CNN Based Cyber Defense Architecture for 802.11 Networks

The 802.11 protocol, which forms the basis of Wi-Fi technology, enables computers, smartphones and other devices to connect to each other wirelessly via wireless local area networks (WLAN). In particular, the advent of the Internet of Things (IoT) has rendered Wi-Fi networks an indispensable component of modern connectivity. This situation is characterised by a multitude of security vulnerabilities. This study proposes a convolutional neural network (CNN) based method for the detection of cyber attacks on wireless networks. The performance of the proposed method is evaluated using Wireshark files from the AWID3 (Aegean Wi-Fi Intrusion Dataset) dataset. The following attacks were decomposed into 64 features with Tshark software; deauthentication, disassociation, (re)association, rogue AP, Krack, Kr00k, and evil twin attacks. Additionally, PCAP files belonging to the normal class in the AWID3 dataset were decomposed into 64 features. The most effective 36 features from the 64-feature dataset were selected with the InfoGain attribute evaluation algorithm. Each row of the dataset, comprising 36 features, was transformed into a 6x6 and 1-channel (grayscale) image. In the final stage of the process, the 6x6x1 dimensional images were classified using a convolutional neural network (CNN) deep learning model, resulting in an accuracy rate of 99.93%. Upon examination of the results and graphs obtained in the study, it becomes evident that the proposed model demonstrates remarkable efficacy in classifying the 8-class dataset extracted from the AWID3 dataset. Moreover, the study serves to assess the effectiveness of feature selection in the new dataset comprising 8 classes and 64 features extracted from the AWID3 dataset.