We consider a wireless system consisting of $K$ legal users, one access point (AP)and one active eavesdropper. The eavesdropper is assumed to attack the system in the uplink phase. Focusing on intrusion detection, we introduce a framework to create datasets that are then put into support vector machine (SVM)classifiers. The characteristics of the three features (i.e., MEAN, RATIO and SUM)in our datasets are formulated from post-processing signals. Based on the three defined features, artificial training data (ATD)is also formed and used at the AP. By training SVM models, we show the high feasibility of detecting active eavesdroppers in many cases. The performance of our proposed approach is evaluated in terms of accuracy and through numerical examples.
|Title of host publication||2019 IEEE Conference on Communications and Network Security (CNS): Workshops: Workshop on Physical-layer Methods for Security and Privacy in 5G and the IoT: Proceedings|
|Publication status||Published - 19 Aug 2019|
|Event||IEEE Conference on Communications and Network Security - Washington, D.C., United States|
Duration: 10 Jun 2019 → 12 Jun 2019
|Conference||IEEE Conference on Communications and Network Security|
|Period||10/06/2019 → 12/06/2019|