IEICE Technical Committee Submission System
Conference Paper's Information
Online Proceedings
[Sign in]
Tech. Rep. Archives
 Go Top Page Go Previous   [Japanese] / [English] 

Paper Abstract and Keywords
Presentation 2023-03-02 14:30
[Poster Presentation] A Study on Eliminating Malicious Node in Federated Learning
Reon Akai, Minoru Kuribayashi, Nobuo Funabiki (Okayama Univ) EMM2022-84
Abstract (in Japanese) (See Japanese page) 
(in English) Federated learning (FL) has been proposed to aggregate deep learning models trained at each node in order to utilize privacy-sensitive data stored separately at each.
However, the presence of malicious nodes among the nodes can degrade the performance of the entire system.
In this study, we investigate a method to prevent performance degradation of the entire system by excluding malicious nodes in federated averaging (FedAvg).
The weight parameters trained at malicious nodes must differ significantly in their statistical characteristics compared to the weight parameters of the normal nodes.
Therefore, the proposed method excludes outliers from the weight parameters received from multiple nodes in the deep neural network (DNN) model.
Our simulations show that this process does not cause much degradation in learning efficiency.
Furthermore, we proposed a method for FedAvg to save the weight parameters each round of training, and to load the file and resume learning, so that the entire system can return to the state before the performance degradation.
Keyword (in Japanese) (See Japanese page) 
(in English) Federated Learning / Deep Learning / Machine Learning / Federated Averaging / / / /  
Reference Info. IEICE Tech. Rep., vol. 122, no. 412, EMM2022-84, pp. 89-94, March 2023.
Paper # EMM2022-84 
Date of Issue 2023-02-23 (EMM) 
ISSN Online edition: ISSN 2432-6380
Copyright
and
reproduction
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034)
Download PDF EMM2022-84

Conference Information
Committee EMM  
Conference Date 2023-03-02 - 2023-03-03 
Place (in Japanese) (See Japanese page) 
Place (in English) Fukue culture hall 
Topics (in Japanese) (See Japanese page) 
Topics (in English)  
Paper Information
Registration To EMM 
Conference Code 2023-03-EMM 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) A Study on Eliminating Malicious Node in Federated Learning 
Sub Title (in English)  
Keyword(1) Federated Learning  
Keyword(2) Deep Learning  
Keyword(3) Machine Learning  
Keyword(4) Federated Averaging  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Reon Akai  
1st Author's Affiliation Okayama University (Okayama Univ)
2nd Author's Name Minoru Kuribayashi  
2nd Author's Affiliation Okayama University (Okayama Univ)
3rd Author's Name Nobuo Funabiki  
3rd Author's Affiliation Okayama University (Okayama Univ)
4th Author's Name  
4th Author's Affiliation ()
5th Author's Name  
5th Author's Affiliation ()
6th Author's Name  
6th Author's Affiliation ()
7th Author's Name  
7th Author's Affiliation ()
8th Author's Name  
8th Author's Affiliation ()
9th Author's Name  
9th Author's Affiliation ()
10th Author's Name  
10th Author's Affiliation ()
11th Author's Name  
11th Author's Affiliation ()
12th Author's Name  
12th Author's Affiliation ()
13th Author's Name  
13th Author's Affiliation ()
14th Author's Name  
14th Author's Affiliation ()
15th Author's Name  
15th Author's Affiliation ()
16th Author's Name  
16th Author's Affiliation ()
17th Author's Name  
17th Author's Affiliation ()
18th Author's Name  
18th Author's Affiliation ()
19th Author's Name  
19th Author's Affiliation ()
20th Author's Name  
20th Author's Affiliation ()
Speaker Author-1 
Date Time 2023-03-02 14:30:00 
Presentation Time 75 minutes 
Registration for EMM 
Paper # EMM2022-84 
Volume (vol) vol.122 
Number (no) no.412 
Page pp.89-94 
#Pages
Date of Issue 2023-02-23 (EMM) 


[Return to Top Page]

[Return to IEICE Web Page]


The Institute of Electronics, Information and Communication Engineers (IEICE), Japan