Adversarial Robustness on CIFAR-10 (L-inf, eps=4/255)
64.91Robust Accuracy (AA)Xiao et al. (2020) ADV
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Xiao et al. (2020) ADVtraining=adversarial2021.02 | 64.91 | 63.56 | -1.35 | 157 | 100 | 1.57 | 179 | |
| Xiao et al. (2020) ADVtraining=adversarial, attack_time_budget_A3=T=302021.02 | 64.91 | 17.7 | -47.21 | 157 | 2,280 | 0.07 | 1,548 | |
| Madry et al. (2018)2021.02 | 44.78 | 44.69 | -0.09 | 25 | 20 | 1.25 | 88 | |
| Guo et al. (2018)2021.02 | 22.3 | 12.14 | -10.16 | 19 | 17 | 1.12 | 99 | |
| Xiao et al. (2020)2021.02 | 19.82 | 11.11 | -8.71 | 49 | 22 | 2.23 | 189 | |
| Metzen et al. (2017)2021.02 | 6.17 | 3.04 | -3.13 | 21 | 13 | 1.62 | 80 | |
| Pang et al. (2019)non-differentiable=true2021.02 | 4.14 | 3.94 | -0.2 | 28 | 24 | 1.17 | 237 | |
| Papernot et al. (2015)2021.02 | 2.85 | 2.71 | -0.14 | 4 | 4 | 1 | 84 | |
| Buckman et al. (2018)non-differentiable=true2021.02 | 2.29 | 1.96 | -0.33 | 9 | 7 | 1.29 | 116 | |
| Das et al. (2017) + Lee et al. (2018)2021.02 | 0.59 | 0.11 | -0.48 | 6 | 2 | 3 | 40 |