Hybrid GWQBBA model for optimized classification of attacks in Intrusion Detection System

In order to keep computer networks secure, Network Intrusion Detection Systems (IDS) look for a variety of threats and illegal application usage that firewalls miss. When building efficient network intrusion detection systems, the feature selection method is crucial. Numerous bio-inspired metaheuristic algorithms were exercised to make the process of identifying abnormal or normal network traffic more efficient with less features and improved accuracy in a shorter amount of time. As a result, this work aims to propose a network Intrusion Detection System (IDS) model that employs hybridization of bio-inspired metaheuristic algorithms in order to identify the generic attack. Among the two goals of the suggested model is the simplification of Network IDS feature selection. This goal was accomplished by combining different types of bioinspired metaheuristic algorithms Grey Wolf Optimization Algorithm and Quantum Binary Bat Algorithm into a single model. The subsequent objective is to use machine learning classifiers to identify the generic attack and assess the optimality of the selected features. By utilizing the Naive Bayes, K Nearest Neighbor and Random Forest (RF) classifiers, this goal was accomplished. To evaluate the suggested hybrid model, the UNSW-NB15 dataset was employed. In light of the findings, the GWQBBA model successfully reduced the number of features used for classification to 12 while maintaining high levels of accuracy, sensitivity and F-measure across the entire spectrum. The accuracy measure of GWQBBA using Random Forest classifier is obtained as 98.5%.