A novel Approach to LUFlow Intrusion Detection System Using Machine Learning Method

Date
2021
Editor
Journal Title
Volume
Issue
Journal Title
Journal ISSN
Volume Title
Abstract
Recent years have seen significant growth in cybersecurity threats, and the strategies employed by attackers are continuing to improve and become more inventive. Furthermore, the complexity of most datasets, as well as the frequent occurrence of uneven class distributions, emphasizes the need for further study. The goal of this study is to use a variety of strategies for dealing with unbalanced datasets to create an effective intrusion detection system using the most recent LUFlow intrusion detection dataset. LUFlow is a flow-based network intrusion detection data collection that includes a reliable ground truth based on harmful behavior correlation. The efficacy of sampling strategies on LUFlow is thoroughly examined and empirically tested using machine learning classifiers such as Random Forest, Decision Tree, Gradient Boosting, and Ada boost. When dealing with the unbalanced class distribution with fewer samples, the suggested system was able to identify attacks with up to 100% F1-score, making it more practical in real-time data fusion situations that target data categorization. The study of the datasets used for training and testing in the detection model is also important since greater dataset quality may help advance offline intrusion detection. Benchmark datasets such as KDD99 and NSL-KDD Cup 99 are obsolete and have significant flaws, making them unfit for testing anomaly-based network intrusion detection systems. By combining the most popular feature selection methods and classifiers, such as Random Forest Classification (RF), Decision Tree (DT) Classification, Ada Boost (AB), and Gradient Boosting, the CIDDS-001 dataset has an ideal union of feature selection methodologies and classifiers (GB). Gradient boosting, Ada boost, and the decision tree classifier were found to be 99.72 percent, 99.34 percent, and 99.87 percent accurate, respectively.
Description
Keywords
TECHNOLOGY::Electrical engineering, electronics and photonics::Electrical engineering
Citation
Department Name
Electrical and Computer Engineering
Publisher
North South University
Printed Thesis
DOI
ISSN
ISBN