Sliding mode observer for anomaly detection in TCP/AQM networks

In recent years, traffic monitoring and especially anomaly estimation have constituted an open problem in the network community. This paper proposes to design an observer-based sliding mode control for time delay systems representing TCP fluid flow model. The main advantage of the sliding observer lies in estimating, in a finite time, system states while exhibiting fundamental robustness against perturbation signals like attacks or more generally anomalies. Fault is then reconstructed from the observer dynamics using smoothing techniques. The feasibility of the proposed method is studied via Simulink.