ABSTRACT
Distributed Denial of Service (DDoS) attack is known to be the most dangerous attack in cyber-threats which is identically collective in the world of today's interconnected networks, as it mainly disrupts and makes vital services unavailable. Hackers are skilled in using multi- classification DDoS attacks to evade the detection for exploitation of the targeted networks. A proactive and efficient detection mechanism is much needed to secure the networks from multiclass DDoS attacks. However, implementing DDoS detection schemes is very difficult due to the factors such as cost, complexities, and inflexibility. Most of the work has been done to tackle these attacks; still, the open question is to find out an efficient, smart and sustainable model amongst available choices. Therefore, in this research, the author proposes a novel, lightweight Cuda- powered Deep Neural Network Long Short-Term Memory (Cu-DNNLSTM) enabled smart and sustainable threat intelligence system for large and distributed enterprise networks and is simply termed as Multiclass DDoS Detection mechanism (MDDM). The proposed approach is implemented using the current state of the art Canadian Institute of Cybersecurity (CIC)-DDoS2019 dataset that is publicly available and has been thoroughly evaluated. The preliminary results achieves 99.60% detection accuracy with a relatively low ratio of False Positives (FP) (i.e., 0.0003). Additionally, the proposed approach has also been compared with two other DL models to show the promising performance of the proposed approach. Finally, the proposed approach is highly scalable, cost effective, flexible and sustainable to be customized for any emerging computational and communication paradigm.
