Hybrid machine learning mechanism for intrusion detection systems in network security

Abstract

The rapid evolution of machine learning (ML) has also caused it to pave its way in the development of intelligent intrusion detection systems (IDSs). As such, the work in this paper looks at attacks involved in a communication network and the generation of a dataset for further use in training the ML models. The major contribution of this work is the implementation of a hybrid ML-based IDS. More specifically, the system operates using a combination of isolation forest for anomaly-based IDS (AIDS), random forest for signature-based intrusion detection system (SIDS) and feature selection method. The results demonstrate that the proposed hybrid model provides accuracies of above 98% and which is higher compared to the individual ML models by using a subset of eight features from the generated dataset.