Optimalisasi Layanan Keamanan Captive Portal Menggunakan Klasifikasi Logistic Regression

Abstract

Privacy has become a major concern with the rapid adoption of various smart devices and internet connections. The randomized MAC (Media Access Control) address for each device was implemented for privacy. Problems arose when implementing randomized MAC addresses on captive portals with connection limitations per user. Random classification by VOUI of the device used to assist the device elimination decisions in the captive portal. MAC address data was obtained from devices connected to the captive portal. The data is processed to be grouped into two separate classes, whether random or not, with four device Mac address threshold and four random percentage threshold. Logistic regression was used to determine the classification with the highest level of accuracy. Of the 16 experiments, it was found that all of them had an accuracy above 92%. The maximum accuracy of 95% was obtained in an experiment using a Mac address threshold of 6 and a random percentage threshold of 50%. This indicates that a value of 6 for the Mac address threshold and a value of 50% for the random percentage threshold can be used for random Mac address classification.