Rogue Access Points Detection Based on Theory of Semi-Supervised Learning

It is very dangerous for wireless client to connect with rogue access point. Attackers could eavesdrop or modify client’s information via rogue access point, therefore, rogue access point can be seen as the most serious threats in wireless local area network (WLAN). In this paper, we proposed a novel approach that can detect rogue access points (AP) quickly and accurately. We take advantage of Time-stamp field and signal field in the 802.11 beacon frame as the data in Gaussian distribution algorithm and Native Bayes Classify to generate the fingerprint of access point. The fingerprint is unique to each access point, which cannot be spoofed. In the detection process, we add sliding window and Semi-Supervised Learning, that give our method the ability to take dynamic self-adjustment. Experimental results indicated that the proposed approach could detect rogue access points more quickly and accurately compare with existing methods.