Hybrid Ant-Baby Optimizer and BiLSTM framework for high-performance IoT intrusion detection.

Area of Science:

• Cybersecurity
• Network Security
• Artificial Intelligence

Background:

• Internet of Things (IoT) networks face significant challenges in intrusion detection due to high-dimensional data, class imbalance, and evolving attack patterns.
• Existing methods often struggle with efficiency and interpretability when dealing with complex IoT traffic.
• Effective intrusion detection is crucial for securing the rapidly expanding landscape of connected devices.

Purpose of the Study:

• To propose a novel hybrid intrusion detection framework for IoT networks.
• To enhance detection accuracy, computational efficiency, and interpretability in intrusion detection systems.
• To address the limitations of high-dimensional data and class imbalance in IoT network traffic.

Main Methods:

• A hybrid framework combining the Ant-Baby Optimizer (ABO) for feature selection and a Bidirectional Long Short-Term Memory (BiLSTM) neural network for classification.
• The ABO algorithm employs mutual information for selecting informative features, reducing dimensionality and improving efficiency.
• Selected features are processed by BiLSTM to capture temporal dependencies in network traffic data.

Main Results:

• Achieved superior detection accuracy of 97.32% on the CICIoT2023 dataset.
• Demonstrated balanced per-class precision, recall, and F1-scores, with a macro F1-score of 0.9732.
• Exhibited low inference latency (0.219 ms/sample) and high processing speed (>4,500 samples/sec).

Conclusions:

• The proposed hybrid framework is effective and outperforms state-of-the-art methods in IoT intrusion detection.
• The framework offers a balance of performance, efficiency, interpretability, and adaptiveness for real-world applications.
• Future work includes cross-domain validation, attention mechanisms, and online learning for enhanced zero-day attack detection.