Robust estimator for loss tomography

Network tomography has been considered the most promising methodology in network measurement in particular for the characteristics of a large network. A large number of methods have been proposed to estimate various characteristics, including link-level loss rates, delay distribution, and so on. However, the robustness of the estimators has been very much ignored that may affect the applicability of the estimators in practice. This paper aims to correct this by introducing robust procedures to the estimators developed for loss tomography, where the estimators proposed previously are analyzed under the scrutiny of robustness, such as the breakdown point of the estimators. Apart from that, a number of estimators are proposed here to overcome the weakness of previous estimators.