Research Article
Automatic Detection of Android Malware via Hybrid Graph Neural Network
Table 5
Experimental results of different features comparisons.
| Features | Binary-classify | Five-classify | Acc | P | R | F1 | Acc | P | R | F1 |
| Permissions(P)+APIs | 96.38 | 96.41 | 96.50 | 96.45 | 94.30 | 94.28 | 93.98 | 94.13 | Source code ASTs (ASTs) | 89.40 | 89.56 | 88.79 | 89.17 | 86.78 | 87.01 | 86.32 | 86.66 | Source code sequence (S) | 95.21 | 95.33 | 94.20 | 94.76 | 91.42 | 90.60 | 90.46 | 90.53 | ASTs+S | 97.03 | 97.12 | 97.14 | 97.13 | 96.32 | 96.20 | 96.18 | 96.19 | P+APIs+ASTs | 97.31 | 96.95 | 97.10 | 97.02 | 95.01 | 95.10 | 94.69 | 94.89 | P+APIs+S | 99.01 | 98.86 | 98.52 | 98.69 | 98.26 | 98.34 | 98.19 | 98.26 | P+APIs+ASTs+S | 99.62 | 99.67 | 99.50 | 99.58 | 99.20 | 99.15 | 98.89 | 99.01 |
|
|