LoGD-ai: An Efficient Network Intrusion Detection System Using a Soft Voting-Based Ensemble Learner

Abstract Network attacks detection is still a difficult task because of the large data volumes needed for training a cutting-edge machine learning algorithms to unearth network invasions. Recently, a number of data mining methods have been put out for network intrusion detection. Meanwhile, each of them encounters certain difficulties resulting from the sophisticated nature of threats the existing models cannot detect. The NSL-KDD dataset containing four different form of attacks was employed in this paper. In this paper, we demonstrate the efficiency of LoGD-ai, a soft voting-based ensemble learner as network intrusion detection system to classify network data to normal and malicious. This soft-voting based ensemble classifier employs logistic regression, gradient boosting and decision tree algorithms for intrusion detection. Finally, the performance of the LoGD-ai classifier is compared to the popular gradient boosting machine (GBM), random forests (RF), and AdaBoost. Scoring 98.05% on accuracy, the LoGD-ai classifier performs better than GBM, RF and AdaBoost classifiers, efficiently detecting and classifying network traffic as normal or malicious. Compared to gradient boosting, our proposed methodology increases accuracy by 0.52%, which is critical for network intrusion detection.