Enhanced Intrusion Detection for ICS Using MS1DCNN and Transformer to Tackle Data Imbalance
2024

Improving Intrusion Detection in Industrial Control Systems

Sample size: 349346 publication 10 minutes Evidence: high

Author Information

Author(s): Zhang Yuanlin, Zhang Lei, Zheng Xiaoyuan

Primary Institution: School of Artificial Intelligence and Data Science, Hebei University of Technology, Tianjin, China

Hypothesis

Can a dual-channel feature extraction model enhance intrusion detection performance in industrial control systems while addressing data imbalance?

Conclusion

The proposed model achieved an accuracy of 95.11% and an F1 score of 95.12%, significantly outperforming traditional models.

Supporting Evidence

  • The model achieved an accuracy of 95.11% and an F1 score of 95.12%.
  • Hybrid oversampling improved the representation of minority classes.
  • The proposed model significantly outperformed traditional machine learning and deep learning models.

Takeaway

This study created a smart system to catch bad guys trying to break into important machines, making sure it doesn't miss any sneaky attacks.

Methodology

The study used a dual-channel model combining MS1DCNN and WDTransformer, applying SMOTE and Borderline-SMOTE for data balancing.

Potential Biases

Potential bias in the dataset due to the imbalance of attack types.

Limitations

The study may not generalize to all types of industrial control systems or attack scenarios.

Participant Demographics

The dataset included 159,600 normal traffic records and 210,340 attack traffic records from various attack types.

Statistical Information

P-Value

0.0489

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.3390/s24247883

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication