Domain Knowledge-Guided Training for NIDS: A Class-Agnostic Evaluation of Robustness on Imbalanced Datasets

Key Points

• This research investigates the effectiveness of domain knowledge-guided training for improving the robustness of network intrusion detection systems on imbalanced datasets.
• Evaluation of existing NIDS approaches using deep learning techniques
• Analysis of performance on imbalanced datasets
• Implementation of domain knowledge-guided training strategies
• Enhanced robustness of NIDS against diverse attack scenarios
• Improved detection rates in imbalanced settings
• Reduced false positive rates compared to traditional methods

Abstract