CN109992977A - A data anomaly point cleaning method based on secure multi-party computing technology - Google Patents

Abstract

The invention belongs to the technical field of information security, and discloses a method for cleaning abnormal data points based on a secure multi-party computing technology. The method for cleaning abnormal data points based on the secure multi-party computing technology includes: unifying the data of two parties A and B. It is a matrix format, with the same dimensions, and the last dimension is the AVF value of the piece of data; participant A and participant B use Yao's encryption algorithm in the secure multi-party computation algorithm ABY to encrypt the data matrix; server A and server B pair The encrypted data set uploaded by each participant is cleaned for abnormal data points. The invention combines the secure multi-party computing technology and the AVF abnormal value detection algorithm, uses the existing secure multi-party computing tool ABY algorithm, realizes the efficient detection of high-dimensional data, and uses the secure multi-party computing technology on the premise of ensuring a certain efficiency. Yao's encryption algorithm ensures the security of the data privacy of all parties.

Description

A data anomaly point cleaning method based on secure multi-party computing technology

technical field

The invention belongs to the technical field of information security, and in particular relates to a method for cleaning abnormal data points based on a secure multi-party computing technology.

Background technique

At present, the closest existing technology: joint data source refers to the fact that in the process of machine learning training, multiple participants have the same type of data. Fusion of these data can expand the scale of the training data set and improve the accuracy of the model training results. . Since the development of machine learning, the quality of the model depends to a large extent on the scale and quality of the data set. Therefore, joint data source learning has become a major trend in the development of machine learning. However, with the advantages of joint data source training, there is a new problem of data privacy and security protection of multiple data sources. In some scenarios, the data owned by each participant may be sensitive to privacy, such as some commercial data or some customers' data. Private information, such as medical information or property information, has extremely high requirements for privacy protection, and it is naturally difficult to share it at will.

With the increasing demand for data fusion, algorithms for protecting data privacy have emerged one after another. For example, in the method of adding a trusted third party, multiple participants jointly authenticate a trusted third party, upload their own plaintext data to the third party, and the third party performs tasks such as data cleaning and training. The trusted third party is often some Credible organizations, or some cloud computing providers that offer paid services. The advantage of this is to realize the privacy protection of data, and also achieve the purpose of data fusion. However, this algorithm has certain security risks. Trusted third parties are often honest but curious. If there is unforeseen data leakage during the process of collecting data for processing, or encountering malicious third parties stealing data information, it is often will cause serious consequences.

With the integration of technologies in various fields, the idea of cryptography has been applied to the field of joint data source training, that is, using mature encryption algorithms to encrypt the data of each participant, and then collect the encrypted data and send it to trusted third parties. The three parties, the trusted third party does not own sensitive plaintext data, but only possesses ciphertext data that seems meaningless after encryption. Equivalent to performing the same operation on plaintext, this encryption method ensures the feasibility of ciphertext training, which greatly ensures the privacy of data. However, there are also practical problems with such algorithms. The biggest problem is the game between security and efficiency. The existing homomorphic encryption algorithms often require a lot of time and computing resources to obtain the results, and the privacy requirements are not so high. In high scenarios, this algorithm has very low efficiency and is not suitable for mass promotion.

Prior art 1 proposes an algorithm that uses a homomorphic encryption algorithm to solve the multi-data source joint data outlier cleaning algorithm. The homomorphic encryption algorithm is used to encrypt the data of each party, and then the AVF outlier detection algorithm is used to detect the outliers in the data set. Screening and cleaning are performed, but due to the efficiency limitation of homomorphic encryption itself, its encryption and decryption requires more time and computing resources, resulting in a relatively low computational efficiency of the algorithm, which cannot meet a large number of data processing needs; the second prior art proposes A privacy-preserving data cleaning scheme based on the LOF outlier detection algorithm is proposed, but due to the nature of whether the data is an outlier or not based on the data distribution density, if the dimension of the data is high, it cannot effectively distinguish the abnormality according to the difference in distribution density. Therefore, this technology has a certain problem of low processing efficiency in the face of high-dimensional data sets.

To sum up, the problems existing in the prior art are:

(1) Existing algorithms that use homomorphic encryption algorithms to solve joint data outlier cleaning of multiple data sources have low computational efficiency and cannot meet a large number of data processing needs.

(2) The existing privacy-preserving data cleaning schemes based on the LOF outlier detection algorithm have the problem of low processing efficiency when faced with high-dimensional data sets.

In view of the problems of the above technologies, a new technology that can balance computing efficiency and security is required, which can improve the low efficiency and high energy consumption of traditional homomorphic encryption algorithms, and can also ensure the necessary data privacy and security requirements. To adapt to actual implementation examples, it is also necessary to support the processing of high-dimensional data.

The significance of solving the above technical problems:

After improving the problems existing in the above technologies, the algorithm can be more adapted to the actual use environment, improve the actual use efficiency, increase the implementability of the algorithm, and better protect the privacy and security of sensitive data.

