CNN-based Intrusion Classification for IEEE 802.11 Wireless Networks

WiFi has been widely deployed to facilitate home, office, or even stadium-scale wireless access to the Internet, and will be an essential part of the future wireless network through being integrated with 5G cellular networks. However, security threats are still a big concern for WiFi due to the open share nature of wireless medium and the easy-access to WiFi intrusion tools. Therefore, it plays a critical role to identify the experienced attacks for applying further intrusion handling methods. However, existing intrusion detection methods based on traditional machine learning technique usually bears low detection accuracy and need a lot of human intervention. In this backdrop, this paper presents a convolutional neural network (CNN)-based intrusion detection algorithm for WiFi networks. Firstly, we present the detection framework, which first conducts data pre-processing, and then a CNN is trained for attack identification. To reduce both the risk of data overfit and the network training time, Dropout technique is adopted and different network structures are investigated. Experimental results on an open data set named AWID have shown that our algorithm greatly outperform existing ones, and its identification accuracy is higher than 99%.