Abstract:
Privacy has become a big concern for both countries and individuals when using the internet. Many countries and standards committees have created regulations or guidelines addressing these privacy issues. Some companies adapt by implementing random MAC addresses. Recently, some operating systems have made random MAC addresses the default option instead of device MAC addresses. Random MAC address detection is necessary due to problems arising in certain scenarios in captive portals. This research proposes a MAC address classification formula with two threshold variables. Data was taken from the database of devices that successfully logged in to the captive portal. The class whether random or not is determined by the Organizationally Unique Identifier part of the given MAC address of the device. It was challenged with Gaussian Naïve Bayes, Logistic Regression, K-nearest neighbors, and New Support Vector Classification to get the threshold value with the highest accuracy and F1-score. These threshold values are used to replace the variables in the classification formula. The results of the classifiers provide the same accuracy pattern, with accuracy values between 93.7993% and 98.1139%, and F1-score values between 93.8424% and 98.1342%. Gaussian Naïve Bayes produces the optimum both accuracy and F1-score. Random MAC address detection can be implemented in a captive portal.