Targeted Attack on ImageNet-Compatible 10-Targets (all-source)
96CNX RobustnessS4ST-TMI
Evaluation results
| Method | Links | |||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| S4ST-TMI# Data=N/A, From same dataset of victim models (‡)=false, Victim models involved for parameter tuning (♢)=false2024.10 | 96 | 86.2 | 94.5 | 95 | 94.7 | 69.1 | 88 | 84.9 | 80.1 | 66.5 | 95.8 | 68.1 | 99.5 | 18.1 | 28.3 | 77.7 | 23.56 | |
| SASD-WS-DI-TMI# Data=1.2M, From same dataset of victim models (‡)=true, Victim models involved for parameter tuning (♢)=true2024.10 | 88.6 | 85.2 | 92.2 | 93.3 | 91.2 | 58.2 | 70.4 | 68.8 | 61.6 | 47.2 | 95.3 | 88.3 | 99.8 | 12.4 | 20.8 | 71.6 | 23.3 | |
| CFM-RDI-TMI# Data=N/A, From same dataset of victim models (‡)=false, Victim models involved for parameter tuning (♢)=true2024.10 | 88.1 | 81.7 | 89 | 91 | 88.2 | 56.3 | 72.3 | 72.7 | 65.3 | 49.4 | 96 | 79.6 | 99.7 | 17 | 26.5 | 71.5 | 23.55 | |
| H-Aug-TMI# Data=N/A, From same dataset of victim models (‡)=false, Victim models involved for parameter tuning (♢)=true2024.10 | 86.3 | 69 | 84.2 | 82.8 | 82.5 | 42.9 | 72.3 | 65.2 | 57.2 | 40.9 | 80.8 | 36.4 | 92.7 | 5.5 | 8.7 | 60.5 | 23.41 | |
| M3D# Data=50K, From same dataset of victim models (‡)=true, Victim models involved for parameter tuning (♢)=true2024.10 | 84 | 65.9 | 83.5 | 85.9 | 84.4 | 60.8 | 67.7 | 62 | 62.2 | 41 | 88 | 62.9 | 94.6 | 5.2 | 9.2 | 63.8 | 19.2 | |
| ODI-TMI# Data=N/A, From same dataset of victim models (‡)=false, Victim models involved for parameter tuning (♢)=true2024.10 | 74.3 | 62.6 | 73.6 | 75.1 | 70.8 | 37.6 | 53.3 | 55.5 | 42.8 | 32.2 | 78.1 | 45.2 | 96.8 | 6.1 | 9.6 | 54.2 | 23.51 | |
| LRS-DI-TMI# Data=1.2M, From same dataset of victim models (‡)=true, Victim models involved for parameter tuning (♢)=true2024.10 | 72.3 | 55.1 | 61.3 | 72.9 | 60.5 | 45.6 | 47.2 | 50.9 | 41.5 | 35.1 | 85 | 41.5 | 93.3 | 24.5 | 31.7 | 54.6 | 23.5 | |
| IAA# Data=5K, From same dataset of victim models (‡)=true, Victim models involved for parameter tuning (♢)=true2024.10 | 63.2 | 30.7 | 55.2 | 58.7 | 54.6 | 15.9 | 33.2 | 24.6 | 27.5 | 17 | 81.6 | 42.7 | 98.7 | 1.4 | 2.4 | 40.5 | 22.82 | |
| DRA-DI-TMI# Data=1.2M, From same dataset of victim models (‡)=true, Victim models involved for parameter tuning (♢)=true2024.10 | 58.3 | 69.2 | 58.9 | 67.9 | 61.7 | 69.1 | 55.9 | 63.1 | 55.6 | 46.6 | 86.5 | 90.2 | 94.3 | 73.4 | 71.8 | 68.2 | 23.53 | |
| C-GSP# Data=1.2M, From same dataset of victim models (‡)=true, Victim models involved for parameter tuning (♢)=true2024.10 | 57.2 | 20.5 | 47.1 | 48.5 | 46.7 | 22.6 | 26.5 | 19.7 | 22.1 | 9.8 | 50.7 | 17.4 | 72.1 | 1.4 | 1.8 | 30.9 | 15.41 | |
| TTP# Data=50K, From same dataset of victim models (‡)=true, Victim models involved for parameter tuning (♢)=true2024.10 | 54.7 | 38.4 | 56.6 | 57.9 | 60.6 | 30 | 35.6 | 24.4 | 28.6 | 11.4 | 68.7 | 27.8 | 84.1 | 2.9 | 5.4 | 39.1 | 19.87 | |
| SU-DI-TMI# Data=N/A, From same dataset of victim models (‡)=false, Victim models involved for parameter tuning (♢)=true2024.10 | 54.7 | 17.1 | 41.6 | 43 | 33.2 | 6.5 | 19.2 | 18.8 | 11.7 | 10.2 | 43.8 | 11.2 | 89.1 | 0.8 | 1 | 26.8 | 23.52 | |
| DFA# Data=50K, From same dataset of victim models (‡)=true, Victim models involved for parameter tuning (♢)=true2024.10 | 25.3 | 1.6 | 15 | 16.3 | 13.8 | 0.7 | 3.1 | 1.6 | 1.7 | 3.9 | 19.9 | 7.8 | 56.4 | 0.3 | 0.3 | 11.2 | 21.19 |