An SDN-based Hybrid-DL-driven cognitive intrusion detection system for IoT ecosystem

Abstract

Smart homes, healthcare, transportation, agriculture, manufacturing, and many more sectors can all benefit from IoT technology. This revolutionary way of interacting with the world could increase efficiency, convenience, and productivity. However, it also raises concerns regarding security, privacy, and data management. To mitigate security threats in IoT networks, this article presents a software-defined networking-enabled cognitive hybrid-deep learning model for intrusion detection in IoT ecosystem. The primary purpose of this framework is to continuously and efficiently identify cybersecurity threats in the IoT environment. With the influence of the cognitive computing paradigm, the designed system can analyse, understand, and even respond to various traffic approaching IoT devices. The proposed model is trained and tested using the cutting-edge N-BaIoT and CICDDoS2019 datasets. The experimental results demonstrate a high level of accuracy, yielding an acceptable false positive rate in conjunction with reasonable testing time. Furthermore, the architecture considers the limited resources of IoT devices, ensuring that they are not overburdened in the process. The proposed model obtained an acceptable accuracy rate of 99. 86 %, 99. 96 % precision, 99. 903 % recall, and 99.93 % F1-Score. The proposed model outperformed the other hybrid-DL models, such as Cu-GRU+LSTM and Cu-GRU+DNN.