A framework for efficient network anomaly intrusion detection with features selection

An intrusion Detection System (IDS) provides alarts against intrusion attacks where a traditional firewall fails. Machine learning algorithms aim to detect anomalies using supervised and unsupervised approaches. Features selection techniques identify important features and remove irrelevant and redundant attributes to reduce the dimensionality of feature space. This paper presents a features selection framework for efficient network anomaly detection using different machine learning classifiers. The framework applies different strategies by using filter and wrapper features selection methodologies. The aim of this framework is to select the minimum number of features that achieve the highest accuracy. UNSW-NB15 dataset is used in the experimental results to evaluate the proposed framework. The results show that by using 18 features from one of the filter ranking methods and applying J48 as a classifier, an accuracy of 88% is achieved.