SOTA Model Inversion Attack benchmarks and papers with code

Benchmarks

Dataset Name	SOTA Method	Metric
VGGFace2	PPA	Top-1 Acc93.88	35	3d ago
CelebA (private) and FFHQ (public) (test)	KEDMI + LOMMA	Attack Accuracy77.27	24	2d ago
CelebA	LOKT	Attack Acc93.93	16	3d ago
Standard MI attack datasets (CelebA, Pubfig83, FFHQ, Facescrub)	LOKT	Queries (M)12.16	15	3d ago
CASIA	SMILE	Acc@180.95	12	3d ago
VGGFace	PPA	Top-1 Accuracy97.96	11	3d ago
CelebA (test)	KEDMI	Attack Accuracy74	10	2d ago
Facescrub	LOKT	Attack Accuracy58.6	8	3d ago
Pubfig83	LOKT	Attack Accuracy94.4	8	3d ago
CelebA (private) FFHQ (public) on VGG16 (test)	GMI + LOMMA	Top-5 Attack Accuracy86.33	8	2d ago
CelebA private FFHQ public on face.evoLve (test)	KEDMI + LOMMA	Top-5 Attack Accuracy94	8	2d ago
CelebA (private) FFHQ (public) on IR152 (test)	KEDMI + LOMMA	Top-5 Attack Accuracy92	8	2d ago
Model Inversion Attack Query Complexity Statistics	RLBMI	Query Steps (Step 1)0	5	3d ago
CelebA	VMI + LOMMA	Top-5 Attack Accuracy91.02	4	2d ago
CelebA 128x128 resolution (test)	LOKT	KNN Distance1,320.16	4	3d ago
Facescrub (private identities)	KED-MI	Attack Accuracy0.48	4	3d ago
Pubfig83 (private identities)	BREP-MI	Attack Accuracy0.66	4	3d ago
CelebA private identities	KED-MI	Attack Accuracy82	4	3d ago
MNIST	VMI	Accuracy0.95	4	3d ago
CXR 128x128 (test)	VMI	Accuracy69	4	3d ago
CelebA 64x64 (test)	VMI	Accuracy0.55	4	3d ago
MNIST original (test)	VMI	Accuracy0.95	3	3d ago

Showing 22 of 22 rows