Evaluation of Supervised Machine Learning Algorithms for Intrusion Detection in Wireless Network using KDDCUP’99 and NSLKDD datasets

Abstract

The Security of computers is a fundamental part of present-day life. The security principles remain the same whether a computer is a single node or a stand-alone system in a large corporate network. Since the nature of wireless networks is to communicate with unknown devices it presents unique security challenges and, they are more vulnerable to attackers. If this property of wireless networks is disabled by completely nullifying the amount of flexibility provided by this communication medium then no new hosts can join the network. Intrusion Detection System (IDS) becomes an interesting topic in research and particularly in the machine learning field for computer network communities. The quality of the data collected from the network traffic is an important aspect of this research area. Most importantly the majority of current IDS are data-driven. The intrusion detection experiments for this paper are conducted using the most popular datasets KDDCUP’99 and its derivative NSL-KDD, to improve the existing classification methods. Various Machine Learning (ML) classifiers are trained using these two datasets and their performance is recorded to conduct a vigorous collation of both.