On the Robustness of Root Locus based Routing Attack-Detection in Reconfigurable Wireless Networks

Data routing is one of the most essential functions having a direct impact on network performance. Due to the lack of any fixed infrastructure, data routing must be achieved through collaboration of nodes in Reconfigurable Wireless Networks (RWN). This implies that a malicious node can participate in the routing process and launch an attack to gain control over data traffic. Intrusion Detection Systems (IDS) are typically used to recognize malicious activities. Any IDS for routing attack-detection in RWN should be able to detect a wide variety of attacks and must have into consideration the computing and power restrictions that some nodes could have. A mathematical framework for a root locus based IDS for routing in RWN has been presented in recent literature achieving promising attack-detection accuracy for a case study. This root locus based IDS consists in modeling the dynamic behavior of neighboring nodes at a given instant as a linear system. In this paper, we analyze through simulations the attack-detection performance of that root locus based IDS for the worst case, where the number of delayed samples of the signals used to characterize the system, is minimum. We show that even for these severely restricted models, the analyzed attack-detection scheme is robust enough to be used in a wide range of network conditions. We analyze the method performance for a wide variety of routing attacks, the effects of node mobility, attacker's position and different node densities.