Intrusion Detection Classification on UNSW-NB15 (Adversarial Robustness)
95.01Accuracy (Clean)LARAR
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| LARAR2026.05 | 95.01 | 31.45 | 33.18 | 42.51 | |
| Base ADVNNMethod Type=Adversarial training, Loss Function=Composite loss function2026.05 | 94.88 | 27 | 28.7 | 38.29 | |
| Vanilla NNArchitecture=Two-layer MLP, Hidden Units=256 and 128, Training Strategy=Standard cross-entropy loss on clean data only2026.05 | 94.23 | 12.47 | 8.92 | 15.34 |