A new Distance-based Feature Selection (DFS) approach in KDDCUP99 dataset for anomaly detection with Negative Selection Algorithm (NSA)

A new Distance-based Feature Selection (DFS) approach in KDDCUP99 dataset for anomaly detection with Negative Selection Algorithm (NSA)

Khashayar Khosharay1 Mahdi Mollamotalebi2

1) Department of Computer, Buinzahra Branch, Islamic Azad University, Buinzahra, Iran
2) Department of Computer, Buinzahra Branch, Islamic Azad University, Buinzahra, Iran

IOI Code : XYCC-ABKDF
Publication : International Conference on Science and Engineering(icesconf.com)
Abstract :
    One of the most interesting and challenging task in intrusion detection system is feature selection. In this paper, a new distance-based feature selection approach is proposed in order to reduce the dimensions of KDDCUP99 dataset. One of the main problems of KDDCUP99 dataset is the curse dimensionality. So in order to solve this challengeable problem, feature selection algorithms are used widely. In this research, we use Euclidean distance to measure the distances between normal and abnormal data features. Our proposed feature selection algorithm, leads us to select 11 features from 41 ones. To detect anomaly data, negative selection algorithm is used. The results shows better performance in anomaly detection compared with recent works.

Keywords : feature selection dimensions kddcup99 anomaly detection negative selection