AI Paper Atlas

Research map / Federated learning and privacy / Privacy attacks and secure learning

When Does Data Augmentation Help With Membership Inference Attacks?

Yigitcan Kaya, Tudor Dumitras

ICML 2021 · Federated learning and privacy › Privacy attacks and secure learning · approach: Membership inference and auditing · 15 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
Apply augmentation (SL, LS, DL, RC, CO, GA, MU) tuned for maximum accuracyVGG CNNs · CIFAR-100Advpow: varies by mechanismMixed
Label smoothing for accuracyVGG CNN · CIFAR-100Advpow: Acc +10%, Advpow 60.2 to 75.6Hurt
Random cropping at high intensity (RAD<25%) (ablation)VGG CNN · CIFAR-100Advpow: 60.2 to 2.9 at P=25 with Acc 42.6%Improved
Random cropping at moderate intensity (RAD<10%) (ablation)VGG CNN · CIFAR-100Advpow: 60.2 to 7.1 at P=14 with Acc 52.8%Improved
Gaussian augmentationVGG CNN · CIFAR-100Advpow: 58.2/60.2 baseline vs 58.8/62.5 at accuracy-tunedNo clear effect
Add label smoothing on top of random cropping (ablation)VGG CNN · CIFAR-100Advpow: RC alone P=4 Advpow 32.0 vs 38.4 with α=0.90Hurt
Label smoothing on non-image tasks (ablation)fully connected networks · Texas-100Adv: 55.0% to 64.6% at α=0.10Hurt
Label smoothing on non-image tasks (ablation)fully connected networks · Purchase-100Adv: 14.6% to 20.5% at α=0.10 with Acc dropMixed
Differential privacy (DP-SGD) with eps<=1VGG CNN · CIFAR-100Acc: 54.3% baseline to 6.8%Hurt
Fewer training epochs (3-7 vs 35) (ablation)VGG CNN · CIFAR-10Adv: 65-90% less Adv, with 5-17% lower AccImproved

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