Enhancing IoT Security with IDSIoT2024 Dataset: A Three-Tier Intrusion Detection Framework
Manasa Koppula, L. M. I. Leo Joseph · 2025
The rise of IoT devices has increased security vulnerabilities, requiring strong solutions against cyber threats. We propose a cutting-edge Three-Tier Intrusion Detection System (TT-IDS) trained using real-time IoT network data to address IoT security challenges. The proposed methodology is significant as it combines a hybrid feature selection method with a deep autoencoder for feature extraction in the first tier. This is followed by incorporating five Tree-Based Algorithms with careful hyperparameter tuning in the second tier, and a finely tuned stacking ensemble method in the third tier. The TT-IDS proposal demonstrated exceptional effectiveness, with a notable $99.84 \%$ training accuracy and $99.52 \%$ testing accuracy for multi-class classification across 12 classes, including normal data. This innovative approach not only pushes the boundaries of intrusion detection but also underscores the vital significance of the proposed methodology in delivering a robust and efficient solution for strengthening IoT ecosystems against a wide range of security threats.