A Hybrid Deep Learning-Powered SDN-Based Intrusion Detection Architecture for Cognitive IoT Security

Abstract

Various sectors, including smart homes, healthcare, transportation, agriculture, and manufacturing, derive significant advantages from Internet of Things (IoT) technology. This innovative method of interacting with the world enhances efficiency, convenience, and productivity. Nonetheless, it raises concerns regarding security, privacy, and data governance. This article introduces a cognitive hybrid Deep Learning (DL) approach, facilitated by Software-Defined Networking (SDN), to address security concerns in IoT networks through intrusion detection. The principal objective of this methodology is to consistently and effectively detect cybersecurity risks within the IoT ecosystem. Inspired by the cognitive computing paradigm, the proposed system can analyze, comprehend, and respond to diverse traffic targeting IoT devices. The suggested model is trained and evaluated utilizing the advanced N-BaIoT and CICDDoS2019 datasets. The experimental results exhibit a high degree of accuracy, producing an acceptable false positive rate alongside a tolerable testing duration. Moreover, the architecture takes into account the constrained resources of IoT devices, guaranteeing they are not excessively taxed during the process. The proposed model achieved an accuracy rate of 99.86%, precision of 99.96%, recall of 99.903%, and F1-Score of 99.93%. The proposed model surpassed existing hybrid DL methods.