AI Paper Atlas

Research map / Federated learning and privacy / Differential privacy

Optimal privacy guarantees for a relaxed threat model: Addressing sub-optimal adversaries in differentially private machine learning

Georgios Kaissis, Alexander Ziller, Stefan Kolek, Anneliese Riess, Daniel Rueckert

NeurIPS 2023 · Federated learning and privacy › Differential privacy · approach: Private online and user-level learning · 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
calibrate DP-SGD noise to the relaxed threat model (RTM) guarantee instead of the standard DP threat modelDP-SGD ResNet-9 · CIFAR-10 at (1, 1e-5)-DPTop-1 accuracy: 0.67 -> 0.81Improved
calibrate DP-SGD noise to the RTM guarantee instead of the DPTMDP-SGD BERT · Stanford SNLI at (1, 1.8e-6)-DPTop-1 accuracy: 0.78 -> 0.80Improved
calibrate DP-SGD noise to the RTM guarantee instead of the DPTMDP-SGD ResNet-18 · ImageNet at (10, 1e-6)-DPTop-1 accuracy: 0.46 -> 0.48Improved
increase dimension from d = 1 to d = 30 for the Gaussian mechanism under RTMGaussian mechanism trade-off function · Delta2/sigma = 1lambda-DR privacy (blessing of dimensionality)Improved
adversary cannot reduce effective gradient dimension (50-dim latent linear network) versus d = 1 worst casegradient-space MI attack · Diabetes dataset, Delta2/sigma = 2hypothesis test powerHurt
switch from DPTM to RTM adversary (no access to candidate record)Gaussian mechanism MI attack · power at alpha = 10^-3 vs Delta2/sigmaadversary power 1 - betaHurt
add Poisson subsampling (rate 0.3) to the mechanism under RTMGaussian mechanism · worst-case 1-d taskadversary power lossImproved
audit with Dirac-canary gradient-space attack instead of LiRA logit-space attackempirical privacy auditing · 1-d synthetic binary task, Delta2/sigma = 1tightness to theoretical boundImproved
evaluate reconstruction attack bound under RTM instead of DPTM (prior kappa = 0.1)DP-SGD on CIFAR-10, SNLI, ImageNet · Reconstruction robustnessprobability of reconstructionImproved

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