Distributed intrusion detection scheme using dual-axis dimensionality reduction for Internet of things (IoT)

The immense growth in the cyber world has given birth to various types of cybercrimes in the Internet of things (IoT). Cybercrimes have breached the multiple levels of cybersecurity that is one of the major issues in the IoT networks. Due to the rise in IoT applications, both devices and services are prone to security attacks and intrusions. The intrusion breaches the data packet extracted from different nodes deployed in the IoT network. Most of the intrusive attacks are very near variants of previously marked cyberattacks containing many repetitive data and features. And to detect the intrusion, the data packet needs to be analyzed. This article presents a novel scheme, i.e., dual-axis dimensionality reduction, that utilizes Kalman filter and salp swarm algorithm (coded as KF-SSA) for analyzing and minimizing the data packet. The proposed data reduction scheme is utilized with KELM-based multiclass classifier to efficiently detect intrusion in the IoT network (KF-SSA with KELM). The proposed method’s overall results are evaluated using standard intrusion detection datasets, i.e., NSL-KDD, KYOTO 2006+ (2015), CICIDS2017, and CICIDS2018 (AWS). The result from the proposed data reduction technique obtains highly reduced data, i.e., 70.% for NSL-KDD and 86.43% for CICIDS2017. The analyzed result shows high detection accuracy of 99.9% for NSL-KDD and 95.68% for CICIDS2017 with decreased computational time.