Effective and robust detection of jamming attacks

In this paper we present and evaluate anomaly-based intrusion detection algorithms for detecting attacks at the physical layer of wireless networks, by seeking for changes in the Signal-to-Noise ratio statistical characteristics. Two types of algorithms are proposed: simple threshold algorithms and cumulative sum (cusum) algorithms. Performance evaluation is performed in terms of the detection probability, false alarm rate, detection delay and the robustness of the algorithms to different detection threshold values. The algorithms are applied locally to measurements collected from three locations of an experimental network and under two attack intensities. The results show that the cumulative sum algorithms are more robust and achieve higher performance under both attack intensities. Next, we use the Dempster-Shafer algorithm to fuse the outputs provided by the above locally executed algorithms at different nodes, thus forming a collaborative intrusion detection system. The evaluation shows that the robustness substantially increases while the performance remains high, for both types of attacks.