Building Scalable Anomaly Identification Systems to IoT Threat Mitigation using Machine learning Techniques

Abstract

The quick growth of IoT technologies has led to major cybersecurity issues for detecting abnormal signs that point to security risks or operational problems. The study offers a machine learning technique for identifying irregularities in Internet of Things networks by analyzing key performance indicators, such as packet loss, congestion in the lungs and bandwidth, and latency. The method includes data preprocessing as its first step, followed by SHAP-based feature importance analysis, then classification through Random Forest (RF) and Support Vector Machine (SVM). The analysis included 1000 entries before anomalies were found through the Tukey method and then classified. The experimental data shows SVM performs better than RF, producing accuracy at 96.5% with precision at 95.9% while recall reaches 96.2%, and the resulting F1-score comes out at 96.0%. SVM achieves effective anomaly detection in IoT environments according to comparison results obtained through Logistic Regression and Convolutional Neural Networks (CNN). The research shows that machine learning brings prospective improvements to IoT security because it enables preemptive anomaly detection, which results in better real-time defense capabilities against threats.