Membership Inference Attack on TinyImageNet (MIA Score, Performance Gap)
87.7MIA ScoreSALUN
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SALUNUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=30%, Backbone=ResNet2026.04 | 87.7 | 0.3 | |
| RLUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=30%, Backbone=ResNet2026.04 | 87.5 | 0.1 | |
| RetrainingUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=30%, Backbone=ResNet2026.04 | 87.4 | 0 | |
| SCRUBUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=30%, Backbone=ResNet2026.04 | 87.4 | 0 | |
| IEU w/GAUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=30%, Backbone=ResNet2026.04 | 87.3 | 0.1 | |
| FTUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=30%, Backbone=ResNet2026.04 | 87 | 0.3 | |
| IEU w/NoisyUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=30%, Backbone=ResNet2026.04 | 86.9 | 0.4 | |
| IEU w/GA+NoisyUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=30%, Backbone=ResNet2026.04 | 86.9 | 0.5 | |
| SALUNUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=50%, Backbone=ResNet2026.04 | 84.2 | 1.1 | |
| RLUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=50%, Backbone=ResNet2026.04 | 83.3 | 0.2 | |
| RetrainingUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=50%, Backbone=ResNet2026.04 | 83.1 | 0 | |
| SCRUBUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=50%, Backbone=ResNet2026.04 | 83.1 | 0 | |
| FTUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=50%, Backbone=ResNet2026.04 | 83 | 0.1 | |
| IEU w/GAUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=50%, Backbone=ResNet2026.04 | 83 | 0.2 | |
| IEU w/NoisyUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=50%, Backbone=ResNet2026.04 | 82.9 | 0.2 | |
| IEU w/GA+NoisyUnlearning Scenario=Random Data Forgetting, Forgetting Ratio=50%, Backbone=ResNet2026.04 | 82.8 | 0.3 |