AI Paper Atlas

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

Measuring Data Reconstruction Defenses in Collaborative Inference Systems

Mengda Yang, Ziang Li, Juan Wang, Hongxin Hu, Ao Ren, Xiaoyang Xu, Wenzhe Yi

NeurIPS 2022 · Adversarial robustness and security › Attacks on text and graphs · approach: Membership inference and inversion · 0 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
replace ERA decoder input with SFD-distilled featuresInverse-Network attack · CelebA 2nd conv block, adversarial defenseMSE: 0.124 -> 0.032Improved
replace ERA with SFDInverse-Network attack · CelebA 2nd conv block, adversarial defenseSSIM: 0.370 -> 0.618Improved
SFD against noise mask defenseSFD attack · CelebA 2nd conv blockMSE: 47.7%Improved
SFD against dropout defenseSFD attack · CelebA 2nd conv blockMSE: 40.9%Improved
move split point from 2nd conv block to avgpoolSFD attack gain over ERA · CelebA, adversarial defenseMSE improvement: 0.092 -> 0.003Hurt
move split point from 2nd conv block to avgpoolSFD attack gain over ERA · CelebA, adversarial defenseSSIM improvement: 0.248 -> 0.015Hurt
split at 1st fc layer instead of conv blockSFD attack · CelebA, adversarial defenseSSIM: -0.002No clear effect
increase adversarial lambda from 1.0 to 5.0ERA attack · CIFAR10 1st conv blockSSIM: 0.945 -> 0.309Hurt
increase adversarial lambda from 1.0 to 5.0target VGG-16 classifier · CIFAR10 1st conv blockaccuracy: 83.1 -> 76.5Hurt
increase noise scale b from 1.0 to 10.0target VGG-16 classifier · CIFAR10 1st conv blockaccuracy: 85.5 -> 82.3Hurt
increase dropout rate r from 0.5 to 0.9target VGG-16 classifier · CIFAR10 1st conv blockaccuracy: 84.9 -> 79.9Hurt
SFD under strongest defense hyper-parametersSFD attack · CIFAR10 1st conv blockSSIM improvement over ERA: 0.181Improved

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