An Intrusion Detection System over the IoT Data Streams Using eXplainable Artificial Intelligence (XAI).

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Machine Learning

Background:

• Network and IoT systems face increasing intrusions, necessitating advanced intrusion detection systems (IDS).
• Traditional AI/ML methods in IDSs struggle with data complexity and lack transparency, hindering effective threat prediction.
• Security professionals require explainable models for reliable decision-making in intrusion detection.

Purpose of the Study:

• To propose a novel IDS architecture integrating deep learning (DL) and eXplainable AI (XAI).
• To develop transparent and effective AI-driven models for network and IoT intrusion detection.
• To empower security analysts with understandable predictions for enhanced threat mitigation.

Main Methods:

• Implemented three DL models: 1-D Convolutional Neural Networks (1-D CNNs), Deep Neural Networks (DNNs), and TabNet.
• Trained and evaluated models on seven diverse datasets from the TON_IOT repository.
• Integrated XAI techniques to interpret model predictions and identify key features for classification.

Main Results:

• The 1-D CNN model achieved 99.24% accuracy on the network dataset.
• CNN and DNN models reached 100% accuracy on most IoT datasets.
• TabNet demonstrated the lowest performance across all tested datasets.

Conclusions:

• The proposed DL-based IDS with XAI effectively enhances network and IoT security.
• The developed models provide high accuracy and crucial explainability for security analysts.
• This approach offers a transparent and powerful solution for real-time intrusion detection.