A Semi-Supervised Learning Approach to IEEE 802.11 Network Anomaly Detection

With the remarkable development of Wi-Fi network, network security has become a key concern over the years. In order to face the increasing number of wireless network intrusion activities, an effective intrusion detection system is necessary. In this paper, a deep learning approach based on ladder network which self-learns the features necessary to detect network anomalies and perform attack classification accurately was proposed. And using focal loss as a loss function to enhance the discriminative ability of the model to classify difficult samples. In experiments on Aegean Wi-Fi Intrusion Dataset (AWID) public data-set, the network records was classified into 4 types: normal record, injection attack, impersonation attack, flooding attack. This paper achieved the classification accuracies of these four types of records are 99.77%, 82.79%, 89.32%, 73.41% respectively, and achieved an overall accuracy of 98.54%.