Intrusion detection for wifi network: A deep learning approach

Abstract

With the popularity and development of Wi-Fi network, network security has become a key concern in the recent years. The amount of network attacks and intrusion activities are growing rapidly. Therefore, the continuous improvement of Intrusion Detection Systems (IDS) is necessary. In this paper, we analyse different types of network attacks in wireless networks and utilize Stacked Autoencoder (SAE) and Deep Neural Network (DNN) to perform network attack classification. We evaluate our method on the Aegean WiFi Intrusion Dataset (AWID) and preprocess the dataset by feature selection. In our experiments, we classified the network records into 4 types: normal record, injection attack, impersonation attack and flooding attack. The classification accuracies we achieved of these 4 types of records are 98.4619 $$\%$$, 99.9940 $$\%$$, 98.3936 $$\%$$ and 73.1200 $$\%$$, respectively.