Fig. 2
From: Graph-augmented multi-modal learning framework for robust android malware detection
The proposed GIT-GuardNet model integrates static features, call graph structures, and temporal behavioral patterns through three specialized encoders–Transformer-based Static Encoder, Graph Attention Network (GAT), and Temporal Transformer. These modality-specific representations are fused via a Cross-Attention Fusion module, followed by a neural classifier for final malware prediction.
