HIDM: Hybrid Intrusion Detection Model for Industry 4.0 Networks Using an Optimized CNN-LSTM with Transfer Learning.

Area of Science:

• Cybersecurity
• Industrial Automation
• Machine Learning

Background:

• Industry 4.0 adoption integrates advanced technologies like AI and IoT, creating new security vulnerabilities.
• Existing intrusion detection systems (IDS) face challenges in securing complex industrial environments.
• The need for robust security solutions is critical for the reliable operation of industrial automation systems.

Purpose of the Study:

• To propose a novel hybrid intrusion detection model (HIDM) for Industry 4.0 environments.
• To enhance intrusion detection accuracy using optimized Convolutional Neural Network-LSTM (OCNN-LSTM) and transfer learning (TL).
• To evaluate the effectiveness of the proposed HIDM against traditional models on real-world industrial datasets.

Main Methods:

• Developed a hybrid intrusion detection model (HIDM) integrating OCNN-LSTM with transfer learning (TL).
• Optimized CNN parameters using the Grey Wolf Optimizer (GWO) for improved prediction accuracy.
• Employed TL to transfer knowledge and enhance the training process of the OCNN-LSTM model.
• Conducted multi-class classification analysis on ToN-IoT and UNW-NB15 datasets.

Main Results:

• The OCNN-LSTM model with TL achieved higher precision: 92.7% on ToN-IoT and 94.25% on UNW-NB15.
• The proposed HIDM demonstrated superior performance compared to OCNN-LSTM without TL, CNN, and LSTM models.
• Key performance metrics including precision, F-measure, recall, accuracy, and detection rate were significantly improved.

Conclusions:

• The hybrid intrusion detection model (HIDM) effectively enhances security in Industry 4.0 environments.
• Transfer learning significantly boosts the accuracy and efficiency of OCNN-LSTM for intrusion detection.
• The proposed model offers a promising solution for detecting threats in industrial automation systems.