COMBATING MODAL ALIASING IN MPTCP ANOMALY DETECTION WITH ENSEMBLE EMPIRICAL MODE DECOMPOSITION METHOD (EEMD)

Abstract

nternet usage has immensely transformed the traffic and it is important to have improved techniques for conveying data and information. It is later known as Multipath Transmission Control Protocol (MPTCP), which becomes one of the solutions that offers the capability in enhancing the reliability and availability of connections through utilizing several paths for the transmission of data. However, this dynamic characteristic of MPTCP is a challenge about handling of anomalies because of the changes and interferences between the paths. Some traditional approaches such as Empirical Mode Decomposition (EMD) that were applied in prior studies detecting the MPTCP networks always lacks the way of dealing with the modal aliasing that combines the signal components hence reducing its capability of detecting anomalies. This thesis presents an improvement of the EEMD method used to solve the problems of modal aliasing in the process of the MPTCP anomaly detection. The solution given below enhances capability in detecting anomaly by minimizing the addition of modal aliasing; which in turn leads to better signal to noise rates hence better identification of anomalous networks. Algorithms proved through experimental data show that this proposed EEMD-based method is superior to the standard EMD method and can be applied to real-life MPTCP anomaly detection scenarios as an effective solution. The NS-3 setting confirmed EEMD outperformed EMD because it produced 94.6% accurate results alongside 92.3% precision and 95.1% recall. The model shows better anomaly detection through fewer false alarms and it finds transient problems which normal detection methods cannot locate. Therefore, the conclusion reached from this research outlines a more efficient approach towards identification of network anomalies and hence improve the security and reliability of the intended MPTCP based communication networks.