Paper Abstract and Keywords |
Presentation |
2022-03-08 10:00
Input predictive attack by keyboard acoustic emanations using BERT and its countermeasures Masahiro Iida (Teikyo Univ.), Mitsuaki Akiyama (NTT), Masaki Kamizono (DTCY), Takahiro Kasama (NICT), Yuichi Hattori (Secure Cycle Inc.), Hiroyuki Inoue (Kyoto Sangyo Univ.), Atsuo Inomata (Osaka Univ.) ICSS2021-67 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
The Keyboard Acoustic Emanations has been proposed to estimate the input key from keystroke sounds as a kind of side-channel attack. In the previous study, it was necessary to estimate word by word using dictionary information from multiple input key candidates included at the top of the estimation result. Therefore, although it can be estimated on a word-by-word basis, it was difficult to estimate correctly on a sentence-by-sentence basis.
Also it has only constructed and evaluated the estimation model assuming the input of English sentences, so that the effectiveness of the model for languages other than English has not been clarified.
In this study, we propose a method to improve the estimation accuracy for natural language texts by applying the BERT model, a natural language processing model to Keyboard Acoustic Emanations in order to take into account the connections between words. Furthermore we executed some evaluation experiments and showed that our method is able to estimate both English and Japanese input texts with more high accuracy. In addition, we discuss effective countermeasures based on the findings of this experiments. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Keyboard Acoustic Emanations / BERT / Side-channel Attacks / / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 121, no. 410, ICSS2021-67, pp. 49-54, March 2022. |
Paper # |
ICSS2021-67 |
Date of Issue |
2022-02-28 (ICSS) |
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 |
ICSS2021-67 |
|