Paper Abstract and Keywords |
Presentation |
2011-03-04 13:25
An anomaly detection method using gamma-divergence and its application to system call sequence Shintaro Murakami, Masanori Kawakita, Jun'ichi Takeuchi (ISIT/Kyushu Univ.) IT2010-98 ISEC2010-102 WBS2010-77 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
Nowadays, damage by malwares is increasing. In this paper, we discuss anomaly detection by machine learning and its application to this problem. When a malware infects a computer, it takes control of the process and system call sequence shows abnormal behaviors. Tatara et. al. proposed a method which detects anomaly by using Markov model. However, we have no data warranted not to be infected, hence training data may involve abnormal data. Moreover, the ratio of abnormal data to normal data may be large, because infected hosts output abnormal system calls period of time. For these reasons, the learning of behaviors of normal sequences requires some robust learning method. For this problem, we employ $\gamma$-divergence, by which it is possible to learn the normal behaviors even if the ration of abnormal data is large. In our study, we apply $\gamma$-divergence to estimate Markov model. We report the results of experiment using real data. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Intrusion Detection / $\gamma$-divergence / Robust Estimation / / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 110, no. 443, ISEC2010-102, pp. 191-197, March 2011. |
Paper # |
ISEC2010-102 |
Date of Issue |
2011-02-24 (IT, ISEC, WBS) |
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 |
IT2010-98 ISEC2010-102 WBS2010-77 |
|