Paper Abstract and Keywords |
Presentation |
2019-03-01 15:45
On Machine Learning Attack Tolerance for PUF-based Device Authentication System Tomoki Iizuka (UTokyo), Yasuhiro Ogasahara, Toshihiro Katashita, Yohei Hori (AIST), Hiromitsu Awano (Osaka Univ.), Makoto Ikeda (UTokyo) VLD2018-133 HWS2018-96 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
Double-Arbiter PUF (DAPUF) and PL-PUF are known to be highly resistant to machine learning attacks.
In this paper, we proposed a deep neural network-based modeling attack for DAPUF and PL-PUF.
Proposed network successfully predicts unknown responses of DAPUF with probability 21.1% higher than the conventional method. Furthermore, we presented that prediction rate of PL-PUF depends on the response bit to be predicted and number of oscillation cycles.
Based on experimental results, we examined the vulnerability of the authentication systems for machine learning attacks considering the environmental variation. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
PUF / hardware security / deep neural network / machine learning attack / modeling attack / / / |
Reference Info. |
IEICE Tech. Rep., vol. 118, no. 458, HWS2018-96, pp. 237-242, Feb. 2019. |
Paper # |
HWS2018-96 |
Date of Issue |
2019-02-20 (VLD, HWS) |
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 |
VLD2018-133 HWS2018-96 |
|