AI Paper Atlas

Research map / Federated learning and privacy / Differential privacy

SecretGen: Privacy Recovery on Pre-trained Models via Distribution Discrimination

Zhuowen Yuan, Fan Wu, Yunhui Long, Chaowei Xiao, Bo Li

ECCV 2022 · Federated learning and privacy › Differential privacy · approach: Differentially private synthetic data · 12 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add cutout transformations to the discrimination metric (ablation)SecretGen · CelebA, face.evoLVe, center maskattack accuracy (Protocol 1): 0.528 -> 0.550Improved
add cutout transformations to the discrimination metric (ablation)SecretGen · CelebA, face.evoLVe, face T maskattack accuracy (Protocol 1): 0.256 -> 0.273Improved
add cutout transformations to the discrimination metric (ablation)SecretGen · CelebA, VGG16, center maskattack accuracy (Protocol 1): 0.584 -> 0.599Improved
add cutout transformations to the discrimination metric (ablation)SecretGen · CelebA, ResNet152, center maskattack accuracy (Protocol 1): 0.594 -> 0.592No clear effect
add ground truth label knowledgeSecretGen · CelebA whitebox, face.evoLVe, center maskattack accuracy (Protocol 1): 0.551 -> 0.788Improved
add ground truth label knowledgeSecretGen · CelebA blackbox, face.evoLVe, center maskattack accuracy (Protocol 1): 0.447 -> 0.603Improved
use SecretGen instead of GMI, both with GT labelSecretGen vs GMI · CelebA whitebox, face.evoLVe, face T maskattack accuracy (Protocol 1): 0.467 -> 0.695Improved
replace GMI (GT label) with SecretGen without GT labelSecretGen · CelebA whitebox, face.evoLVe, center maskattack accuracy (Protocol 1): 0.595 -> 0.551Hurt
replace GMI (GT label) with SecretGen without GT labelSecretGen · CelebA whitebox, VGG16, center maskattack accuracy (Protocol 2): 0.955 -> 0.928Hurt
apply purification defense to the target modelSecretGen · CelebA blackbox, VGG16, center maskattack accuracy (Protocol 1): 0.351 -> 0.328Hurt
apply purification defense to the target modelSecretGen · CelebA blackbox, VGG16, face T maskattack accuracy (Protocol 2): 0.837 -> 0.747Hurt
replace sequential cutout with flipping, grayscale or color jitter (ablation)pseudo label predictor · CelebA, face.evoLVelabel prediction accuracyHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy