AI-Driven Threat Intelligence for IoT Networks: Leveraging Machine Learning for Enhanced Intrusion Detection

As Internet of Things (IoT) devices continue to spread, they also create many new entry points for cyberattacks. Traditional security methods struggle to keep up, which makes smarter and more adaptive defenses necessary. This paper introduces an Artificial Intelligence (AI)–driven threat intelligence framework designed to improve intrusion detection in diverse IoT networks. The framework combines Machine Learning (ML) and Deep Learning (DL) models to detect malicious activity more accurately across different types of network traffic. To evaluate the approach, three widely used benchmark datasets—UNSW-NB15, CIC-IDS2017, and IoT-Botnet—were used. Experimental results show that the proposed hybrid Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) model performs very well. It achieved 97% accuracy, a 0.95 F1-score, and a 0.98 Receiver Operating Characteristic – Area Under the Curve (ROC-AUC) on the UNSW-NB15 dataset, outperforming traditional ML models such as Random Forest, which reached 94% accuracy. While DL models provided better detection performance and stronger generalization, ML models proved to be much faster, with nearly three times lower inference latency—about 3 milliseconds per network flow. This makes them more suitable for real-time deployment at the IoT edge, where computing resources are limited. Overall, the proposed hybrid approach strikes a practical balance between detection accuracy and processing speed, offering a scalable and robust foundation for AI-based IoT threat intelligence in real-world environments.