Bio-Inspired Metaheuristic Algorithm for Network Intrusion Detection System of Architecture

Abstract

By identifying different kinds of attacks and application misuse that firewalls normally aren't able to identify, network intrusion detection systems (IDS) are intended to keep computer networks safe. When creating a network intrusion detection system, feature selection techniques are crucial. Several bionic meta-heuristic algorithms are used to quickly categorize network traffic as problematic or normal, then decrease features to demonstrate higher accuracy. Thus, in order to detect frequent attacks, this research proposes a hybrid model of network intrusion detection system (IDS) based on an algorithm inspired by a hybrid bionic element. There are two goals for the suggested model. The first step is to minimize the number of features that are chosen in Network IDS. By combining biosensing metaheuristics with hybrid models, this objective is accomplished. The algorithms used in this chapter are particle swarm optimization (PSO), multiverse optimizer (MVO), grey wolf optimization (GWO), moth flame optimization (MFO), firefly algorithm (FFA), whale optimization algorithm (WOA), bat algorithm (BAT), genetic bee colony (GBC) algorithm, artificial bee colony algorithm (ABC), fish swarm algorithm (FSA), cat swarm optimization (CSO), artificial algae algorithm (AAA), elephant herd optimization (EHO), cuckoo search optimization algorithm (CSOA), lion optimization algorithm (LOA), and cuttlefish algorithm (CFA) algorithm. Using machine learning classifiers, the second objective is to identify frequent attacks. SVM (support vector machine), C4.5 (J48) decision trees, and RF (random forest) classifiers are used to accomplish this purpose. Thus, the goal of the suggested model is to pinpoint frequent attacks. The data indicates that J48 is the top classifier when it comes to model building time when compared to SVM and RF. The data indicates that when it came to feature reduction for classification, the MVO-BAT model decreased the features to 24, whereas the MFO-WOA and FFA-GWO models lowered the accuracy, sensitivity, and F-measure of all features to 15. The accuracy, sensitivity, and F-measure of each feature are the same for every classifier.