Enhanced Intrusion Detection for IoT Networks Using Machine Learning Approach

Iot devices (Internet of Things) have experienced explosive growth, creating major security vulnerabilities due to hardware limitations and heterogeneous design profiles. This paper presents an innovative intrusion detection solution for IoT ecosystems based on advanced machine learning techniques applied to the TON_IoT dataset. By merging ensemble-based classification with strategic dimension reduction methods, our framework simultaneously elevates detection precision and minimizes computational overhead. We performed comprehensive testing across the diverse TON_IoT dataset, incorporating numerous attack vectors and IoT configurations. Our solution demonstrates exceptional performance with $\mathbf{9 8 . 7 \%}$ accuracy, $97.5 \%$ precision, and $96.8 \%$ recall metrics, substantially exceeding benchmark approaches. The paper also contributes a new mathematical formulation for optimizing feature extraction through combined information gain assessment and correlationbased filtering. Empirical results confirm our approach offers robust protection for IoT infrastructures against sophisticated and emerging security threats.