Reducing U2R and R2L category false negative rates with support vector machines
The KDD Cup '99 is commonly used dataset for training and testing IDS machine learning algorithms. Some of the major downsides of the dataset are the distribution and the proportions of U2R and R2L instances, which represent the most dangerous attack types, as well as the existence of R2L attac...
Saved in:
Published in | Serbian journal of electrical engineering Vol. 11; no. 1; pp. 175 - 188 |
---|---|
Main Authors | , |
Format | Journal Article |
Language | English |
Published |
Faculty of Technical Sciences in Cacak
2014
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | The KDD Cup '99 is commonly used dataset for training and testing IDS machine
learning algorithms. Some of the major downsides of the dataset are the
distribution and the proportions of U2R and R2L instances, which represent
the most dangerous attack types, as well as the existence of R2L attack
instances identical to normal traffic. This enforces minor category detection
complexity and causes problems while building a machine learning model
capable of detecting these attacks with sufficiently low false negative rate.
This paper presents a new support vector machine based intrusion detection
system that classifies unknown data instances according both to the feature
values and weight factors that represent importance of features towards the
classification. Increased detection rate and significantly decreased false
negative rate for U2R and R2L categories, that have a very few instances in
the training set, have been empirically proven.
nema |
---|---|
ISSN: | 1451-4869 2217-7183 |
DOI: | 10.2298/SJEE131007015M |