Towards DDoS-aware Routing in SDN using Cross-layer Knowledge Transfer

Routing packets efficiently is the main purpose of today’s networking. Mainly, efficiency deals with how securely are packets delivered taking in consideration the exponential growth of network threats, particularly the Distributed Denial of Service (DDoS) Attacks. However, security and routing are separately addressed in literature and little work was devoted to security-aware routing.In this paper, we point out the need to merge security and routing in SDN instead of addressing them separately. We propose a new DDoS-aware routing scheme using a cross-layer knowledge transfer between flows and switches. This scheme consists of learning features from flow-based DDoS attack datasets then transferring knowledge to switches. Transferred knowledge is used to establish reliable paths between end-points that avoid suspicious switches. Mainly, evaluation of a switch risk is based on a computation of a Suspicion Risk Factor (SRF) using a set of features learned from an ML-based classification of flows. Finally, SRFs of switches are used to seek for DDoS-aware paths. Evaluation results and analysis show that the proposed scheme outperforms the baseline protocol OSPF in terms of finding the most secure path and offered throughput.