SUMMARY OF THE INVENTION

Aiming at the problems existing in the prior art, the present invention provides a method for cleaning abnormal data points based on a secure multi-party computing technology.

The present invention is implemented in the following way: a method for cleaning abnormal data points based on secure multi-party computing technology, and the method for cleaning abnormal data points based on secure multi-party computing technology includes:

The first step is to unify the data of the two parties A and B into a matrix format, with the same dimensions, and the last dimension is the AVF value of the data;

In the second step, participant A and participant B use Yao's encryption algorithm in the secure multi-party computation algorithm ABY to encrypt the data matrix;

In the third step, server A and server B perform data anomaly cleaning on the encrypted data set uploaded by each participant.

Further, in the first step, participant A and participant B unify their own data set formats according to regulations:

Among them, D ₁ represents the N×(M+1) dataset matrix of participant A, a _ij represents any data in participant A’s dataset, avf _ai represents the AVF value of participant A’s ith data, i∈[ 1, N], j∈[1, M], M, N∈N ⁺ ; D ₂ represents the P×(M+1) dataset matrix of participant B, and b _kj represents any data in the dataset of participant B , avf _bk represents the AVF value of the kth data of participant A, k∈[1, P], j∈[1, M], M, P∈N ⁺ . Two of the parties have the same data dimension.

Further, in the second step, participant A and participant B encrypt their own data sets according to the regulations, including:

1) Use Yao's encryption algorithm in the secure multi-party computation encryption ABY algorithm to encrypt the data set D ₁ of the participant A:

in, Indicates the part where the encrypted data set is handed over to server A, Indicates that the encrypted data set is handed over to server B, Enc represents Yao's encryption algorithm, and D ₁ represents the data set of participant A;

Specifically, encrypt each element as follows:

in, Indicates the part where the encrypted data is handed over to server A, Indicates that the encrypted data is handed over to server B, and a _ij represents any data of participant A; The part indicating that the encrypted AVF value of the i-th data of the participant A is handed over to the server A, Represents the part where the encrypted AVF value of the i-th data of the participant A is handed over to the server B, and avf _ai represents the AVF value of the i-th data of the participant A;

2) Use the following formula to represent the encrypted data set of participant A:

Among them, X ₁₀ represents the encrypted data set of participant A held by server A, X ₁₁ represents the encrypted data set of participant A held by server B, i ∈ [1, N], j ∈ [1, M], M, N∈N ⁺ ;

3) Use Yao's encryption algorithm in the secure multi-party computation encryption ABY algorithm to encrypt the data set D ₂ of the participant B:

in, Indicates the part where the encrypted data set is handed over to server A, Indicates that the encrypted data set is handed over to server B, Enc represents Yao's encryption algorithm, and D ₂ represents the data set of participant B;

Specifically, encrypt each element as follows:

in, Indicates the part where the encrypted data is handed over to server A, Indicates that the encrypted data is handed over to server B, and b _kj represents any data of participant A; The part indicating that the AVF value of the k-th data of the encrypted participant B is handed over to the server A, Represents the part where the encrypted AVF value of the k-th data of the participant B is handed over to the server B, and avf _bk represents the AVF value of the k-th data of the participant B;

4) Use the following formula to represent the encrypted data set of participant B:

Among them, X ₂₀ represents the encrypted data set of participant B held by server A, X ₂₁ represents the encrypted data set of participant B held by server B, k ∈ [1, P], j ∈ [1, M], M, P∈N ⁺ ;

5) Participant A and Participant B respectively upload the encrypted data to the corresponding server.

Further, in the third step, server A and server B perform data anomaly cleaning on the encrypted data set uploaded by each participant, which specifically includes:

1) Server A extracts the last one-dimensional data in the encrypted data set of participant A obtained by itself:

Server A sorts A ₁₀ using the sorting algorithm in Yao's encryption algorithm in secure encryption algorithm ABY:

A' ₁₀ =Sort(A ₁₀ );

Among them, A ₁₀ represents the last one-dimensional data in the encrypted data set of participant A owned by server A, A' ₁₀ represents the data after A ₁₀ is sorted in descending order, and Sort() represents the sorting algorithm in Yao's encryption algorithm;

Sort X ₁₀ at the same time based on A ₁₀ , that is, arrange X ₁₀ in descending order of A _10. After sorting is completed:

Among them, X′ ₁₀ is the data set submitted by participant A to server A after the last one-dimensional data of X ₁₀ , that is, A ₁₀ is the benchmark after the descending sorting is completed, i∈[1, N], j∈[1, M] , M, N∈N ⁺ ;

A fixed value Thre is specified, indicating the threshold value of the AVF value within the normal range, and the data in A' ₁₀ is compared with Thre in order:

Res _i =Comp(A' _10i , Thre);

Among them, A' _10i represents the element in A' ₁₀ , i∈[1, N], N∈N ⁺ , Comp() represents the comparison size algorithm in Yao's encryption algorithm, Res _i represents the result of comparing A' _10i with Thre , if the value of Res _i is 1, it means that A' _10i ≥ Thre; if the value of Res _i is 0, it means that A' _10i <Thre, compare the data in A' ₁₀ with Thre in order, until Res _i =0, Stop the comparison and keep the first i rows of data in X' ₁₀ :

