AI Paper Atlas

Research map / Federated learning and privacy / Differential privacy

Optimizing Canaries for Privacy Auditing with Metagradient Descent

Matteo Boglioni, Terrance Liu, Andrew Ilyas, Steven Wu

ICLR 2026 · Federated learning and privacy › Differential privacy · approach: Private training with public data · 3 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 metagradient canaries instead of random canariesDP-SGD auditing, procedure B · MNIST DP training, eps=8empirical epsilon lower bound: 0.016 -> 0.669Improved
use metagradient canaries instead of random canariesDP-SGD auditing, procedure B · Fashion-MNIST DP finetuning, eps=8empirical epsilon lower bound: 0.056 -> 1.483Improved
use metagradient canaries instead of random canariesDP-SGD auditing, procedure B · CIFAR-10 DP training, eps=8empirical epsilon lower bound: 0.405 -> 0.732Improved
use metagradient canaries instead of random canariesDP-SGD auditing, procedure B · CIFAR-100 DP finetuning, eps=8empirical epsilon lower bound: 0.354 -> 1.286Improved
use metagradient canaries instead of random canariesDP-SGD auditing, procedure A · CIFAR-100 DP finetuning, eps=8empirical epsilon lower bound: 0.342 -> 0.770Improved
use metagradient canaries instead of random canariesDP-SGD auditing, procedure A · CIFAR-10 DP training, eps=8empirical epsilon lower bound: 0.285 -> 0.392Improved
use metagradient canaries instead of random canaries at strict privacy eps=1DP-SGD auditing, procedure B · CIFAR-10 DP training, eps=1empirical epsilon lower bound: 0.030 -> 0.000No clear effect
use metagradient canaries instead of random canaries at strict privacy eps=1DP-SGD auditing, procedure A · CIFAR-100 DP training, eps=1empirical epsilon lower bound: 0.075 -> 0.074No clear effect
use procedure B instead of procedure A with metagradient canariesmetagradient canaries · CIFAR-10 DP training, eps=8empirical epsilon lower bound: 0.392 -> 0.732Improved
add DP finetuning from pretrained model instead of training from scratch (ablation)metagradient canaries, procedure B · CIFAR-10, eps=8empirical epsilon lower bound: 0.732 -> 1.207Improved
use ClipBKD canaries instead of metagradient canariesone-run auditing · CIFAR-10empirical epsilon lower boundHurt
use Gradient Max canaries instead of metagradient canariesone-run auditing · CIFAR-10 vs MNISTempirical epsilon lower boundMixed

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