Smart Cyber Defense: Machine Learnig Powered Intrusion Detection in 802.11 Networks

Abstract

Wireless networks have become an indispensable element of our digital world. From our cell phones to our smart home systems, from our workplaces to public transportation, we rely on wireless connections everywhere. This technology, which overcomes the physical limitations of wired connections, offers theeasiest way to access information. The efficiency and ease of the 8 0 2. 1 1 protocol contribute to the extensive adoption of Wi-Fi networks. The widespread use of Wi-Fi networks also brings many security vulnerabilities. In this study, a Wireless Attack Detection System (Wi-ADS) dataset has been developed to detect cyber attacks on Wi-Fi networks. The dataset, which includes attacks that occur in wireless networks as well as attacks encountered in daily life in both wireless and wired networks, is a comprehensive dataset consisting of Normal, Krack, Evil Twin, Deauthentication, Disassociation, Rogue AP, MiTM, Smurf and SYN_Flood attack classes. The common features identified for all data were subjected to classification with Random Forest (RF), K Nearest Neighbor (KNN), Light Gradient Boosting Machine (LGBM), Decision Tree (DT) and Support Vector Machine (SVM) machine learning (ML) models. RF and LGBM classifiers, which are the most successful in default settings, were subjected to parameter tunning to improve their performance. As a result of the optimal hyper parameters, the LGBM classifier achieved an accuracy of 9 6. 3 9 %. The obtained result demonstrates the success of the proposed Wi-ADS dataset and the optimization process in detecting cyber threats in wireless networks. © 2024 IEEE.