Optimized intrusion detection for IoT networks using Cauchy-Gaussian hybrid evolutionary feature selection

Area of Science:

• Cybersecurity and Network Engineering
• Machine Learning Applications

Background:

• Internet of Things (IoT) networks are vulnerable to cyberattacks due to the sensitive data they handle.
• Intrusion Detection Systems (IDS) are crucial for IoT security, but developing low-complexity systems remains a challenge.
• Existing intrusion classification methods often struggle with the scale and complexity of IoT environments.

Purpose of the Study:

• To propose a novel, low-complexity Intrusion Detection System (IDS) method for the Internet of Things (IoT).
• To address the challenge of building efficient and effective intrusion detection models for resource-constrained IoT devices.
• To enhance the performance and reduce the complexity of IDS through active feature selection and ensemble machine learning.

Main Methods:

• A novel Cauchy-Gaussian genetic-arithmetic optimizer-driven variance-based active feature selection method is introduced.
• The method employs a two-phase approach: KD-tree representation for active sample learning based on feature variance, followed by optimizer-driven feature selection.
• Ensemble machine learning, specifically the Bagging algorithm, is utilized with the selected features for intrusion classification.

Main Results:

• The proposed method achieved high accuracy rates of 99.88% (CICIDS2017) and 99.72% (IoTID20) using the Bagging algorithm.
• A significantly low false positive rate was observed: 0.000801 (CICIDS2017) and 0.000165 (IoTID20).
• The active feature selection approach demonstrably reduced the complexity of wrapper methods compared to conventional optimization techniques.

Conclusions:

• The novel active feature selection method combined with ensemble learning offers a highly effective and low-complexity solution for Intrusion Detection Systems in IoT.
• The Cauchy-Gaussian genetic-arithmetic optimizer enhances feature selection diversity and convergence, outperforming traditional methods.
• The validated results on benchmark datasets confirm the method's superiority in accuracy and efficiency for IoT security.