Simulation Training Auxiliary Model Based on Neural Network and Virtual Reality Technology

Area of Science:

• Computer Science
• Cybersecurity
• Virtual Reality

Background:

• Training simulators are increasingly software-based, virtualized, and networked.
• Network intrusion poses significant threats to computer and network security.
• Virtual Reality (VR) technologies are maturing, offering new simulation possibilities.

Purpose of the Study:

• To propose a VR-based intrusion detection (ID) simulation training system.
• To develop an improved ID model for enhanced network security training.
• To address challenges with unbalanced datasets in ID systems.

Main Methods:

• Developed a VR-based simulation training system.
• Proposed an ID model using Convolutional Neural Networks (CNN) and Long Short-term Memory (LSTM).
• Implemented data oversampling techniques for unbalanced datasets.
• Utilized 3DSMAX for process visualization and equipment modeling.

Main Results:

• The CNN-LSTM model demonstrated superior performance compared to traditional methods like BP and GA.
• The F1 index, a key evaluation metric, showed significant improvement, especially for D4.
• CNN-LSTM outperformed GA by 12.75% and BP by 14.07% in performance.

Conclusions:

• The VR-based ID simulation training system effectively meets its training objectives.
• The developed CNN-LSTM model significantly enhances intrusion detection capabilities.
• The simulation training approach proves impressive and effective for cybersecurity education.