Machine Learning-Based Intrusion Detection for IoT Networks Using an Optimized Random Forest Classifier
Volume: 15 - Issue: 07 - Date: 26-07-2026
Approved ISSN: 2278-1412
Published Id: IJAECESTU503 | Page No.: 101-110
Author: Goury Vishwakarma
Co- Author: Prof. Ankita Tiwari,Prof. Swati Khanve
Abstract:-The rapid expansion of Internet of Things (IoT) devices across critical sectors has introduced significant security challenges. These devices, marked by constrained computational capacity, diverse communication protocols, and reliance on wireless connectivity, are increasingly vulnerable to sophisticated cyberattacks. Traditional signature- and rule-based intrusion detection systems fall short in IoT settings, as they struggle with zero-day threat identification, produce high false positive rates, and require substantial processing power. This study explores a machine learning-based intrusion detection strategy employing a Random Forest classifier fine-tuned for IoT traffic analysis. The approach includes thorough data preprocessing and taxonomy-based label consolidation. Experimental validation using the "Intrusion in IoT" dataset yields outstanding overall results, achieving a 99.32% test accuracy. Nevertheless, further examination highlights a notable trade-off between global accuracy and recall for minority classes, with particularly poor performance observed for infrequent attack categories. This work presents both an optimized classification pipeline and an empirical assessment of the accuracy–imbalance trade-off.
Key Words:-Internet of Things Security, Network Intrusion Detection, Machine Learning, Random Forest, Class Imbalance.
Area:-Engineering
Download Paper:
Preview This Article