Among them, I=i, is the first i row data retained after sorting, j∈[1, M], M∈N ⁺ , X″ ₁₀ is the data set of participant A owned by server A after the final data cleaning is completed;

2) Server A extracts the last one-dimensional data in the encrypted data set of participant B obtained by itself:

Server A sorts A ₂₀ using the sorting algorithm in Yao's encryption algorithm in secure encryption algorithm ABY:

A' ₂₀ =Sort(A ₂₀ );

Among them, A ₂₀ represents the last one-dimensional data in the encrypted data set of participant B owned by server A, A' ₂₀ represents the data after A ₂₀ is sorted in descending order, and Sort() represents the sorting algorithm in Yao's encryption algorithm;

Sort X ₂₀ at the same time based on A ₂₀ , that is, sort X ₂₀ in descending order of A _20. After sorting is completed:

Among them, X′ ₂₀ is the data set submitted by participant B to server A after the last one-dimensional data of X ₂₀ , that is, A ₂₀ is the benchmark after the descending sorting is completed, k∈[1, P], j∈[1, M] , M, P∈N ⁺ ;

A fixed value Thre is specified, indicating the threshold value of the AVF value within the normal range, and the data in A' ₂₀ is compared with Thre in order:

Res _k =Comp(A' _20k , Thre);

Among them, A' _20k represents the element in A' ₂₀ , k∈[1, P], P∈N ⁺ , Comp() represents the comparison size algorithm in Yao's encryption algorithm, Res _k represents the result of comparing A' _20k with Thre , if the value of Res _k is 1, it means that A' _20k ≥ Thre; if the value of Res _k is 0, it means that A' _20k <Thre, compare the data in A' ₂₀ with Thre in order until Res _k =0, Stop the comparison and keep the first k rows of data in X' ₂₀ :

Among them, K=k, is the first k rows of data retained after sorting, j∈[1, M], M∈N ⁺ , X″ ₂₀ is the data set of participant B owned by server A after the final data cleaning is completed;

3) Server B extracts the last one-dimensional data in the encrypted data set of participant A obtained by itself:

Server B sorts A ₁₁ using the sorting algorithm in Yao's encryption algorithm in secure encryption algorithm ABY:

A' ₁₁ =Sort(A ₁₁ );

Among them, A ₁₁ represents the last one-dimensional data in the encrypted data set of participant A owned by server B, A' ₁₁ represents the data after A ₁₁ is sorted in descending order, and Sort() represents the sorting algorithm in Yao's encryption algorithm;

Sort X ₁₁ at the same time based on A ₁₁ , that is, sort X ₁₁ in descending order of A _11. After sorting is completed:

Among them, X′ ₁₁ is the data set submitted by participant A to server B after the last one-dimensional data of X ₁₁ , that is, A ₁₁ is the benchmark after the descending sorting is completed, i∈[1, N], j∈[1, M] , M, N∈N ⁺ ;

A fixed value Thre is specified, indicating the threshold value of the AVF value within the normal range, and the data in A' ₁₁ is compared with Thre in order:

Res _i =Comp(A' _11i , Thre);

Among them, A' _11i represents the element in A' ₁₁ , i∈[1, N], N∈N ⁺ , Comp() represents the comparison size algorithm in Yao's encryption algorithm, Res _i represents the result of comparing A' _11i with Thre , if the value of Res _i is 1, it means that A' _11i ≥ Thre; if the value of Res _i is 0, it means that A' _11i <Thre, compare the data in A' ₁₁ with Thre in order, until Res _i =0, Stop the comparison and keep the first i rows of data in _X'11 :

Among them, I=i, is the first i row data retained after sorting, j∈[1, M], M∈N ⁺ , X″ ₁₁ is the data set of participant A owned by server B after the final data cleaning is completed;

4) Server B extracts the last one-dimensional data in the encrypted data set of participant B obtained by itself:

Server B sorts A ₂₁ using the sorting algorithm in Yao's encryption algorithm in secure encryption algorithm ABY:

A' ₂₁ =Sort(A ₂₁ );

Among them, A ₂₁ represents the last one-dimensional data in the encrypted data set of participant B owned by server B, A' ₂₁ represents the data sorted by A ₂₁ in descending order, and Sort() represents the sorting algorithm in Yao's encryption algorithm.

Sort X ₂₁ at the same time based on A ₂₁ , that is, sort X ₂₁ in descending order of A ₂₁ , after sorting is completed:

Among them, X′ ₂₁ is the data set submitted by participant B to server B after the last one-dimensional data of X ₂₁ , that is, A ₂₁ is the benchmark after the descending sorting is completed, k∈[1, P], j∈[1, M] , M, P ∈ N ⁺ .

