Analysis of Performance Metrics on the Conjuncture of Intrusions in IEEE 802.11 Networks with Machine Learning at Hospital N.S.C.

Abstract

The security present in IEEE 802.11 networks becomes more relevant every day. However, security on the IEEE     802.11 network has not kept pace with threats with as much significance. For this reason, the proposal arises to design          an Intrusion Detection System-IDS based on machine learning that will be able to have self-improvement, since it will             create a safe environment, capable of detecting all disguised threats, Deauthentication, Eapol -logoff (Eapol) and Beacon Flood, where they were launched on a real corporate network. With this, correlated the performance metrics, and among them, which values the quality of the classification, the Matthews Correlation Coefficient. The Deauthentication anomaly above the Naive Bayes classifier was obtained (88.71%), whereas the quality value of the Logistic Regression (Logistic) classifier was equated to (88.69%), and nevertheless, the J48 presented a lower value  of (88.47%).

Despite this, the identification of the Beacon Flood attack was due to the Naive Bayes algorithm showing the highest detection rate (100.00%), followed by Logistic (99.95%) and J48 having the lowest value            (98.85 %). As a result, in the detection of the Eapol anomaly, the classifications presented similarity of (100.00%) and     the others, with the presentation of a detection, due to non-anomalous data (Normal), the Naive Bayes was affected by             (89.92 % ), followed by Logistic maintaining (89.89%), while J48 was tested with a lower rate (89.67%). With the  study evidences provide the possibility that it is possible to develop an intrusion detection system based on wireless networks.

DOI:https://doi.org/10.56238/Connexpemultidisdevolpfut-116