Performance Analysis of a Machine Learning Enabled Anomaly Intruder Detector in Wireless Networks
R. Ramyea, Nehru Kasthuri, S Preethi, J. Kavivarman, R Harish, M E Krishnakanth, P N Harish Prabu · 2022 IEEE 6th Conference on Information and Communication Technology (CICT) · 2022
Rapid technological development in wireless networks has offered a smart way of living. Those wireless networks encounter several other security problems which are difficult to ignore. In order to sustain a reliable operation and wireless access in these networks, intruders are to be identified and deleted. A univariate support vector machine (SVM) -based intrusion detection is proposed that uses NSL-KDD dataset with 42 attributes. Among the 42 features, the most important and suitable features are selected by using univariant feature selection. The intruding attacks such as Denial of Service (DoS), Remote to Local (R2L), User to Root (U2R), and probe are considered for the analysis. SVM uses linear kernel function to distinguish between the two classes. The average accuracy rate and precision rate for all the attacks are 98.5% and 97.4% respectively. On an average, recall rate and F measure score is obtained as 94.2%. The adaptive moment (ADAM) optimization technique is employed to reduce the average loss function of 0.038% for 5 epoch rates.