We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
An Optimized Network Intrusion Detection System for Attack Detection based on Supervised Machine Learning Models in an Internet-of-Things Environment.
- Authors
Alhomoud, Adeeb
- Abstract
In this paper, an optimized classification approach based on a support vector machine (SVM) classifier is proposed to maximize the accuracy of a machine learning model employed by a network intrusion detection system to detect malicious attacks in an Internet-of-Things (IoT) environment. In addition, an experiment study based on the TON_IoT dataset is conducted in terms of classifier performance metrics, such as the false positive rate, true positive rate, precision, F-measure, recall, Matthews correlation coefficient, receiver operating characteristic area, precision-recall curve area, and a confusion matrix. It is demonstrated that the model accuracy is maximized and the number of rounds minimized for the confusion matrix to converge and correctly classify all various attacks. Finally, classifier errors are compared in terms of kernel types. This part of the investigation shows that the SVM classifier based on a polynomial kernel outperformed radial basis function and normalized polynomial kernels in terms of classifier errors, time to build the model, correctly classified instances, incorrectly classified instances, and kappa statistics.
- Subjects
MACHINE learning; SUPERVISED learning; INTRUSION detection systems (Computer security); RADIAL basis functions; RECEIVER operating characteristic curves; SUPPORT vector machines
- Publication
International Journal of Advances in Soft Computing & Its Applications, 2023, Vol 15, Issue 2, p1
- ISSN
2710-1274
- Publication type
Article
- DOI
10.15849/IJASCA.230720.01