Distributed Ensemble Method Using Deep Learning to Detect DDoS Attacks in IoT Networks

The widespread adoption of Internet of Things (IoT) devices has increased exponentially in recent years. Consequently, the security risks and vulnerabilities related to these unsecured IoT devices are also continuously increasing. Among the significant challenges facing the IoT environment is the threat of Distributed Denial of Service (DDoS) attacks. Several solutions are available in the literature to detect DDoS attacks. However, these detection mechanisms can easily be evaded by attackers using advanced tools and techniques, posing difficulty in detecting such lethal attacks in real time. Therefore, this paper proposes a novel distributed ensemble method for detecting lethal IoT traffic-based DDoS attacks. This method comprises two key stages: first, developing a distributed ensemble method using the breathtaking capabilities of the H2O.ai distributed machine learning platform and the ensemble learning technique. Secondly, this method was deployed on the Apache Storm stream processing framework, to swiftly analyze incoming network streams and categorize them into eleven distinct classes, including benign traffic and ten types of attacks, in near real time. The proposed method accurately identifies specific target categories within a multi-attack classification scenario by utilizing the expertise of various models. Ultimately, the prediction for a target class is determined based on the model with the highest detection rate. The effectiveness of this method has been examined using different configured scenarios. The experimental results show that our method can identify various attack categories more accurately with 99%+ accuracy and 8.45 s quicker than non-ensemble methods.