Table 5 FGSM and PGD attacks after adversarial training.
From: Adversarial susceptibility analysis for water quality prediction models
Model | Clean Accuracy | FGSM (Ɛ=0.1) | FGSM (Ɛ=0.2) | PGD alpha = 0.02 (Ɛ=0.1) | PGD alpha = 0.02 (Ɛ=0.2) |
|---|---|---|---|---|---|
Random Forest | 98% | 40.5 | 43 | 42.95 | 40.95 |
SimpleNN | 74.77 | 68.19 | 59.05 | 65.27 | 64.72 |
