Distributed multi-task learning privacy protection method and system against inference attack

Abstract

The distributed multi-task learning privacy protection method and system against reasoning attacks, through each task node to perform model training based on local data, and realize joint model training by sharing knowledge; the present invention proposes privacy protection model training based on homomorphic cryptography The mechanism enables task nodes to realize multi-task learning model training on the premise of ensuring the privacy of training data, and makes the efficiency of model training independent of the amount of sample data, improving the efficiency of machine learning model training; a model release method based on differential privacy is designed, It can resist identity inference attacks launched by model users when accessing machine learning models. The system includes key generation center, central server, task nodes and model users. The invention can ensure the data privacy of the task nodes during the model training process and after the model release, and promote the large-scale application of multi-task machine learning.

Description

technical field [0001] The invention belongs to the field of information security, and in particular relates to a distributed multi-task learning privacy protection method and system against reasoning attacks, which can be used for collaborative training of multi-task models of large-scale data with different distributions. Background technique [0002] With the development of cloud computing and big data technology, machine learning technology has been applied on a large scale, especially in the fields of image recognition and intelligent speech recognition. The recognition accuracy of machine learning models has exceeded that of the human brain. Machine learning is often based on massive data for model training. However, the training data may come from different data sources, resulting in different distributions of the collected data. Therefore, traditional machine learning model training methods are difficult to be directly applied to multi-data distribution model trainin...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a privacy protection method and system for distributed multi-task learning against inference attacks to ensure privacy protection during the training process of the model in the prior art. Data privacy of task nodes after model release, promoting large-scale application of multi-task machine learning

Images