A SHAP-guided forward feature selection with XGBoost reduces APT-phase detection to 12 features, but the reported 94% macro F1 is likely biased because the test split was used during selection.
APT datasets and attack modeling for automated detection methods: A review,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Explainable AI for Enhancing IDS Against Advanced Persistent Kill Chain
A SHAP-guided forward feature selection with XGBoost reduces APT-phase detection to 12 features, but the reported 94% macro F1 is likely biased because the test split was used during selection.