A fixed value Thre is specified, indicating the threshold value of the AVF value within the normal range, and the data in A' ₂₁ is compared with Thre in order:

Res _k =Comp(A' _21k , Thre);

Among them, A' _21k represents the element in A' ₂₁ , k∈[1, P], P∈N ⁺ , Comp() represents the comparison size algorithm in Yao's encryption algorithm, Res _k represents the result of comparing A' _21k with Thre , if the value of Res _k is 1, it means that A' _21k ≥ Thre; if the value of Res _k is 0, it means that A' _21k <Thre, compare the data in A' ₂₁ with Thre in order, until Res _k =0, Stop the comparison and keep the first k rows of data in _X'21 :

Among them, K=k, is the first k rows of data retained after sorting, j∈[1, M], M∈N ⁺ , X″ ₂₁ is the data set of participant B owned by server B after the final data cleaning is completed;

5) The finally obtained X″ ₁₀ , X″ ₁₁ , X″ ₂₀ , and X″ ₂₁ are the data set after the final data cleaning is completed.

Another object of the present invention is to provide a machine learning system applying the method for cleaning abnormal data points based on the secure multi-party computing technology.

To sum up, the advantages and positive effects of the present invention are as follows: the present invention combines the secure multi-party computing technology and the AVF outlier detection algorithm, and utilizes the existing secure multi-party computing tool ABY algorithm to achieve efficient detection of high-dimensional data, and Under the premise of ensuring a certain efficiency, the Yao's encryption algorithm in the secure multi-party computing technology is used to ensure the security of the data privacy of all parties.

Table 1 Technical performance comparison

Description of drawings

FIG. 1 is a flowchart of a method for cleaning abnormal data points based on a secure multi-party computing technology provided by an embodiment of the present invention.

FIG. 2 is a schematic diagram of a scenario of an embodiment provided by an embodiment of the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

In view of the existing algorithms that use homomorphic encryption algorithm to solve the multi-data source joint data outlier cleaning, the computational efficiency is low and cannot meet the needs of a large number of data processing; the existing privacy protection data cleaning solutions based on the LOF outlier detection algorithm face the Inefficient processing of high-dimensional datasets. The invention is mainly used to realize the safe data abnormal point cleaning algorithm in the joint data source environment; based on the secure multi-party computing technology, the data abnormal point cleaning work under the joint machine learning scenario of multiple data sources under the premise of privacy protection is realized.

The application principle of the present invention will be described in detail below with reference to the accompanying drawings.

As shown in FIG. 1 , the method for cleaning abnormal data points based on the secure multi-party computing technology provided by the embodiment of the present invention includes the following steps:

S101: Participant A and Participant B unify their own data set formats according to the regulations: unify the data of the two parties A and B into a matrix format with the same dimensions, and the last dimension is the Attribute Value Frequency (AVF )value;

S102: Participant A and Participant B encrypt their own data sets according to regulations: Participant A and Participant B encrypt the data matrix using Yao's encryption algorithm in the secure multi-party computing algorithm ABY;

S103: Server A and server B perform data anomaly point cleaning on the encrypted data set uploaded by each participant.

The application principle of the present invention will be further described below with reference to specific embodiments.

The method for cleaning abnormal data points based on the secure multi-party computing technology provided by the embodiment of the present invention specifically includes the following steps:

Step 1, Participant A and Participant B unify their own data set format according to the regulations:

Among them, D ₁ represents the N×(M+1) data set matrix of the participant A, a _ij represents any data in the data set of the participant A, and avf _ai represents the Attribute Value Frequency (AVF) of the i-th data of the participant A Value, i∈[1, N], j∈[1, M], M, N∈N ⁺ ; D ₂ represents the P×(M+1) dataset matrix of participant B, and b _kj represents participant B Arbitrary data in the dataset, avf _bk represents the Attribute Value Frequency (AVF) value of the kth data of participant A, k∈[1, P], j∈[1, M], M, P∈N ⁺ . The data dimension (ie, the value of M) of the two parties is the same.

Step 2, Participant A and Participant B encrypt their own data sets according to the regulations:

2a) Use Yao's encryption algorithm in the secure multi-party computation encryption ABY algorithm to encrypt the data set D ₁ of the participant A:

in, Indicates the part where the encrypted data set is handed over to server A, Indicates that the encrypted data set is handed over to server B, Enc represents Yao's encryption algorithm, and D ₁ represents the data set of participant A.

Specifically, encrypt each element as follows:

in, Indicates the part where the encrypted data is handed over to server A, Indicates that the encrypted data is handed over to server B, and a _ij represents any data of participant A; The part indicating that the encrypted AVF value of the i-th data of the participant A is handed over to the server A, The AVF value representing the ith piece of data of the encrypted participant A is handed over to the server B, and avf _ai represents the AVF value of the ith piece of data of the participant A.

2b) Use the following formula to represent the encrypted data set of participant A:

Among them, X ₁₀ represents the encrypted data set of participant A held by server A, X ₁₁ represents the encrypted data set of participant A held by server B, i ∈ [1, N], j ∈ [1, M], M, N∈N ⁺ .

