Anomaly Detection with Various Machine Learning Classification Techniques over UNSW-NB15 Dataset

Abstract

The exponential growth of computers and devices connected to the Internet and the variety of commercial services offered creates the need to protect Internet users. As a result, intrusion detection systems (IDS) are becoming an essential part of each computer-communication system, detecting and responding to malicious network traffic and computer abuse. In this paper, an IDS based on the UNSW-NB15 dataset has been implemented. The results obtained indicate F1 Score and Recall values of 76.1% and 85.3% for the Naive Bayes algorithm, 78.2% and 96.1% for Logistic Regression algorithm, 88.3% and 95.4% for Decision Tree classifier, and 89.3% and 98.5% for Random Forest.

Keywords

• anomaly detection
• intrusion detection system
• machine learning
• network security
• unsw-nb15 dataset