A Privacy Preserving Method for Collaborative Deep Learning Model Training

Abstract

The invention discloses a privacy protection method for collaborative deep learning model training, including: proposing a collaborative distributed deep learning model training method, participants use existing training data to calculate the gradient of model parameters locally, and calculate the obtained The gradient data of the model is sent to the parameter server to update the model parameters; a privacy protection mechanism based on the double trapdoor public key cryptography algorithm is proposed, so that the participants can realize the security training of the deep learning model on the premise of ensuring the privacy of their own training data; the detailed design The granular deep learning model release method ensures that only the data owners who participate in the training can obtain the model, ensuring the fairness of model training. The results of the simulation test show that the present invention can provide accurate model training services under the premise of ensuring the data privacy of the participants. It can provide privacy protection for next-generation computer technologies such as artificial intelligence.

Description

technical field [0001] The invention belongs to the field of information security and relates to a privacy protection method, which can be used for collaborative security training of deep learning models in large-scale data. Background technique [0002] Machine learning is becoming a new engine for the development of the digital economy, especially driven by new theories and technologies such as mobile Internet, big data, supercomputing, sensor networks, and brain science, as well as the strong demand for economic and social development, machine learning will further empower All walks of life can promote the in-depth development of the digital economy. As a branch of machine learning, deep learning has also received more and more attention in industry and academia, and is widely used in medical diagnosis, speech recognition, image recognition and other fields. Deep learning is often based on massive data for model training. By analyzing the hidden correlation between data,...

AI Technical Summary

Problems solved by technology

The present invention proposes a privacy protection method for collaborative deep learning model training, which can solve the problem of multi-source data-oriented deep learning model security training, and solve the problems of traditional privacy protection schemes such as high computational overhead and model training accuracy cannot be guaranteed. In order to provide technical support for large-scale security applications of deep learning

Images