AI Paper Atlas

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

Gaussian Membership Inference Privacy

Tobias Leemann, Martin Pawelczyk, Gjergji Kasneci

NeurIPS 2023 · Federated learning and privacy › Privacy attacks and secure learning · approach: Membership inference and auditing · 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
use the analytical LRT-based GLiR attack with known gradient distribution parametersGLiR attack on simulated gradients · Simulated gradient distribution, single stepmatch to analytical trade-offImproved
estimate gradient mean and covariance from background data instead of knowing them (ablation)GLiR attack · Real model gradients on Purchase, CIFAR-10, Adultattack strength vs analytical curveHurt
compose five SGD steps instead of oneGLiR attack · CIFAR-10 at FPR=0.25attack performance (TPR): mu = 1.14 -> mu = 2.54Improved
compose five SGD steps instead of oneGLiR attack · Purchase and Adultattack performance (TPR)No clear effect
calibrate noise for mu-GMIP instead of mu-GDP at the same munoisy SGD · CIFAR-10, Purchase, Adulttest accuracyImproved
increase mu above 0.86 under mu-GMIPnoisy SGD · CIFAR-10required noiseImproved
push mu toward 0 (high privacy) for both GDP and GMIPnoisy SGD · CIFAR-10, Purchase, Adulttest accuracyHurt
use loss-based attacks such as LiRA instead of the gradient attackmembership inference attack · Appendix C.4 comparisonattack success rateHurt
increase d relative to batch size none-step SGD bound · Theory (Corollary 5.1)attacker successHurt
increase gradient susceptibility Kone-step SGD bound · Theory (Theorem 5.1)attacker successHurt
increase d and n so the trade-off converges to the Gaussian curveTheorem 5.1 trade-off function · Figure 1 (n=4,d=3; n=10,d=7; n=22,d=15)closeness to mu-GMIP curveImproved

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