2c) Use Yao's encryption algorithm in the secure multi-party computation encryption ABY algorithm to encrypt the data set D ₂ of the participant B:

in, Indicates the part where the encrypted data set is handed over to server A, Indicates that the encrypted data set is handed over to server B, Enc represents Yao's encryption algorithm, and D ₂ represents the data set of participant B.

Specifically, encrypt each element as follows:

in, Indicates the part where the encrypted data is handed over to server A, Indicates that the encrypted data is handed over to server B, and b _kj represents any data of participant A; The part indicating that the AVF value of the k-th data of the encrypted participant B is handed over to the server A, The AVF value representing the k-th piece of data of the encrypted participant B is handed over to the server B, and avf _bk represents the AVF value of the k-th piece of data of the participant B.

2d) Use the following formula to represent the encrypted data set of Party B:

Among them, X ₂₀ represents the encrypted data set of participant B held by server A, X ₂₁ represents the encrypted data set of participant B held by server B, k ∈ [1, P], j ∈ [1, M], M, P∈N ⁺ .

2e) Participant A and Participant B respectively upload the encrypted data to the corresponding server.

Step 3, server A and server B clean the encrypted data set uploaded by each participant for abnormal data points:

3a) Server A extracts the last one-dimensional data in the encrypted data set of Participant A obtained by itself, namely:

Server A sorts A ₁₀ using the sorting algorithm in Yao's encryption algorithm in secure encryption algorithm ABY:

A' ₁₀ =Sort(A ₁₀ );

Among them, A ₁₀ represents the last one-dimensional data in the encrypted data set of participant A owned by server A, A' ₁₀ represents the data sorted by A ₁₀ in descending order, and Sort() represents the sorting algorithm in Yao's encryption algorithm.

Sort X ₁₀ at the same time based on A ₁₀ , that is, arrange X ₁₀ in descending order of A _10. After sorting is completed:

Among them, X′ ₁₀ is the data set submitted by participant A to server A after the last one-dimensional data of X ₁₀ , that is, A ₁₀ is the benchmark after the descending sorting is completed, i∈[1, N], j∈[1, M] , M, N ∈ N ⁺ .

A fixed value Thre is specified, indicating the threshold value of the AVF value within the normal range, and the data in A' ₁₀ is compared with Thre in order:

Res _i =Comp(A' _10i , Thre);

Among them, A' _10i represents the element in A' ₁₀ , i∈[1, N], N∈N ⁺ , Comp() represents the comparison size algorithm in Yao's encryption algorithm, Res _i represents the result of comparing A' _10i with Thre , if the value of Res _i is 1, it means that A' _10i ≥ Thre; if the value of Res _i is 0, it means that A' _10i <Thre, compare the data in A' ₁₀ with Thre in order, until Res _i =0, Stop the comparison and keep the first i rows of data in X' ₁₀ :

Among them, I=i, is the first i row of data retained after sorting, j∈[1, M}, M∈N ⁺ , X″ ₁₀ is the data set of participant A owned by server A after the final data cleaning is completed.

3b) Server A extracts the last one-dimensional data in the encrypted data set of participant B obtained by itself, namely:

Server A sorts A ₂₀ using the sorting algorithm in Yao's encryption algorithm in secure encryption algorithm ABY:

A' ₂₀ =Sort(A ₂₀ );

Among them, A ₂₀ represents the last one-dimensional data in the encrypted data set of participant B owned by server A, A' ₂₀ represents the data sorted by A ₂₀ in descending order, and Sort() represents the sorting algorithm in Yao's encryption algorithm.

Sort X ₂₀ at the same time based on A ₂₀ , that is, sort X ₂₀ in descending order of A _20. After sorting is completed:

Among them, X′ ₂₀ is the data set submitted by participant B to server A after the last one-dimensional data of X ₂₀ , that is, A ₂₀ is the benchmark after the descending sorting is completed, k∈[1, P], j∈[1, M] , M, P ∈ N ⁺ .

A fixed value Thre (same as Thre above) is specified, indicating the threshold value of the AVF value within the normal range, and A′ ₂₀

The data in , compare the size with Thre in order:

Res _k =Comp(A' _20k , Thre);

Among them, A' _20k represents the element in A' ₂₀ , k∈[1, P], P∈N ⁺ , Comp() represents the comparison size algorithm in Yao's encryption algorithm, Res _k represents the result of comparing A' _20k with Thre , if the value of Res _k is 1, it means that A' _20k ≥ Thre; if the value of Res _k is 0, it means that A' _20k <Thre, compare the data in A' ₂₀ with Thre in order until Res _k =0, Stop the comparison and keep the first k rows of data in X' ₂₀ :

Among them, K=k, is the first k rows of data retained after sorting, j∈[1, M], M∈N ⁺ , X″ ₂₀ is the data set of participant B owned by server A after the final data cleaning is completed.

