Information: Join today and make your research activities more affordable! Technical workshop participation fees and annual registration fees are available at member rates.
Notice: [Important] Announcement of Changes to Registration Fee Payment and Manuscript Upload Procedures for IEICE Technical Meetings
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 2024-07-25 14:00
Privacy protection of training datasets in CNN transfer learning models
Takumi Katsuie, Kozo Okano, Shinpei Ogata (Shinshu Univ.), Shin Nakajima (NII) SS2024-1 KBSE2024-7
Abstract (in Japanese) (See Japanese page) 
(in English) Transfer learning, one of the machine learning methods, has attracted attention as a technique that can create highly accurate machine learning models with a small amount of training data by using the knowledge of trained models. However, machine learning models have a problem that an attacker can extract the training data. Therefore, DP-SGD was developed as a privacy-preserving machine learning method. In this report, we perform membership inference attacks on four cases of transfer learning models using either or both DP-SGD and SGD to check whether the training data is protected. From the results, we found that when the privacy of the training data of the source model is protected and the transfer learning is performed, the transfer learning does not weaken the privacy protection of the training data of the source model, but the privacy of the training data of the target model is not protected. Therefore, it is necessary to use DP-SGD for transfer learning. In order to efficiently create a transfer learning model with privacy-preserving training data, we found that DP-SGD should be used only for transfer learning.
Keyword (in Japanese) (See Japanese page) 
(in English) Machine learning / Transfer learning / Differential privacy / DP-SGD / Membership Inference Attack / / /  
Reference Info. IEICE Tech. Rep., vol. 124, no. 133, SS2024-1, pp. 1-6, July 2024.
Paper # SS2024-1 
Date of Issue 2024-07-18 (SS, KBSE) 
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 SS2024-1 KBSE2024-7

Conference Information
Committee KBSE SS IPSJ-SE  
Conference Date 2024-07-25 - 2024-07-27 
Place (in Japanese) (See Japanese page) 
Place (in English)  
Topics (in Japanese) (See Japanese page) 
Topics (in English)  
Paper Information
Registration To SS 
Conference Code 2024-07-KBSE-SS-SE 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Privacy protection of training datasets in CNN transfer learning models 
Sub Title (in English)  
Keyword(1) Machine learning  
Keyword(2) Transfer learning  
Keyword(3) Differential privacy  
Keyword(4) DP-SGD  
Keyword(5) Membership Inference Attack  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Takumi Katsuie  
1st Author's Affiliation Shinshu University (Shinshu Univ.)
2nd Author's Name Kozo Okano  
2nd Author's Affiliation Shinshu University (Shinshu Univ.)
3rd Author's Name Shinpei Ogata  
3rd Author's Affiliation Shinshu University (Shinshu Univ.)
4th Author's Name Shin Nakajima  
4th Author's Affiliation National Institute of Informatics (NII)
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 ()
21st Author's Name  
21st Author's Affiliation ()
22nd Author's Name  
22nd Author's Affiliation ()
23rd Author's Name  
23rd Author's Affiliation ()
24th Author's Name  
24th Author's Affiliation ()
25th Author's Name  
25th Author's Affiliation ()
26th Author's Name / /
26th Author's Affiliation ()
()
27th Author's Name / /
27th Author's Affiliation ()
()
28th Author's Name / /
28th Author's Affiliation ()
()
29th Author's Name / /
29th Author's Affiliation ()
()
30th Author's Name / /
30th Author's Affiliation ()
()
31st Author's Name / /
31st Author's Affiliation ()
()
32nd Author's Name / /
32nd Author's Affiliation ()
()
33rd Author's Name / /
33rd Author's Affiliation ()
()
34th Author's Name / /
34th Author's Affiliation ()
()
35th Author's Name / /
35th Author's Affiliation ()
()
36th Author's Name / /
36th Author's Affiliation ()
()
Speaker Author-1 
Date Time 2024-07-25 14:00:00 
Presentation Time 30 minutes 
Registration for SS 
Paper # SS2024-1, KBSE2024-7 
Volume (vol) vol.124 
Number (no) no.133(SS), no.134(KBSE) 
Page pp.1-6 
#Pages 6 
Date of Issue 2024-07-18 (SS, KBSE) 

[Return to Top Page]

[Return to IEICE Web Page]
The Institute of Electronics, Information and Communication Engineers (IEICE), Japan