Deep Learning Based Distributed Denial of Service Attack Detection and Mitigation in the Software Defined Network

Abstract

Todays modern world is enhanced by various technological advancements. The development of smart components paved the way for smarter things in various arenas, which made a great impact on the improvement of traditional networks named Software Defined Networking(SDN). The main feature of SDN is its programmability and this enables the network to function via programs. Apart from the benefits, the emerging SDN network is mustered up with several limitations too. Such limitations are network security, scalability, performance, interoperability and reliability. Among them, the major limitation is SDN network security, since the attacks like DDoS attacks can halt the total network. This research work is solely worked on the detection of DDoS attacks in SDN networks. The impact of Distributed Denial of Service (DDoS) attack mainly depends on the scale of the attack which results in the loss of necessary services, increased remediation cost, loss of productivity and reputational damage. Thereby, a reliable attack detection technique is necessary to counteract DDoS attacks. Accordingly, four novel detection techniques are contributed to this thesis. newline As the first contribution, review on various SDN-DDoS attack detection methodologies are reviewed and also introduced a detection model worked with Bidirectional Gated Recurrent Units (Bi-GRU) classifier which is trained by statistical and flow-based features. Here, the Bi-GRU classifier is optimally tuned by the Self-Improved Honey Badger Algorithm (SI-HBA). newline The second research contribution is based on attack detection and mitigation processes. Whereas, the detection process is evolved with an optimized DNN, where the weight is optimized by the Self Improved Moth Flame Optimization (SIMFO) approach during the training process. Also, proposed a Bait based mitigation process to remove the attacker from the network. newline The third research contribution relies on DDoS attack detection model in SDN network with a deep 2-layer classification model via Mouse Updated Aquila Expl