Publication

Abstract : Wi-Fi networks have evolved rapidly, resulting in improved performance through increased datarates. In these networks, access points periodically broadcast beacon frames for dynamic network management. However, these frames are typically unprotected and hence can be leveraged by adversaries to forge beacons; thus resulting in numerous attacks such as beacon flooding, deauthentication and DoS attacks, and WPA2 Handshake cracks. Thus, a mechanism for classifying between valid and forged beacons depending on the application requirements is necessary. However, developing such a mechanism is challenging due to the large volume of network traffic coupled with the need for rapid detection of anomalies. In this work, we develop a mechanism for signature-based detection using machine learning for detecting beacon forgeries in Wi-Fi networks. In particular, we leverage supervised learning models to analyze traffic patterns that are further clustered and the generated signatures are classified between normal and abnormal behavior. Finally, we evaluate the proposed signature-based approach in terms of classification accuracy, energy consumption, and execution time of the algorithms on a custom attack dataset generated using Wireshark and on the publicly available NSL-KDD dataset.