Figure 1.
The Process of Code Standardization
Figure 2.
The Process of Generating the PDG
Figure 3.
The Process of Slicing PDG
Figure 4.
The Process of Extracting Features from the Slice Graph
Figure 5.
Vulnerability detection model architecture
TABLE I.
TABLE TYPE STYLES
| Parameter | Value |
|---|---|
| Loss Function | CrossEntropyLoss |
| Optimization Algorithm | Adam |
| Learning Rate | 0.0001 |
| Weight Decay | 0.001 |
| Batch Size | 16 |
| Training Epochs | 500 |
| Max Steps | 10000 |
Figure 6.
Results of the Model Training in the Proposed Network
Figure 7.
Ablation experiment
Figure 8.
Ablation experiment
Figure 9.
loss comparison
Figure 10.
Performance comparison of different models under evaluation