Abstract
Exponential growth of mobile devices and increased usage of data by mobile users creates a high data traffic problem in cellular networks. Data offloading to Wi-Fi networks provides an alternative to relieve the congestion that occurs in such cellular networks. Wi-Fi Access Point to which data is offloaded must be chosen carefully as there is a possibility of Malicious Access Points (MAPS) in the network, which could trick the users to connect with them instead of Legitimate Access Points (LAPs). Thus, the offloaded data may be used for devious purposes by the MAP which results in a severe security breach like military and defense. Therefore, it is necessary to detect and weed out such MAPs. A normalized K Nearest Neighbors (KNN) algorithm, which is a supervised learning technique, is used to learn from a set of given data pertaining to previous history of Access Points with their characteristic information and decides if the Access Point is malicious or non-malicious. In this paper, we propose a KNN based Evading Malicious Access Point (EMAP) algorithm that identifies MAPs, by using a combination of Round Trip Time (RTT) probes sent and beacon frames received by the user, thus offloading safely to a LAP. The results obtained show that our algorithm has an efficiency of 85% in high as well as low traffic conditions, as compared to lower and variable efficiency of existing proposed methods for identifying MAPs.