Improving Intrusion Detection for IoT Networks Using SMOTE and PCA on NSL-KDD Dataset

This paper focuses on detecting anomalies in IoT devices, which is an important component of maintaining their security. To accomplish this, we use the well-known NSL-KDD dataset, which represents real-world IoT network traffic. Our strategy addresses the issue of imbalanced data by employing the SMOTE oversampling technique, which improves our classifier's performance by mitigating data scarcity in the minority class. Following that, we improve our anomaly detection procedure by using Principal Component Analysis (PCA) to identify features. PCA speeds up the subsequent classification processes by lowering data dimensionality while keeping critical information. The Random Forest Classifier, known for its robustness and capacity to handle varied data patterns, is used for the actual classification task. Through extensive experiments, we demonstrate that our proposed method surpasses other existing approaches, achieving an impressive 99.8% accuracy rate in detecting anomalies in IoT devices. Our research presents a promising avenue to enhance IoT security, ensuring safer and more reliable interconnected systems.