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 2021-03-04 14:15
[Poster Presentation] Detection of Adversarial Examples in CNN Image Classifiers Using Features Extracted with Multiple Strengths of Filter
Akinori Higashi, Minoru Kuribayashi, Nobuo Funabiki (Okayama Univ.), Huy Hong Nguyen, Isao Echizen (NII) EMM2020-70
Abstract (in Japanese) (See Japanese page) 
(in English) Deep learning has been used as a new method for machine learning, and its performance has been significantly improved. Adversarial examples are known as attacks to machine learning system by injecting malicious noise to inputs such as images, sounds, videos so as to fool the system. Research on fooling image classifiers has been reported as a potential threat to CNN-based systems. In this paper, we propose a new method for detecting adversarial examples by using the sensibilities of image classifiers. Since adversarial examples are generated by adding noise, we focus on the behavior of the output of the image classifier to the noise removal filter. We change the strength of the noise removal filter and observe its output to determine whether it is an adversarial example or not. With the increase of the filter strength, the entropy of the image is expected to decrease and adversarial noises are removed as well. Therefore, the output of the softmax function of the image classifier is expected to change significantly in the case of adversarial examples, while it is stable in the case of normal images. A framework for detecting simple adversarial examples by using the response characteristics to noise removal operations. We conducted experiments against typical adversarial example generating attacks and quantitatively evaluated its performance.
Keyword (in Japanese) (See Japanese page) 
(in English) Adversarial Example / Noise Removal Filter / CNN / Image Classifier / / / /  
Reference Info. IEICE Tech. Rep., vol. 120, no. 418, EMM2020-70, pp. 19-24, March 2021.
Paper # EMM2020-70 
Date of Issue 2021-02-25 (EMM) 
ISSN Print edition: ISSN 0913-5685  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 EMM2020-70

Conference Information
Committee EMM  
Conference Date 2021-03-04 - 2021-03-05 
Place (in Japanese) (See Japanese page) 
Place (in English) Online 
Topics (in Japanese) (See Japanese page) 
Topics (in English) Image and Sound Quality, Metrics for Perception and Recognition, Human Auditory and Visual System, etc. 
Paper Information
Registration To EMM 
Conference Code 2021-03-EMM 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Detection of Adversarial Examples in CNN Image Classifiers Using Features Extracted with Multiple Strengths of Filter 
Sub Title (in English)  
Keyword(1) Adversarial Example  
Keyword(2) Noise Removal Filter  
Keyword(3) CNN  
Keyword(4) Image Classifier  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Akinori Higashi  
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 Huy Hong Nguyen  
4th Author's Affiliation National Institute of Informatics (NII)
5th Author's Name Isao Echizen  
5th Author's Affiliation National Institute of Informatics (NII)
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
Date Time 2021-03-04 14:15:00 
Presentation Time 15 
Registration for EMM 
Paper # IEICE-EMM2020-70 
Volume (vol) IEICE-120 
Number (no) no.418 
Page pp.19-24 
#Pages IEICE-6 
Date of Issue IEICE-EMM-2021-02-25 


[Return to Top Page]

[Return to IEICE Web Page]


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