Jammer Detection based on Artificial Neural Networks: A Measurement Study

Gecgel S., Goztepe C., Karabulut Kurt G. Z.

ACM Workshop on Wireless Security and Machine Learning (WiseML), Florida, United States Of America, 15 - 17 May 2019, pp.43-48 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1145/3324921.3328788
  • City: Florida
  • Country: United States Of America
  • Page Numbers: pp.43-48
  • Istanbul Technical University Affiliated: Yes


Wireless networks are prone to jamming attacks due to the broadcast nature of the wireless transmission environment. The effect of jamming attacks can be further increased as the jammers can focus their signals on reference signals of the transmitters, to further deteriorate the transmission performance. In this paper, we aim to jointly determine the presence of the jammer, along with its attack characteristics by using neural networks. Two neural network architectures are implemented; deep convolutional neural networks and deep recurrent neural networks. The presence of jammer and the transmitter and the type of the jammer is determined through a diverse set of scenarios that are implemented on software defined radios using orthogonal frequency division multiplexing based signaling. To improve the detection performance, prepossessing techniques are applied. Test results show that the proposed approach can effectively detect and classify the jamming attacks with around 85% accuracy.