Optimizing IoT Security: A Machine Learning Pipeline for Fast and Accurate Intrusion Detection

After Industrial Revolution 4.0, there has been a drastic increase in IoT devices, proportionally affecting network traffic exponentially. This requires the advancement of a more efficient anomaly-detecting system, a technique to identify and counteract cyber threats. Although several systems have been published, they lack generalizability to real-time testing. This paper presents a network intrusion detection and prevention system that utilizes machine learning and ensemble learning algorithms that are trained to ensure a highly efficient and robust anomaly detection system. In this paper, we implement a three-stage prediction pipeline. The effectiveness of the proposed system is determined by calculating performance metrics such as accuracy, recall, precision, F1 score, and inference time and comparing them with those of other deep learning models. Furthermore, the proposed method is capable of classifying the various anomalies into 6 broad groups such as DDoS, DoS, Mirai, Web-based attacks, etc. The generalizability of this system is also proven by deploying it in a real-time IoT network environment for classification and evaluation.