Leveraging Smart Sentry to Detect and Mitigate Cyber Threats in Industrial IoT Networks

The Internet of Things (IoT) has revolutionized digital connectivity but has simultaneously broadened the landscape for cyber threats. This paper focuses on categorizing IoT attacks by integrating multiple machine learning (ML) models and deep learning (DL) techniques. The research presents a binary and multiclass classification approach utilizing algorithms like Random Forest, Decision Tree, Extra Trees Classifier, Support Vector Machine, k-Nearest Neighbors, and a Deep Neural Network. Experiments were conducted using the Edge-IIoTset dataset, reflecting real-world IoT threat scenarios. Pre-processing involved Principal Component Analysis (PCA) for dimensionality reduction, Synthetic Minority Over-sampling Technique (SMOTE) to counteract class imbalance, and data normalization. The study provides a comparative analysis of model performance, showing that the DNN model achieved outstanding results—100% accuracy for binary classification, 96.15% for 6-class classification, and 94.68% for 15-class classification. A 10-fold cross-validation was also implemented to ensure model generalization. The findings offer valuable insights for improving the security posture of IoT environments through intelligent detection mechanisms.