Tripartite optimized framework for network based intrusion detection system

Abstract

An unsecured network pays the way for various illegal access. Therefore, an efficient security system is needed to safeguard the data from attackers. Even though various network security methods such as authentication, authorization, and firewall have been used, the system still suffers from various types of security issues. An optimized soft voting classifier (OSOVOC) algorithm is proposed for network-based intrusion detection system (NIDS). It performs the hybridization of two algorithms namely: optimized decision tree (ODT) and optimized support vector machine (OSVM). All three algorithms are optimized using hyper parameter tuning. The final proposed model produces a more accurate result in an unbalanced dataset The primary focus of the proposed OSOVOC algorithm is to eradicate the false alarms, which significantly increases the true positive count and true negative count. The proposed OSOVOC algorithm produces negotiable false alarms, with performance reaching a F1-score of 100% during training phase and 99.87% during testing phase.