Model Inversion Attack on Facescrub
58.6Attack AccuracyLOKT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LOKTTarget Model (T)=FaceNet64, Public Dataset (Dpub)=Facescrub, Design=Sen2023.10 | 58.6 | 1,225.13 | |
| LOKTTarget Model (T)=FaceNet64, Public Dataset (Dpub)=FFHQ, Design=Sen2023.10 | 53.7 | 1,338.67 | |
| LOKTTarget Model (T)=FaceNet64, Public Dataset (Dpub)=Facescrub, Design=S2023.10 | 53.2 | 1,280.7 | |
| LOKTTarget Model (T)=FaceNet64, Public Dataset (Dpub)=FFHQ, Design=S2023.10 | 47.2 | 1,404.85 | |
| LOKTTarget Model (T)=FaceNet64, Public Dataset (Dpub)=Facescrub, Design=CoD2023.10 | 45.7 | 1,296.29 | |
| LOKTTarget Model (T)=FaceNet64, Public Dataset (Dpub)=FFHQ, Design=CoD2023.10 | 44.5 | 1,403.73 | |
| BREPMITarget Model (T)=FaceNet64, Public Dataset (Dpub)=Facescrub2023.10 | 40.2 | 1,236.4 | |
| BREPMITarget Model (T)=FaceNet64, Public Dataset (Dpub)=FFHQ2023.10 | 37.3 | 1,456.59 | |
| MIDREAttack=Standard inversion attack2024.09 | 15.97 | — | |
| MIDRE (Adapt.Att)Attack=Adaptive attack (attacker knows masking portions)2024.09 | 10.5 | — |