Secure Three-Party Linear Regression Without Accuracy Loss
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing privacy-preserving computation methods for linear regression, such as differential privacy, homomorphic encryption, secure multi-party computation, and federated learning, face issues of reduced accuracy, high computational overhead, and increased communication time, failing to provide dual protection for data and models.
Innovation Solution
A privacy-preserving computation method and system for secure three-party linear regression using secure two-party and three-party matrix hybrid multiplication and inversion protocols, ensuring data privacy by allowing participants to compute without accessing intermediate results, thus improving accuracy and reducing communication time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If differential privacy is applied by adding noise to data or objective function, then privacy protection is achieved, but model accuracy decreases and usability is reduced
Solution Approach 1:
The patent introduces secure multi-party computation protocols as intermediaries between data owners and the regression modeling process. Instead of directly adding noise to data (differential privacy), the system uses secret sharing and secure computation protocols that allow the regression model to be trained on encrypted/shared data without revealing individual data points, thus achieving privacy protection without sacrificing model accuracy
Solution Approach 2:
The patent segments the regression computation process into multiple secure computation steps: secure matrix multiplication, secure matrix inversion, and secure linear combination operations. Each step is performed on secret-shared data, allowing the complete regression pipeline to execute while maintaining privacy without requiring noise addition that would degrade accuracy
2Reliability
If homomorphic encryption is used for ciphertext computations, then privacy security is enhanced, but computational overhead increases exponentially and computation speed decreases
Solution Approach 1:
The patent replaces expensive homomorphic encryption operations with more efficient secure multi-party computation primitives based on secret sharing. The secret sharing scheme uses lightweight field operations and simple communication protocols rather than complex cryptographic transformations, dramatically reducing computational overhead while maintaining security
Solution Approach 2:
The patent substitutes the cryptographic mechanism of homomorphic encryption with a different security mechanism based on information-theoretic secret sharing and secure computation protocols. This replacement eliminates the exponential computational overhead of homomorphic encryption while achieving the same privacy protection goal through mathematical secrecy rather than cryptographic transformation
3Reliability
If secure multi-party computation protocols are used for distributed regression computation, then input privacy and computational accuracy are ensured, but communication time increases significantly with more participants
Solution Approach 1:
The patent merges the secure computation of matrix multiplication and matrix inversion into an integrated protocol for computing the regression coefficients. By combining these operations and optimizing the communication patterns, the system reduces the number of separate communication rounds required compared to executing standard secure multi-party computation protocols for each operation separately
Solution Approach 2:
The patent optimizes the communication efficiency by changing the parameters of the secure computation protocol, specifically by using optimized secret sharing schemes and communication patterns that minimize the number of interaction rounds. The protocol is tailored to the specific structure of the linear regression computation, allowing for more efficient parameter selection and communication optimization
Data Source
AI summary
Disclosed is a privacy-preserving computation method and system for secure three-party linear regression, relating to the technical field of privacy-preserving computation. The method includes: processing, by two participants, two private transposed matrices and two private matrices by using a 2PHMP to obtain ∂1 and ∂2; processing, by the two participants, ∂1 and ∂2 by using a 2PIP to obtain u1 and u2; splitting, by the first participant, a first private transposed matrix into v1 and Δ; obfuscating and overlaying, by the second participant, a second private transposed matrix with Δ to obtain v2; and processing v2, u1, u2, v1, and a third private matrix by using a 3PHMP to obtain a regression coefficient matrix.