3c) Server B extracts the last one-dimensional data in the encrypted data set of Participant A obtained by itself, namely:

Server B sorts A ₁₁ using the sorting algorithm in Yao's encryption algorithm in secure encryption algorithm ABY:

A' ₁₁ =Sort(A ₁₁ );

Among them, A ₁₁ represents the last one-dimensional data in the encrypted data set of participant A owned by server B, A' ₁₁ represents the data sorted by A ₁₁ in descending order, and Sort() represents the sorting algorithm in Yao's encryption algorithm.

Sort X ₁₁ at the same time based on A ₁₁ , that is, sort X ₁₁ in descending order of A _11. After sorting is completed:

Among them, X′ ₁₁ is the data set submitted by participant A to server B after the last one-dimensional data of X ₁₁ , that is, A ₁₁ is the benchmark after the descending sorting is completed, i∈[1, N], j∈[1, M] , M, N ∈ N ⁺ .

A fixed value Thre (same as Thre above) is specified, indicating the threshold value of the AVF value within the normal range, and the data in A' ₁₁ is compared with Thre in order:

Res _i =Comp(A' _11i , Thre);

Among them, A' _11i represents the element in A' ₁₁ , k∈[1, N], P∈N ⁺ , Comp() represents the comparison size algorithm in Yao's encryption algorithm, Res _i represents the result of comparing A' _11i with Thre , if the value of Res _i is 1, it means that A' _11i ≥ Thre; if the value of Res _i is 0, it means that A' _11i <Thre, compare the data in A' ₁₁ with Thre in order, until Res _i =0, Stop the comparison and keep the first i rows of data in _X'11 :

Among them, I=i, is the first i row data retained after sorting, j∈[1, M], M∈N ⁺ , X″ ₁₁ is the data set of participant A owned by server B after the final data cleaning is completed.

3d) Server B extracts the last one-dimensional data in the encrypted data set of participant B obtained by itself, namely:

Server B sorts A ₂₁ using the sorting algorithm in Yao's encryption algorithm in secure encryption algorithm ABY:

A' ₂₁ =Sort(A ₂₁ );

Among them, A ₂₁ represents the last one-dimensional data in the encrypted data set of participant B owned by server B, A' ₂₁ represents the data sorted by A ₂₁ in descending order, and Sort() represents the sorting algorithm in Yao's encryption algorithm.

Sort X ₂₁ at the same time based on A ₂₁ , that is, sort X ₂₁ in descending order of A ₂₁ , after sorting is completed:

Among them, X′ ₂₁ is the data set submitted by participant B to server B after the last one-dimensional data of X ₂₁ , that is, A ₂₁ is the benchmark after the descending sorting is completed, k∈[1, P], j∈[1, M] , M, P ∈ N ⁺ .

A fixed value Thre (same as Thre above) is specified, indicating the threshold value of the AVF value within the normal range, and the data in A' ₂₁ is compared with Thre in order:

Res _k =Comp(A' _21k , Thre);

Among them, A' _21k represents the element in A' ₂₁ , k∈[1, P], P∈N ⁺ , Comp() represents the comparison size algorithm in Yao's encryption algorithm, Res _k represents the result of comparing A' _21k with Thre , if the value of Res _k is 1, it means that A' _21k ≥ Thre; if the value of Res _k is 0, it means that A' _21k <Thre, compare the data in A' _21k with Thre in order, until Res _k =0, Stop the comparison and keep the first k rows of data in _X'21 :

Among them, K=k, is the first k rows of data retained after sorting, j∈[1, M], M∈N ⁺ , X″ ₂₁ is the data set of participant B owned by server B after the final data cleaning is completed.

3e) The finally obtained X″ ₁₀ , X″ ₁₁ , X″ ₂₀ , and X″ ₂₁ are the data set after the final data cleaning is completed.

The above descriptions are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention shall be included in the protection of the present invention. within the range.

Claims (5)

1. A data abnormal point cleaning method based on a secure multi-party computing technology is characterized by comprising the following steps:

step one, unifying data of two parties A and B into a matrix format, wherein the data have the same dimensionality, and the last dimension is the AVF value of the data;

secondly, encrypting the data matrix by the participant A and the participant B by using a Yao's encryption algorithm in a secure multi-party computing algorithm ABY;

and thirdly, the server A and the server B perform data anomaly point cleaning on the encrypted data sets uploaded by the participants.

2. The secure multi-party computing technology based data anomaly cleansing method according to claim 1, wherein said first step participant a and participant B unify their own data set formats as specified:

wherein D is₁A data set matrix of N (M +1) representing party A, a_ijRepresenting arbitrary data in the participant a dataset, avf_aiAVF value representing participant A ith data, i ∈ [1, N]，j∈[1，M]，M，N∈N⁺；D₂A dataset matrix representing P x (M +1) of party B, B_kjRepresenting arbitrary data in the participant B dataset, avf_bkAVF value representing data of k-th item of participant A, k ∈ [1, P]，j∈[1，M]，M，P∈N⁺(ii) a Where the data dimensions of both participants are the same.

