An optimized model for network intrusion detection systems in industry 4.0 using XAI based Bi-LSTM framework.

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Industrial Networks

Background:

• Industry 4.0 environments face increased cyber threats due to limited resources and network heterogeneity.
• These threats pose significant financial and reputational risks, including sensitive data theft.
• Existing security measures struggle with the diverse nature of industrial networks.

Purpose of the Study:

• To develop a novel intrusion detection system (IDS) for Industry 4.0 networks.
• To enhance the precision and interpretability of intrusion detection.
• To improve the security and privacy of industrial networking systems.

Main Methods:

• A Bidirectional Long Short-Term Memory based Explainable Artificial Intelligence (BiLSTM-XAI) framework was developed.
• Data preprocessing involved cleaning and normalization.
• Feature selection was performed using the Krill Herd Optimization (KHO) algorithm.
• SHAP and LIME explainable AI algorithms were utilized for result interpretation.

Main Results:

• The BiLSTM-XAI framework demonstrated superior performance in intrusion detection.
• The system achieved a high classification accuracy of 98.2% in experiments.
• The use of XAI algorithms improved the interpretability of prediction results.

Conclusions:

• The proposed BiLSTM-XAI framework offers a precise and interpretable solution for detecting intrusions in Industry 4.0.
• This approach significantly enhances security and privacy within industrial networks.
• The study highlights the effectiveness of advanced AI techniques in addressing modern cybersecurity challenges.