Analyzing Federated Learning with Enhanced Privacy Preservation

For machine learning in online social networks or mobile environments, data privacy-preserving is a very important issue. In the past, artificial intelligence and machine learning algorithms that operating on a stand-alone machine must analyze each data in order to build an accurate model when train...

Full description

Saved in:
Bibliographic Details
Main Authors: Tseng, Sheng-Po, Yeh, Lo-Yao, Wu, Lee-Chi, Tsai, Pei-Yu
Format: Conference Proceeding
Language:English
Subjects:
Analytical models
Collaborative work
Data privacy
Federated learning
Machine learning
Machine learning algorithms
Privacy-preserving
Smart contracts
Social networking (online)
Training
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:For machine learning in online social networks or mobile environments, data privacy-preserving is a very important issue. In the past, artificial intelligence and machine learning algorithms that operating on a stand-alone machine must analyze each data in order to build an accurate model when training a model. However, this method is insufficient in terms of privacy protection, because it must collect data from different sources; on the other hand, if these algorithms analyze only a part of the data, the accuracy of the models constructed by artificial intelligence and machine learning algorithms will be very low. To solve the above problems, this study starts from the privacy protection aspect of industrial data management, combined with the concept of federated learning, and hopes to improve the accuracy of algorithm modules while ensuring data privacy. Therefore, in this paper, we analyze the impact of federated learning under different protocol mechanisms on the accuracy of algorithm modules, and based on this, we can apply this module to integrate blockchain smart contract technology in the future.
ISSN:2375-0324
DOI:10.1109/MDM55031.2022.00097