decrypt the first ciphertext distance to obtain the second plaintext distance; the private key of the first party and the public key of the first party form a set of public-private key pairs .
4. The method according to claim 1, wherein the arbitrary round of iteration is the first iteration, and the first center data corresponding to each cluster in the first cluster set is randomly initialized data.
5. The method of claim 1, wherein the second party has a second private data set, and the second private data set includes a plurality of second private data;

   After the selecting the cluster corresponding to the shortest plaintext distance as the cluster to which the first private data currently belongs, the method further includes:

   For the first private data set, determine the first average value of each first private data belonging to the first type of cluster; the first type of cluster is any type of cluster in the first type of cluster;

   Receiving, from the second party, the second average value of each second private data belonging to the first type cluster determined by the second party;

   According to the first average value and the second average value, the first center data corresponding to the first type cluster is updated.
6. 5. The method according to claim 5, wherein after the updating the first center data corresponding to the first type cluster, the method further comprises:

   Jointly judge with the second party whether the change amount of the center data of the various clusters in the general cluster set meets a preset condition for stopping iteration;

   If the result of the judgment is that the amount of change of the center data of each type of cluster in the total cluster set does not meet the preset iterative stop condition, then the next iteration of the multiple rounds of iterative process is performed.
7. The method of claim 6, wherein the method further comprises:

   If the result of the judgment is that the amount of change in the central data of the various clusters in the general cluster set satisfies the preset iterative stop condition, the cluster to which the first private data currently belongs is determined as the first private data The final category cluster.
8. The method according to claim 6, wherein the joint judgment comprises:

   Locally judge whether the change amount of the center data of the various clusters in the first-type cluster set meets a preset iterative stop condition, and obtain the first judgment result;

   Receiving a second judgment result from the second party, where the second judgment result is used to indicate whether the amount of change in the center data of each type of cluster in the second type cluster set meets a preset iterative stop condition;

   According to the first judgment result and the second judgment result, a comprehensive judgment is made to determine whether the change amount of the center data of the various types of clusters in the general cluster set meets a preset iterative stop condition.
9. The method of claim 1, wherein the second party has a second private data set, and the second private data set includes a plurality of second private data;

   The method also includes:

   Based on the homomorphic encryption method, the second ciphertext distance between the second private data and each first central data is jointly calculated with the second party, so that the second party can be based on the second ciphertext After the distance is decrypted, the second plaintext distance between the second private data and the first central data is obtained.
10. The method according to claim 9, wherein the method based on homomorphic encryption and the second party jointly calculate the second ciphertext distance between the second private data and each first central data, comprising :

    Receiving second encrypted data from the second party; the second encrypted data is obtained by the second party encrypting the second private data with the public key of the second party;

    Homomorphically calculating the distance between the second encrypted data and the first central data to obtain the second ciphertext distance;

    Sending the second ciphertext distance to the second party.
11. A device for clustering private data of multiple parties. The multiple parties include a first party and a second party. The first party has a first set of private data. The first set of private data includes multiple first parties. Private data, the device is set on the first party, and is used to perform multiple rounds of iterative processes, including the following units for performing any round of iteration:

    The center determining unit is used to determine the first center data corresponding to each cluster in the first type cluster set; the second party has the second center data currently corresponding to each cluster in the second type cluster set; The first-type cluster set and the second-type cluster set constitute a general-type cluster set;

    An independent calculation unit, configured to calculate the first plaintext distance between the first privacy data and each first center data determined by the center determination unit;

    The first joint calculation unit is configured to jointly calculate the first ciphertext distance between the first private data and each second central data based on the homomorphic encryption method with the second party;

    A decryption unit, configured to decrypt the first ciphertext distance obtained by the first joint computing unit to obtain the second plaintext distance between the first private data and the second central data;

    The cluster determining unit is configured to select the cluster corresponding to the shortest plaintext distance as the current first privacy data according to each first plaintext distance obtained by the independent calculation unit and each second plaintext distance obtained by the decryption unit The class cluster to which it belongs.
12. The apparatus of claim 11, wherein the first joint computing unit comprises:

