Abstract
The increase in connected devices and the constantly evolving methods and techniques of attackers pose a challenge for network intrusion detection systems from conception to operation. As a result, machine learning algorithms are constantly adopted for network intrusion detection systems. A pattern-matching method (signaturebased) is widely used in basic network Intrusion Detection systems (IDS). A more robust method is to use a machine learning classifier to detect anomalies and unseen attacks. However, a single machine learning classifier is unlikely to be able to accurately detect all types of attacks, especially uncommon attacks, e.g., Remote2Local (R2L) and User2Root (U2R), due to a large difference in the patterns of attacks. In this paper, we introduce a -best technique for variable selection and experimented with several classification algorithms, including logistic regression, Decision Tree, and Stochastic Gradient Descent (SGD). Last but not least, we combine model predictions into ensemble predictions to determine whether they will provide us with greater accuracy than the aforementioned techniques. The methods include bagging, boosting, and voting. After utilizing these three, we achieved the best accuracy of 83% for the AdaBoost algorithm from the boosting technique. This accuracy is identical to that of logistic regression.