3. The method for cleansing data anomaly points based on secure multiparty computing technology according to claim 1, wherein said second step of encrypting the owned data set by party a and party B specifically comprises:

1) encrypting data set D of participant A by Yao's encryption algorithm in ABY algorithm by utilizing secure multi-party computation₁And (3) encryption:

wherein,a section representing that the encrypted data set is handed to the server a,representing the part of the encrypted data set handed to server B, Enc representing the Yao's encryption algorithm, D₁A data set representing party a;

specifically, each element is encrypted according to the following formula:

wherein,a section showing that the encrypted data is handed to the server a,a section showing that the encrypted data is handed to the server B, a_ijAny data representing party a;the AVF value representing the encrypted ith piece of data of party a is handed to part of server a,part of server B handed over the AVF value representing encrypted participant a's ith piece of data, AVF_aiAVF value representing participant a's ith data;

2) the encrypted data set of party a is represented by:

wherein, X₁₀An encrypted data set, X, representing a party A held by a server A₁₁An encrypted data set representing party A held by Server B, i ∈ [1, N]，j∈[1，M]，M，N∈N⁺；

3) Encrypting data set D of participant B using Yao's encryption algorithm in ABY algorithm using secure multi-party computation₂And (3) encryption:

wherein,a section representing that the encrypted data set is handed to the server a,representing the part of the encrypted data set handed to server B, Enc representing the Yao's encryption algorithm, D₂A data set representing party B;

specifically, each element is encrypted according to the following formula:

wherein,a section showing that the encrypted data is handed to the server a,a section showing that the encrypted data is handed to the server B, B_kjAny data representing party a;the AVF value representing the encrypted kth piece of data of party B is handed over to part of server a,part of server B handed over the AVF value representing the encrypted kth piece of data of party B, AVF_bkAVF value representing the kth piece of data for party B;

4) the encrypted data set of party B is represented by:

wherein, X₂₀An encrypted data set, X, representing a party B held by a server A₂₁An encrypted data set representing party B held by Server B, k ∈ [1, P ∈]，j∈[1，M]，M，P∈N⁺；

5) And the participant A and the participant B respectively upload the encrypted data to corresponding servers.

4. The method for cleaning data anomaly points based on the secure multi-party computing technology as claimed in claim 1, wherein the third step of cleaning the data anomaly points of the encrypted data sets uploaded by the participants by the server a and the server B specifically comprises:

1) the server A extracts the last one-dimensional data in the encrypted data set of the party A taken by the server A:

server A uses the sorting algorithm in Yao's encryption algorithm in security encryption algorithm ABY to pair A₁₀And (3) sequencing:

A′₁₀＝Sort(A₁₀)；

wherein A is₁₀Last-dimensional data, A ', in the encrypted data set representing party A owned by Server A'₁₀Represents A'₁₀Sorting the finished data in a descending order, wherein the Sort () represents a sorting algorithm in the Yao's encryption algorithm;

with A₁₀Taking X as a reference₁₀Also sorted simultaneously, i.e. according to A₁₀Descending order of X₁₀And after the sorting is finished:

wherein, X'₁₀Is represented by X₁₀Last one-dimensional data, i.e. A₁₀Submitting the data set of the server A for the participant A after the reference descending sorting is completed, wherein i belongs to [1, N ∈]，j∈[1，M]，M，N∈N⁺；

A 'is defined as a fixed value Thre representing a threshold value of AVF value within a normal range'₁₀The data in (1) is compared with Thre in order:

Res_i＝Comp(A′_10i，Thre)；

wherein, A'_10iRepresents A'₁₀The element in (1) is i' [1, N ]]，N∈N⁺Comp () denotes the comparative size algorithm, Res, in Yao's encryption algorithm_iRepresents A'_10iIf Res is the result of comparison with Thre_iValue is 1 and represents A'_10iMore than or equal to Thre; if Res_iValue is 0 and represents A'_10i< Thre, A'₁₀The data in (1) is compared with Thre in sequence until Res_i0, stop comparison, and mix X'₁₀First i row data retention in (1):

wherein, I is I, the first I row of data reserved after sorting, j belongs to [1, M ∈]，M∈N⁺，X″₁₀Of party A owned by Server A after completion of cleaning for final dataA data set;

2) the server A extracts the last one-dimensional data in the encrypted data set of the party B taken by the server A:

server A uses the sorting Algorithm of the Yao's encryption Algorithm of the secure encryption Algorithm ABY to pair A'₂₀And (3) sequencing:

A′₂₀＝Sort(A₂₀)；

wherein A is₂₀Last-dimensional data, A 'in encrypted data set representing party B owned by Server A'₂₀Is represented by A₂₀Sorting the finished data in a descending order, wherein the Sore () represents a sorting algorithm in the Yao's encryption algorithm;

with A₂₀Taking X as a reference₂₀Also sorted simultaneously, i.e. according to A₂₀Descending order of X₂₀And after the sorting is finished:

wherein, X'₂₀Is represented by X₂₀Last one-dimensional data, i.e. A₂₀Submitting the data set of the server A for the participant B after the reference descending sorting is completed, wherein k belongs to [1, P ]]，j∈[1，M]，M，P∈N⁺；

