Synergistic Approaches to Enhance IoT Intrusion Detection: Balancing Features through Combined Learning

Abstract

The Internet of Things (IoT) plays a crucial role in ensuring security by preventing unauthorized access, malware infections, and malicious activities. IoT monitors network traffic as well as device behaviour to identify potential threats and take appropriate mitigation measures. However, there is a need for an IoT Intrusion Detection system with enhanced generalization capabilities, leveraging deep learning and advanced anomaly detection techniques. This study presents an innovative approach to IoT IDS that combines SMOTE-Tomek link and BTLBO, CNN with XGB classifier which aims to address data imbalances, improve model performance, reduce misclassifications, and improve overall dataset quality. The proposed IoT IDS system, using the IoT-23 dataset, achieves 99.90% accuracy and a low error rate, all while requiring significantly less execution time. This work represents a significant step forward in IoT security, offering a robust and efficient IDS solution tailored to the changing challenges of the interconnected world.

Keywords

• min-max normalization
• smote-tomek link
• btlbo algorithm
• cnn with xgb
• adam optimizer