RETRACTED: ICN intrusion detection method based on GA-CNN.

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Industrial Control Systems

Background:

• Industrial control system networks face significant data theft risks from attacks like SQL injection.
• Existing security measures struggle to adequately protect industrial control systems in global communication environments.

Purpose of the Study:

• To propose an advanced network intrusion detection method for industrial control systems.
• To enhance the security and data integrity of industrial networks against sophisticated cyber threats.

Main Methods:

• Utilized genetic algorithm (GA) for data optimization and feature selection.
• Developed a hybrid model combining a one-dimensional multi-scale convolutional neural network (CNN) with a gated recurrent unit (GRU).
• Implemented GA for optimizing feature selection to identify critical data subsets.

Main Results:

• The proposed model achieved training and testing in approximately 8 seconds, outperforming other methods (around 10 seconds).
• Demonstrated high detection rate (96.97%), low packet loss rate (1.256%), and minimal false alarm rate (0.0947%).
• Achieved an intrusion defense success rate exceeding 90%.

Conclusions:

• The novel method offers superior intrusion detection performance and generalization ability for industrial control systems.
• The hybrid CNN-GRU model effectively addresses time-series data dependencies and mitigates gradient issues.
• The approach meets the critical network intrusion detection needs of industrial control environments.