A 'is defined as a fixed value Thre representing a threshold value of AVF value within a normal range'₂₀The data in (1) is compared with Thre in order:

Res_k＝Comp(A′_20k，Thre)；

wherein, A'_20kRepresents A'₂₀Element (b) k ∈ [1, P ]]，P∈N⁺Comp () denotes the comparative size algorithm, Res, in Yao's encryption algorithm_kRepresents A'_20kIf Res is the result of comparison with Thre_kValue is 1 and represents A'_20kMore than or equal to Thre; if Res_kValue is 0 and represents A'_20k< Thre, A'₂₀In order of the data inSize comparison with Thre until Res_k0, stop comparison, and mix X'₂₀First k rows of data retention:

where K is K, the first K rows of data retained after sorting, j e [1, M]，M∈N⁺，X″₂₀The data set of the participant B owned by the server A after the final data cleaning is finished;

3) the server B extracts the last one-dimensional data in the encrypted data set of the party A taken by the server B:

server B uses the sorting algorithm in Yao's encryption algorithm in secure encryption algorithm ABY to pair a₁₁And (3) sequencing:

A′₁₁＝Sort(A₁₁)；

wherein A is₁₁Last-dimensional data, A 'in encrypted data set representing party A owned by server B'₁₁Is represented by A₁₁Sorting the finished data in a descending order, wherein the Sort () represents a sorting algorithm in the Yao's encryption algorithm;

with A₁₁Taking X as a reference₁₁Also sorted simultaneously, i.e. according to A₁₁Descending order of X₁₁And after the sorting is finished:

wherein, A'₁₁Is represented by X₁₁Last one-dimensional data, i.e. A₁₁Submitting the data set of the server B for the participant A after the reference descending sorting is completed, wherein i belongs to [1, N ∈]，j∈[1，M]，M，N∈N⁺；

A 'is defined as a fixed value Thre representing a threshold value of AVF value within a normal range'₁₁The data in (1) is compared with Thre in order:

Res_i＝Comp(A′_11i，Thre)；

wherein, A'_11iRepresents A'₁₁Element in (1, N) is i ∈ [, N ∈ [ ]]，N∈N⁺Comp () denotes the comparative size algorithm, Res, in Yao's encryption algorithm_iRepresents A'_11iIf Res is the result of comparison with Thre_iValue is 1 and represents A'_11iMore than or equal to Thre; if Res_iValue is 0 and represents A'_11i< Thre, A'₁₁The data in (1) is compared with Thre in sequence until Res_i0, stop comparison, and mix X'₁₁First i row data retention in (1):

wherein, I is I, the first I row of data reserved after sorting, j belongs to [1, M ∈]，M∈N⁺，X″₁₁The data set of the participant A owned by the server B after the final data cleaning is finished;

4) the server B extracts the last one-dimensional data in the encrypted data set of the party B taken by the server B:

server B uses the sorting algorithm in Yao's encryption algorithm in secure encryption algorithm ABY to pair a₂₁And (3) sequencing:

A′₂₁＝Sort(A₂₁)；

wherein, A'₂₁Last-dimensional data, A 'in encrypted data set representing party B owned by server B'₂₁Is represented by A₂₁Sorting the finished data in a descending order, wherein the Sort () represents a sorting algorithm in the Yao's encryption algorithm;

with A₂₁Taking X as a reference₂₁Also sorted simultaneously, i.e. according to A₂₁Descending order of X₂₁And after the sorting is finished:

wherein, X'₂₁Is represented by X₂₁Last one-dimensional data, i.e. A₂₁Submitting the data set of the server B for the participant B after the reference descending sorting is completed, wherein k belongs to [1, P ∈]，j∈[1，M]，M，P∈N⁺；

A 'is defined as a fixed value Thre representing a threshold value of AVF value within a normal range'₂₁The data in (1) is compared with Thre in order:

Res_k＝Comp(A′_21k，Thre)；

wherein, A'_21kRepresents A'₂₁Element (b) k ∈ [1, P ]]，P∈N⁺Comp () denotes the comparative size algorithm, Res, in Yao's encryption algorithm_kRepresents A'_21kIf Res is the result of comparison with Thre_kValue is 1 and represents A'_21kMore than or equal to Thre; if Res_kValue is 0 and represents A'_21k< Thre, A'₂₁The data in (1) is compared with Thre in sequence until Res_k0, stop comparison, and mix X'₂₁First k rows of data retention:

where K is K, the first K rows of data retained after sorting, j e [1, M]，M∈M⁺，X″₂₁The data set of the participant B owned by the server B after the final data cleaning is finished;

5) the final X ″)₁₀，X″₁₁，X″₂₀，X″₂₁And cleaning the finished data set for the final data.

5. A machine learning system applying the data anomaly point cleaning method based on the secure multi-party computing technology according to any one of claims 1 to 4.