A Loss Tomography Algorithm in Wireless Sensor Networks Using Gibbs Sampling

End-to-end application data in wireless sensor networks can be a valuable health indicator, if they can be used for network measurement purposes. This paper therefore applies network tomography technology to identify lossy nodes using end-to-end application traffic. Based on the path information piggybacked by data packets and the end-to-end performance observations, the problem of lossy nodes inference is modeled as a Bayesian inference problem and a Markov Chain Monte Carlo (MCMC) algorithm using Gibbs sampling was proposed. The algorithm is evaluated via simulation and achieves high detection and low false positive rates.