An optimal feature based network intrusion detection system using bagging ensemble method for real-time traffic analysis

The enormous growth of cyber threats has become a calamitous issue in today's technically advanced world where data and information play a crucial role in identifying patterns and automatic predictive analysis. Network packet analysis is a pivotal technique in cybersecurity to protect our network and computer from unauthorized access. A network intrusion detection system (NIDS) is a network packet monitoring tool that intently inspects all the incoming and outgoing packets passing through a network and recognizes malicious incidents. This paper proposes a novel NIDS using the decision tree-based Bagging ensemble method, where the NSL-KDD dataset has been used for experimental purposes. The optimal features from the mentioned dataset have been filtered through the application of the wrapper-based Moth Flame optimization (MFO) technique and the effectiveness of the selected features has been evaluated using various machine learning, deep learning, and ensemble learning frameworks. All the experiments have been conducted in accordance with both binary and multiclass categories. Exhaustive performance evaluation confirms that the proposed MFO-ENSEMBLE method achieves an 87.43% detection rate and incurs minimal time overhead amongst all classification techniques. Practical implementation of the proposed methodology in a custom-built real-time test-bed confirms both the novelty as well as the feasibility of this work.