    An encryption subunit, configured to encrypt the first private data with the public key of the first party to obtain the first encrypted data;

    The sending subunit is configured to send the first encrypted data obtained by the encrypting subunit to the second party, so that the second party can homomorphically calculate the difference between the first encrypted data and the second central data To obtain the first ciphertext distance;

    The receiving subunit is configured to receive the first ciphertext distance from the second party.
13. The device according to claim 12, wherein the decryption unit is specifically configured to use the private key of the first party to decrypt the first ciphertext distance to obtain the second plaintext distance; The private key of the party and the public key of the first party form a set of public-private key pairs.
14. The device according to claim 11, wherein the any round of iteration is the first iteration, and the center determining unit is specifically configured to determine the first type cluster currently corresponding to each cluster in the first type cluster set. The central data is randomly initialized data.
15. The apparatus of claim 11, wherein the second party has a second private data set, and the second private data set includes a plurality of second private data;

    The device also includes:

    The mean value determining unit is configured to determine that the first private data set belongs to the first category after the cluster determining unit selects the cluster corresponding to the shortest plaintext distance as the cluster to which the first private data currently belongs A first mean value of each first private data of the cluster; the first type of cluster is any type of cluster in the first type of cluster set;

    A receiving unit, configured to receive, from the second party, the second average value of each second private data belonging to the first type cluster determined by the second party;

    The updating unit is configured to update the first center data corresponding to the first type cluster according to the first average value determined by the average value determining unit and the second average value received by the receiving unit.
16. The device of claim 15, wherein the device further comprises:

    The joint judgment unit is configured to jointly judge with the second party after the update unit updates the first center data corresponding to the first type of cluster, the change of the center data of each type of cluster in the general type cluster set Whether the quantity meets the preset conditions for stopping iteration;

    An iteration triggering unit, configured to perform the multi-round iteration process if the judgment result of the joint judging unit is that the change amount of the center data of the various clusters in the general cluster set does not meet the preset iterative stop condition The next iteration in.
17. The device of claim 16, wherein the device further comprises:

    The final determination unit is configured to: if the judgment result of the joint judgment unit is that the variation of the central data of the various clusters in the general cluster set satisfies a preset stop iteration condition, then the first private data is currently The class cluster to which the first private data belongs is determined as the class cluster to which the first private data ultimately belongs.
18. The apparatus according to claim 16, wherein the joint judgment unit comprises:

    The local judgment subunit is used to locally judge whether the change amount of the center data of the various clusters in the first type cluster set meets the preset stop iteration condition, and obtain the first judgment result;

    The receiving subunit is configured to receive a second judgment result from the second party, where the second judgment result is used to indicate whether the amount of change in the center data of each type of cluster in the second type cluster set meets a preset Stop the iteration condition;

    The comprehensive judgment subunit is used to make a comprehensive judgment based on the first judgment result obtained by the local judgment subunit and the second judgment result received by the receiving subunit. Whether the amount of change meets the preset conditions for stopping iteration.
19. The apparatus of claim 11, wherein the second party has a second private data set, and the second private data set includes a plurality of second private data;

    The device also includes:

    The second joint calculation unit is configured to jointly calculate the second ciphertext distance between the second private data and each first central data based on the homomorphic encryption method, so that the second The party obtains the second plaintext distance between the second private data and the first central data after decryption according to the second ciphertext distance.
20. The apparatus of claim 19, wherein the second joint computing unit comprises:

    A receiving subunit, configured to receive second encrypted data from the second party; the second encrypted data is obtained by the second party encrypting the second private data with the public key of the second party;

    A calculating subunit for homomorphically calculating the distance between the second encrypted data received by the receiving subunit and the first central data to obtain the second ciphertext distance;

    The sending subunit is configured to send the second ciphertext distance obtained by the calculating subunit to the second party.
21. A computer-readable storage medium having a computer program stored thereon, and when the computer program is executed in a computer, the computer is caused to execute the method according to any one of claims 1-10.
22. A computing device, comprising a memory and a processor, the memory stores executable code, and when the processor executes the executable code, the method according to any one of claims 1-10 is implemented.

Images