Table 2 Ablation study of GIT-GuardNet components.
From: Graph-augmented multi-modal learning framework for robust android malware detection
Model variant | Accuracy | F1-Score | AUC-ROC | Precision | Recall |
|---|---|---|---|---|---|
Full GIT-GuardNet | 99.85 | 99.85 | 99.94 | 99.89 | 99.81 |
No Graph Module | 98.94 | 98.91 | 99.23 | 98.85 | 98.97 |
No Temporal Module | 98.61 | 98.57 | 98.96 | 98.66 | 98.49 |
No Cross-Attention Fusion | 97.73 | 97.68 | 97.90 | 97.60 | 97.76 |
Static Only (Transformer) | 96.98 | 96.87 | 97.88 | 96.73 | 97.02 |
