Performance Comparison of Two Sequential Change Detection Algorithms on Detection of In-band Wormholes
Date: March 18 - March 20, 2009
This paper compares the performance of parametric and non-parametric sequential change detection algorithms for detecting in-band wormholes in wireless ad hoc networks. The algorithms considered are the non-parametric cumulative sum (NP-CUSUM) and the repeated sequential probability ratio test (R-SPRT). Theoretical performance of the two is compared using metrics that take into account the algorithms’ repeated nature, and the advantage of the parametric method is illustrated. On the other hand, connections between the parametric and non-parametric methods are made in the proposed worst-case adversary model, where the non-parametric method is shown to be more robust to attack strategy changes. Experimental evaluation of wormhole detection schemes based on the two algorithms is presented. This work has implications for both the theoretical understanding and practical design of wormhole detection schemes based on parametric and nonparametric change detection algorithms.