Data backtracking method based on FATE privacy calculation

Abstract

The invention discloses a data backtracking method based on FATE privacy calculation, which comprises the following steps: S1, FATE service installation and unilateral test completion; s2, FATE interconnection is carried out, and bilateral testing is completed; s3, accessing a user FATE through an independently developed FATE federation service; s4, accessing the FATE federated service to an FATE federated algorithm; s5, the FATE federated service is accessed to the FATE task; s6, the model providing side customizes an encryption algorithm, and provides a mapping API or an offline file; s7, the model providing side extracts the receipts and encrypts the receipts through a user-defined encryption algorithm; s8, the model providing side initiates an FATE uploading task for the encrypted data through the federal service platform, uploads the data to an FATE server and notifies the data acquisition side; s9, the data acquisition side selects an FATE algorithm, and initiates an FATE federated modeling task through the federated service platform; s10, the data acquisition side initiates an FATE downloading task through the federated service platform after the federated modeling task is completed; and S11, after the data downloading task is completed, the data acquisition side processes the data through a customized decryption algorithm and an offline file and then stores the data. According to the method, the problems of data flow, privacy protection and cross-platform compatibility are effectively solved; meanwhile, convenient operation is provided, and the risk of data leakage is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of privacy computing, and in particular to a data backtracking method based on FATE privacy computing. BACKGROUND

[0002] With the development of big data technology, data privacy and security issues are attracting more and more attention. China implemented the Network Security Law of the People's Republic of China in 2017, and the European Union implemented the General Data Protection Regulation (GDPR) in 2018. These regulations require more caution when using and exchanging data, making data sources independent of each other as "data silos", limiting traditional data sharing and cooperation, and urgently needing a technology to enable data cooperation while protecting data privacy.

[0003] However, the modeling framework of FATE solves the problem of "data silos" between enterprises, institutions and departments due to data ownership and privacy restrictions, allowing fragmented data scattered among different subjects (such as user credit data of financial institutions and consumer data of e-commerce) to be jointly modeled, improving model effectiveness, effectively protecting data security and user privacy, and realizing data usability, security compliance, and flow sharing. SUMMARY

[0004] One object of the present application is to provide a data backtracking method based on FATE privacy computing, which effectively solves the problems of data flow, privacy protection and cross-platform compatibility, and provides convenient operation and reduces the risk of data leakage.

[0005] Specific steps

[0006] A data backtracking method based on FATE privacy computing, characterized by the following steps: S1. FATE service installation to ensure that each user completes a unilateral test; S2. Plan user FATE number, confirm the need for interconnection users, exchange FATE interconnection information, and complete a bilateral test; S3. Access user FATE through a self-developed FATE federation service platform; S4. Access FATE algorithm in the self-developed FATE federation service platform, including but not limited to horizontal federated algorithm and vertical federated algorithm; S5. Access FATE tasks in the self-developed FATE federation service platform, including but not limited to data upload tasks, federated modeling tasks, and data download tasks; S6. Model provider side custom encryption algorithm, provide mapping API or offline file; S7. The model provides a side-drawing receipt, and then encrypts the receipt through a custom encryption algorithm. The encrypted data is uploaded to the FATE server through a FATE upload task initiated by the federal service platform and notifies the data acquisition side; S8. The data acquisition side selects a FATE algorithm and initiates a FATE federal modeling task through the federal service platform. After the task is completed, a FATE download task is initiated. S9. After the data download task is completed, the data acquisition side stores the data after processing through a custom decryption algorithm and offline file processing.

[0007] The data backtracking method based on FATE privacy calculation according to claim 1, wherein the step S1 comprises the following steps: S11. The user installs the FATE service; S12. The user completes a one-sided test.

[0008] The data backtracking method based on FATE privacy calculation according to claim 1, wherein the step S2 comprises the following steps: S21. Plan the FATE number of users, and each FATE user has a unique number, server and service port; S22. Confirm the users who need to be interconnected, and exchange the PK, IP and Port of each other; S23. The user configures the information (PK, IP, Port) of the opposite interconnection user to the FATE service of the user, and completes a two-sided test.

[0009] The data backtracking method based on FATE privacy calculation according to claim 1, wherein the step S3 comprises the following steps: S31. Develop an independent FATE service platform; S32. Configure the FATE information of the user to the FATE service platform.

[0010] The data backtracking method based on FATE privacy calculation according to claim 1, wherein the step S4 comprises the following steps: S41. Access the FATE horizontal federal algorithm in the FATE service platform; S42. Access the FATE vertical federal algorithm in the FATE service platform.

[0011] The data backtracking method based on FATE privacy calculation according to claim 1, wherein the step S5 comprises the following steps: S51. Access the FATE upload task in the FATE service platform; S52. Access the FATE federation modeling task in the FATE service platform; S53. Access the FATE download task in the FATE service platform.

[0012] The data backtracking method based on FATE privacy calculation according to claim 1, wherein the step S6 comprises the following steps: S61. The data provider customizes an encryption algorithm, which formulates different offsets for each user; S62. The encryption algorithm f(c, PK_of) = [(ord(c) + u) × 10] / 10, wherein c represents a character, u represents an offset, ord(c) represents a Unicode code point value of the character, and floor() represents a down rounding operation; S63. Each user formulates a different model mapping relationship, provides a model mapping API or an offline file, and according to the formula f([M_n]) = [M_id], wherein M_n represents the name of a model, M_id represents the id of the model, and [M_n] represents a list set.

[0013] The data backtracking method based on FATE privacy calculation according to claim 1, wherein the step S7 comprises the following steps: S71. Extract model data; S72. Cut the data into a set of a specific length and encrypt the data through the encryption algorithm in the step S6. S73. Upload the encrypted data in the step S72 to the FATE server through the FATE federation service platform; S74. After the uploading is completed, notify the data acquisition.

