Performance Evaluation of Supervised Machine Learning Based Intrusion Detection with Univariate Feature Selection on NSL KDD Dataset

Abstract

In order to provide exceptional security in networks and secure sensitive and private data, an efficient technique for detecting intrusions is critical nowadays. Due to the rapid expansion of Internet and network technology use, which also accorded to an escalation in the number of attacks, IDS are currently of more interest to researchers. Network intrusion detection (NID) is used to identify network invasions, which is essential for assuring the security of the Internet of Things (IoT) and have become a quintessential element in nearly any security infrastructure. Lately, machine learning algorithms have been used to offer prospective IDS solutions. Intrusion detection is carried out by SVM, kNN, Decision Tree and Logistic Regression using trained attack patterns. Simulation results demonstrate the competence of the proposed detection system to recognize anomalies and sound an alarm. Additionally, feature selection should be incorporated as a preprocessing step to reduce big datasets and enhance accuracy and system performance. In this study, we present an IDS model with feature selection based on univariate selection that works in conjunction with ML based classifiers such as decision tree, SVM, kNN and logistic regression. Using the NSL-KDD data set, we exemplify how our model can outperform conventional ML classifiers in terms of detection rate, precision, recall.

Detection of Distributed Denial of Service Attacks in Software Defined Networks by Using Machine Learning

“…Conversely, False Negatives (FN) denote the number of positive cases that the model wrongly labels as negative. True Negatives (TN) represent the correct model predictions of negative instances, as explicated in [29].…”