Improving the Routing Process in SDN Using a Combination of the Evidence Theory and ML

Abstract

Today, Denial of Service (DoS) attacks are recognized as a major jeopardy for network performance and generally lead to distrusted paths. Henceforth, Trustworthiness has gain more concerns from researchers which focus on protecting communications by introducing new secure routing approaches. Particularly in SDN, trust-driven secure routing is recognized as a promising solution for providing security efficiently in future networks.

This paper proposes a trust-based routing approach as an alternative to Shortest Path First (SPF). It aims to forward packets through trusted routes in SDN. It is based on the classification of each switch either as benign, malicious or uncertain node according to a trustworthiness score computed by aggregating multi-side recommendations using the Dempster-Shafer (D-S) combination rule. The classification is achieved by a Multi-class Support Vector Machine (MC-SVM) model which aims to build a trust graph that is used to find the most trusted path joining a source to a destination using a trust-driven cost. Simulation results and analysis show that the proposed scheme performs better than the SPF-based baseline routing process in terms of the end-to-end throughput and the end-to-end delay.