Efficient Pearson coefficient calculation method based on third party in federated learning environment

Abstract

The invention relates to a third-party-based efficient Pearson coefficient calculation method in a federated learning environment, in the method, an open-source FATE is selected as an overall calculation communication framework for calculating a Pearson coefficient, two parties participating in feature correlation coefficient calculation are a party A and a party B respectively, and a semi-honest third party is a party C; according to an existing scheme, a large number of homomorphic encryption operations need to be used in order to safely generate Bever Triplets, and in a new scheme, the homomorphic encryption operations can be removed and (a, b, c) Bever Triplets can be safely generated through a semi-honest third party without sacrificing safety, and meanwhile, the two parties respectively obtain an addition secret sharing share of (a, b, c). Due to the fact that a large amount of large integer modular exponentiation operation in Paillier homomorphic encryption in an original scheme does not exist, only tensor dot product and addition and subtraction operation exist, efficiency is greatly improved.

Description

technical field [0001] The invention relates to a third-party efficient Pearson coefficient calculation method in a federated learning environment. Background technique [0002] Here are some basics in this field: [0003] Federated machine learning: The application of privacy computing in the field of machine learning can integrate data from multiple parties without revealing the private data of all parties, and use machine learning algorithms to train models and make predictions. [0004] The security model of federated learning is mostly a semi-honest model. [0005] Feature engineering: A data preprocessing method in machine learning engineering, which screens the feature data of the sample and discretizes the feature data so that a better machine learning model can be trained later. [0006] Pearson coefficient: A method of calculating data correlation, which can be used in feature engineering to calculate the correlation between sample data features, so as to filter ...

AI Technical Summary

Problems solved by technology

[0019] However, this method still uses a large number of Paillier homomorphic public key encryption operations to generate BeaverTriplets (triplets), resulting in inefficient calculation of the overall Pearson correlation coefficient of FATE, which is not practical in the case of large-scale data

For example, when there are tens of millions of data, the calculation task cannot be completed in one day, which makes FATE's data correlation calculation very impractical

Images