Online training of SVMs for real-time intrusion detection

Author

Zhang, Zonghua ; Shen, Hong

Author_Institution

Graduate Sch. of Inf. Sci., Japan Adv. Inst. of Sci. & Technol., Ishikawa, Japan

Volume

1

fYear

2004

fDate

2004

Firstpage

568

Abstract

To break the strong assumption that most of the training data for intrusion detectors are readily available with high quality, conventional SVM, robust SVM and one-class SVM are modified respectively in virtue of the idea from online support vector machine (OSVM) in this paper, and their performances are compared with that of the original algorithms. Preliminary experiments with 1998 DARPA BSM data set indicate that the modified SVMs can be trained online and the results outperform the original ones with less support vectors (SVs) and training time without decreasing detection accuracy. Both of these achievements benefit an effective online intrusion detection system significantly.

Keywords

authorisation; learning (artificial intelligence); support vector machines; telecommunication security; DARPA BSM data set; intrusion detectors; one-class SVM; online support vector machine; online training; real-time intrusion detection; robust SVM; Computer networks; Computer security; Detectors; Earth; Information science; Intrusion detection; Robustness; Support vector machine classification; Support vector machines; Training data;

fLanguage

English

Publisher

ieee

Conference_Titel

Advanced Information Networking and Applications, 2004. AINA 2004. 18th International Conference on

Print_ISBN

0-7695-2051-0

Type

conf

DOI

10.1109/AINA.2004.1283970

Filename

1283970