A machine learning approach to IEEE 802.11 network intrusion detection
Abstract
Intrusions Detection Systems have been developed to protect wireless local area networks against intrusions; there are two types namely Anomaly-based Intrusions Detection Systems and Signature-based Intrusions Detection Systems. Anomaly-based Intrusions Detection Systems are able to detect novel attacks against a system while Signature-based Intrusion systems which are unable to. Unfortunately, Anomaly-based Intrusion Detection Systems that have been developed have high overall accuracy but low accuracy when detecting the different kinds of intrusions namely flooding attacks, impersonation attacks, and injection attacks. To solve this problem, this project proposes a novel method of combining the feature engineering technique of feature selection through recursive feature elimination with the bagging ensemble learning technique. Evaluation results show an improved performance in terms of detecting the three kinds of intrusions.