Backdoor Attack on STL10
100Attack Success Rate (ASR)InkDrop
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| InkDropCondensation Method=DM, Condensation Backbone=ConvNet2026.03 | 100 | — | — | 58.05 | |
| InkDropCondensation Method=DAM, Condensation Backbone=ConvNet2026.03 | 100 | — | — | 53.5 | |
| DoorpingCondensation Method=DAM, Condensation Backbone=ConvNet2026.03 | 100 | — | — | 53.28 | |
| RelaxCondensation Method=DAM, Condensation Backbone=ConvNet2026.03 | 100 | — | — | 53.48 | |
| RelaxCondensation Method=DM, Condensation Backbone=ConvNet2026.03 | 99.96 | — | — | 59.62 | |
| INACTIVEPre-training Dataset=CIFAR10, Trigger=Invisible (Ours)2024.05 | 99.68 | — | 74.02 | — | |
| DRUPEPre-training Dataset=CIFAR10, Trigger=Patch2024.05 | 96.72 | — | 74.43 | — | |
| RelaxCondensation Method=IDM, Condensation Backbone=ConvNet2026.03 | 95.36 | — | — | 65.82 | |
| InkDropCondensation Method=IDM, Condensation Backbone=ConvNet2026.03 | 93.33 | — | — | 64.76 | |
| DoorpingCondensation Method=IDM, Condensation Backbone=ConvNet2026.03 | 31.44 | — | — | 66.09 | |
| BadEncoderPre-training Dataset=CIFAR10, Trigger=CTRL2024.05 | 16.85 | — | 75.73 | — | |
| DoorpingCondensation Method=DM, Condensation Backbone=ConvNet2026.03 | 14.88 | — | — | 57.73 | |
| NaiveCondensation Method=DM, Condensation Backbone=ConvNet2026.03 | 10.32 | — | — | 62.1 | |
| SimpleCondensation Method=DAM, Condensation Backbone=ConvNet2026.03 | 10.32 | — | — | 53.46 | |
| NaiveCondensation Method=IDM, Condensation Backbone=ConvNet2026.03 | 10.24 | — | — | 66.71 | |
| SimpleCondensation Method=IDM, Condensation Backbone=ConvNet2026.03 | 10 | — | — | 65.76 | |
| BadEncoderPre-training Dataset=CIFAR10, Trigger=WaNet2024.05 | 9.78 | — | 72.73 | — | |
| SimpleCondensation Method=DM, Condensation Backbone=ConvNet2026.03 | 9.64 | — | — | 59.72 | |
| NaiveCondensation Method=DAM, Condensation Backbone=ConvNet2026.03 | 8.76 | — | — | 54.9 | |
| BadEncoderPre-training Dataset=CIFAR10, Trigger=Ins-Xpro22024.05 | 5.91 | — | 74.11 | — | |
| BadEncoderPre-training Dataset=CIFAR10, Trigger=Ins-Kelvin2024.05 | 1.16 | — | 74.89 | — | |
| Clean ModelPre-training Dataset=CIFAR10, Trigger=None2024.05 | — | 76.14 | — | — |