Distributed multi-party security model training framework for privacy protection
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a multi-party secure logistic regression model (SLRM). One of the methods includes receiving, at a plurality of secure computation nodes (SCNs), a plurality of random numbers from a random number provider; encrypting, at each SCN, data stored at the SCN using the received random numbers; iteratively updating a secure logistic regression model (SLRM) by using the encrypted data from each SCN; and after iteratively updating the SLRM, outputting a result of the SLRM, wherein the result is configured to enable a service to be performed by each SCN.
1. A computer-implemented method for training multi-party secure logistic regression models, the computer-implemented method comprising:
receiving, by a first secure computation node (SCN) of a plurality of secure computation nodes (SCNs), a first random number of a plurality of random numbers from a random number provider;
encrypting, by the first SCN, data stored at the SCN using the first random number to generate a first encrypted data;
iteratively updating, by the first SCN, a secure logistic regression model (SLRM) by using the first encrypted data to generate an interim computation result that is divided into a plurality of shares that are transmitted to the plurality of SCNs to be used as subsequent inputs for the SLRM to determine a column vector of a prediction error as a difference between a sigmoid fitting function applied to a respective share and a label of a sample of a mini batch of a Stochastic Gradient Descent method, each of the plurality of shares hiding original data such that the original data cannot be recovered by using a single share of the plurality of shares; and
after iteratively updating the SLRM, outputting a result of the SLRM, wherein the result is configured to enable a service to be performed by each SCN of the plurality of SCNs.
2. The computer-implemented method of claim 1 , wherein each random number is at least one of a number, a vector, or a matrix.
3. The computer-implemented method of claim 1 , wherein at least one of the random numbers is generated by a third-party agent.
4. The computer-implemented method of claim 1 , wherein iteratively updating the SLRM is based on a Secret Sharing (SS) scheme.
5. The computer-implemented method of claim 1 , wherein iteratively updating the SLRM comprises:
receiving a notification that a portion of a service process is completed; and
in response to receiving the notification, initiating a second portion of the service process.
6. The computer-implemented method of claim 1 , further comprising:
transmitting, by the first SCN, the first encrypted data to a second SCN of the plurality of SCNs.
7. The computer-implemented method of claim 1 , further comprising:
prior to receiving a plurality of random numbers from the random number provider,
initializing parameters associated with the SLRM.
8. The computer-implemented method of claim 1 , wherein iteratively updating the SLRM continues until a difference between two consecutive iteration results is less than a predetermined threshold.
9. A non-transitory, computer-readable medium storing one or more instructions executable by a computer system to perform operations comprising:
receiving, by a first secure computation node (SCN) of a plurality of secure computation nodes (SCNs), a first random number of a plurality of random numbers from a random number provider;
encrypting, by the first SCN, data stored at the SCN using the first random number to generate a first encrypted data;
iteratively updating, by the first SCN, a secure logistic regression model (SLRM) by using the first encrypted data to generate an interim computation result that is divided into a plurality of shares that are transmitted to the plurality of SCNs to be used as subsequent inputs for the SLRM to determine a column vector of a prediction error as a difference between a sigmoid fitting function applied to a respective share and a label of a sample of a mini batch of a Stochastic Gradient Descent method, each of the plurality of shares hiding original data such that the original data cannot be recovered by using a single share of the plurality of shares; and
after iteratively updating the SLRM, outputting a result of the SLRM, wherein the result is configured to enable a service to be performed by each SCN of the plurality of SCNs.
10. The non-transitory, computer-readable medium of claim 9 , wherein each random number is at least one of a number, a vector, or a matrix.
11. The non-transitory, computer-readable medium of claim 9 , wherein at least one of the random numbers is generated by a third-party agent.
12. The non-transitory, computer-readable medium of claim 9 , wherein iteratively updating the SLRM is based on a Secret Sharing (SS) scheme.
13. The non-transitory, computer-readable medium of claim 9 , wherein iteratively updating the SLRM comprises:
receiving a notification that a portion of a service process is completed; and
in response to receiving the notification, initiating a second portion of the service process.
14. The non-transitory, computer-readable medium of claim 9 , the operations comprising:
transmitting, by the first SCN, the first encrypted data to a second SCN of the plurality of SCNs.
15. The non-transitory, computer-readable medium of claim 9 , the operations comprising:
prior to receiving the plurality of random numbers from the random number provider,
initializing parameters associated with the SLRM.
16. The non-transitory, computer-readable medium of claim 9 , wherein iteratively updating the SLRM continues until a difference between two consecutive iteration results is less than a predetermined threshold.
17. A computer-implemented system, comprising:
one or more computers; and
one or more computer memory devices interoperably coupled with the one or more computers and having tangible, non-transitory, machine-readable media storing one or more instructions that, when executed by the one or more computers, perform one or more operations comprising:
receiving, receiving, by a first secure computation node (SCN) of a plurality of secure computation nodes (SCNs), a first random number of a plurality of random numbers from a random number provider;
encrypting, by the first SCN, data stored at the SCN using the first random number to generate a first encrypted data;
iteratively updating, by the first SCN, a secure logistic regression model (SLRM) by using the first encrypted data to generate an interim computation result that is divided into a plurality of shares that are transmitted to the plurality of SCNs to be used as subsequent inputs for the SLRM to determine a column vector of a prediction error as a difference between a sigmoid fitting function applied to a respective share and a label of a sample of a mini batch of a Stochastic Gradient Descent method, each of the plurality of shares hiding original data such that the original data cannot be recovered by using a single share of the plurality of shares; and
after iteratively updating the SLRM, outputting a result of the SLRM, wherein the result is configured to enable a service to be performed by each SCN of the plurality of SCNs.
18. The computer-implemented system of claim 17 , wherein each random number is at least one of a number, a vector, or a matrix.
19. The computer-implemented system of claim 17 , wherein at least one of the random numbers is generated by a third-party agent.
20. The computer-implemented system of claim 17 , wherein iteratively updating the SLRM is based on a Secret Sharing (SS) scheme.