Privacy-preserving machine learning

Abstract

New and efficient protocols are provided for privacy-preserving machine learning training (e.g., for linear regression, logistic regression and neural network using the stochastic gradient descent method). A protocols can use the two-server model, where data owners distribute their private data among two non-colluding servers, which train various models on the joint data using secure two-party computation (2PC). New techniques support secure arithmetic operations on shared decimal numbers, and propose MPC-friendly alternatives to non-linear functions, such as sigmoid and softmax.

Description

Background technique [0001] Machine learning techniques are widely used in practice to generate predictive models for medicine, finance, recommendation services, threat analysis, and authentication techniques. Huge amounts of data collected over time enable new solutions to old problems, and advances in deep learning have enabled breakthroughs in language, image, and text recognition. [0002] Large Internet companies collect users' online activities in order to train recommender systems that predict their future interests. Health data from various hospitals and government organizations can be used to generate new diagnostic models, while financial companies and payment networks can combine transaction history, merchant data, and account holder information to train more accurate fraud detection engines. [0003] figure 1 A high-level diagram depicting a process 100 for training and using a machine learning model is shown. Process 100 begins with training data, shown as exis...

AI Technical Summary

Problems solved by technology

In this setup, the server has full access to the plaintext data, but wants to ensure that the published model cannot be used to infer the data used during training

Images