[0014] The data backtracking method based on FATE privacy calculation according to claim 1, wherein the step S8 comprises the following steps: S81. After receiving the notification in the step S7, select a FATE federation algorithm that is intended to be used; S82. After confirming the federation algorithm, construct a FATE federation modeling task through the FATE federation service platform. S83. After the task in the step S82 is completed, construct a FATE download task through the FATE federation service platform.

[0015] The data backtracking method based on FATE privacy calculation according to claim 1, wherein the step S9 comprises the following steps: S91. The decryption algorithm is according to the model mapping API or the offline file in the step S6, and according to the formula f(M₍id₎) = M nConvert the model id into the model name, where the name represents the model, the id represents the model, and according to the formula f([n], PK o f) = [chr([n] − PK o f)] number conversion to the corresponding Unicode character, representing a list set, and finally through the formula S = ∏ k =1 n c k Convert the corresponding data, where k represents the kth element in the character sequence, represents the string concatenation operation, and represents the result string after splicing; S92. Store the final data.

Claims

1. A data backtracking method based on FATE privacy computing, characterized in that, Includes the following steps: S1.FATE service installation ensures that each user completes a one-sided test; S2. Plan user FATE numbers, identify users that need to be interconnected, exchange FATE interconnection information, and complete bilateral testing; S3. Access user FATE through the self-developed FATE Federation service platform; S4. Integrate the FATE algorithm into the self-developed FATE Federation service platform. The FATE algorithm includes, but is not limited to, horizontal federation algorithm and vertical federation algorithm. S5. Integrate FATE tasks into the self-developed FATE Federation service platform. FATE algorithms include, but are not limited to, data upload tasks, federated modeling tasks, and data download tasks. S6. The model provides a custom encryption algorithm and offers a mapping API or offline file; S7. After the model provider extracts the receipt, it encrypts it using a custom encryption algorithm. The encrypted data is then uploaded to the FATE server via a FATE upload task initiated by the federated service platform, and the data acquisition side is notified. S8. The data acquisition side selects the FATE algorithm, initiates the FATE federated modeling task through the federated service platform, and continues to initiate the FATE download task after the task is completed. S9. After the data download task is completed, the data is stored after the data is processed by a customized decryption algorithm and offline files.

2. The data backtracking method based on FATE privacy computation according to claim 1, characterized in that, Step S1 includes the following steps: S11. The user installs the FATE service; S12. The user completes the one-sided test.

3. The data backtracking method based on FATE privacy computation according to claim 1, characterized in that, Step S2 includes the following steps: S21. Plan user FATE numbers. Each FATE user has its own unique number, server, and service port. S22. Identify the users who need to connect, and the users exchange their PK, IP, and Port; S23. The user configures the other party's interconnected user information (PK, IP, Port) into their own FATE service to complete the bilateral test.

4. The data backtracking method based on FATE privacy computing according to claim 1, characterized in that, Step S3 includes the following steps: S31. Develop an independent FATE service platform; S32. Configure the user's FATE information to the FATE service platform.

5. A data backtracking method based on FATE privacy computation according to claim 1, characterized in that, Step S4 includes the following steps: S41. Integrate the FATE horizontal federation algorithm into the FATE service platform; S42. Integrate the FATE vertical federation algorithm into the FATE service platform.

6. A data backtracking method based on FATE privacy computation according to claim 1, characterized in that, Step S5 includes the following steps: S51. Integrate the FATE upload task into the FATE service platform; S52. Integrate the FATE Federation modeling task into the FATE service platform; S53. Integrate the FATE download task into the FATE service platform.

7. A data backtracking method based on FATE privacy computing according to claim 1, characterized in that, Step S6 includes the following steps: S61. Data provides a customized encryption algorithm, which sets a different offset for each user; S62. Encryption algorithm f(c, PK_of) = [(ord(c) + u) × 10] / 10 where c represents the character, u represents the offset, PK_of represents the Unicode code point value of the character, u is a random real number uniformly distributed within the interval, and PK_of represents the rounding down; S63. Each user defines a different model mapping relationship, providing a model mapping API or offline files, according to the formula f([M_n]) = [M_id] where M_n represents the model name, M_id represents the model ID, and M_id represents the list set.

8. A data backtracking method based on FATE privacy computation according to claim 1, characterized in that, Step S7 includes the following steps: S71. Extract model data; S72. Cut the data into sets of a specific length and encrypt them using the encryption algorithm from step S6. S73. Upload the encrypted data from step S72 to the FATE server via the FATE Federation Service Platform; S74. Notify the data acquisition test after the upload is complete.

9. A data backtracking method based on FATE privacy computation according to claim 1, characterized in that, Step S8 includes the following steps: S81. After receiving the notification from step S7, select the desired FATE federation algorithm; S82. After confirming the federated algorithm, construct the FATE federated modeling task through the FATE federated service platform. S83. After the task in step S82 is completed, construct the FATE download task through the FATE federated service platform.

10. A data backtracking method based on FATE privacy computation according to claim 1, characterized in that, Step S9 includes the following steps: S91. Decryption calculation: Based on the model mapping API or offline file in step S6, according to the formula f(M₍id₎) = M n Convert the model ID to the model name, where represents the model name and represents the model ID. Simultaneously, according to the formula f([n], PK)... o f) = [chr([n] − PK] o f)] The number is converted to its corresponding Unicode character, representing a list or set, and finally the formula S = ∏ k =1 n c k Convert into corresponding data, where represents the k-th element in the character sequence, represents the string concatenation operation, and represents the concatenated string; S92. Store the final data.