EHFOA-ID: An Enhanced HawkFish Optimization-Driven Hybrid Ensemble for IoT Intrusion Detection

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Internet of Things (IoT)

Background:

• IoT environments face complex intrusion detection challenges due to high-dimensional data and diverse attack vectors.
• Existing methods struggle with class imbalance and efficient feature space navigation.

Purpose of the Study:

• To propose EHFOA-ID, an enhanced intrusion detection framework for IoT.
• To address challenges of high-dimensional traffic, heterogeneous attacks, and class imbalance.

Main Methods:

• Utilizes an Enhanced HawkFish Optimization Algorithm (EHFOA) for joint feature selection and hyperparameter tuning.
• Employs a hybrid deep ensemble learning model to capture spatial, temporal, and contextual relationships.
• Integrates adaptive exploration-exploitation, Lévy flight, and diversity preservation for optimization.

Main Results:

• Achieved >99% accuracy on UNSW-NB15 and >98% on SECOM datasets.
• Reported macro-F1 scores above 0.97 and false alarm rates below 2%.
• Demonstrated superior performance over state-of-the-art intrusion detection approaches.

Conclusions:

• EHFOA-ID offers a robust and efficient solution for IoT intrusion detection.
• The unified optimization-learning pipeline enhances generalization and reduces feature redundancy.
• The framework effectively improves decision reliability against diverse intrusion patterns.