DuRBIN: A Comprehensive Approach to Analysis and Detection of Emerging Threats due to Network Intrusion

Network Intrusion Detection System (NIDS) is a popular software application that scans and monitors for suspicious network traffic and alerts when a potential threat is identified. Although NIDS has its' advantages, there are scenarios pertaining to a busy or large network, encrypted data information, and fake addresses that result in frequent false positives and other shortcomings. It is imperative to understand and compare different machine learning algorithms used in NIDS to address these shortcomings and find a better approach for analysis and detection. This research contributes specifically focused on finding that better approach by comparing different machine learning algorithms, such as Deep Learning, Distributed Random Forest, Gradient Boosting Machine, and Naïve Bayes, and demonstrate the eligibility, applicability, and appropriateness of each algorithm with numerical values and graphs. 'LUFlow' Network Intrusion Detection Data Set was used to conduct the research. Additionally, to aid in visualizing the algorithm outcomes based on malicious networks, benign networks, and outliers, a 'RShiny' interface is developed where users can see different plots for different variables of concern, check predictions, track model metrics, and have a model summary. The derived results of this research can help in the methodological approach of implementing various machine learning algorithms in NIDS while addressing various existing limitations.