The possibility of programming the control and data planes, enabled by the Software-Defined Networking (SDN) paradigm, represents a fertile ground on top of which novel operation and management mechanisms can be fully explored, being Distributed Denial of Service (DDoS) attack detection based on machine learning techniques the focus of this work. To carry out the detection, this paper proposes ORACLE: cOllaboRation of dAta and Control pLanEs to detect DDoS attacks, an architecture that promotes the coordination of control and data planes to detect network attacks. As its first contribution, this architecture delegates to the data plane the extraction and processing of traffic information (with per flow granularity). This is done to ease the calculation and classification of the feature set used in the attack detection, as the needed flow information is already processed when it arrives at the control plane. Besides, as the second contribution, this architecture breaks the limitations to calculate some features that are not possible to implement in a traditional OpenFlow-based environment. In the evaluation of ORACLE, we obtained up to 96% of accuracy in the testing phase, using a K-Nearest Neighbors model.

Available at: https://ieeexplore.ieee.org/abstract/document/9464076