AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

How Does Data Augmentation Affect Privacy in Machine Learning?

Da Yu, Huishuai Zhang, Wei Chen, Jian Yin, Tie-Yan Liu

AAAI 2021 · Adversarial robustness and security › Attacks on text and graphs · approach: Membership inference and inversion · 40 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
Use mean loss across augmented instances instead of a single-instance lossM_loss membership inference · CIFAR10, WRN16-8, k=10MI success rate (%): 61.9 -> 68.9Improved
Use permutation-invariant raw-moment features instead of mean lossM_mean membership inference · CIFAR10, WRN16-8, k=10MI success rate (%): 68.9 -> 70.1Improved
Use augmented-instance losses for membership inferenceM_loss · CIFAR10, WRN16-8, augmentation-trained model versus unaugmented modelMI success rate (%): 62.9 -> 61.9Improved
Use concatenated raw losses as neural-network features rather than permutation-invariant moments (ablation)M_moments · CIFAR10, WRN16-8, k=10MI success rate (%): 70.1 -> 63.1Hurt
Use mean loss instead of concatenated raw losses as neural-network features (ablation)M_NN_loss · CIFAR10, WRN16-8, k=10MI success rate (%): 63.1 -> 68.9Improved
Increase the number of augmented attack instances k (ablation)M_loss, M_mean, and M_moments · CIFAR10 and CIFAR100MI success rate (%)Hurt

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