Automatic Classification of APT Attack Scenarios using Heterogeneous Graph Transformer(HGT)

Heterogeneous APT Pattern Modeling: APT attack scenarios have structurally distinct patterns that homogeneous graph approaches fail to adequately capture.

APT Scenario Auto-Classification Gap: Research on automatic classification of APT attack scenarios is nearly absent.

Heterogeneous APT Pattern Modeling: APT attack scenarios have structurally distinct patterns that homogeneous graph approaches fail to adequately capture.

APT Scenario Auto-Classification Gap: Research on automatic classification of APT attack scenarios is nearly absent.

APT Heterogeneous Knowledge Graph Construction: The study automatically constructs heterogeneous knowledge graphs from APT reports using SecureBERT-based entity extraction and rule-based relation extraction.

HGT-Based APT Scenario Classification: The study classifies APT scenarios using a Heterogeneous Graph Transformer with meta-relation-based attention to learn scenario-specific structural patterns.

APT Heterogeneous Knowledge Graph Construction: The study automatically constructs heterogeneous knowledge graphs from APT reports using SecureBERT-based entity extraction and rule-based relation extraction.

HGT-Based APT Scenario Classification: The study classifies APT scenarios using a Heterogeneous Graph Transformer with meta-relation-based attention to learn scenario-specific structural patterns.

High APT Classification Performance:

Outperforms GNN Baselines:

High APT Classification Performance: