Pith. sign in

REVIEW 5 cited by

Toward Training at ImageNet Scale with Differential Privacy

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2201.12328 v2 pith:FNRMSRHD submitted 2022-01-28 cs.LG

Toward Training at ImageNet Scale with Differential Privacy

classification cs.LG
keywords privacytrainingaccuracydifferentialimagenetmodelsscaletrain
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Differential privacy (DP) is the de facto standard for training machine learning (ML) models, including neural networks, while ensuring the privacy of individual examples in the training set. Despite a rich literature on how to train ML models with differential privacy, it remains extremely challenging to train real-life, large neural networks with both reasonable accuracy and privacy. We set out to investigate how to do this, using ImageNet image classification as a poster example of an ML task that is very challenging to resolve accurately with DP right now. This paper shares initial lessons from our effort, in the hope that it will inspire and inform other researchers to explore DP training at scale. We show approaches that help make DP training faster, as well as model types and settings of the training process that tend to work better in the DP setting. Combined, the methods we discuss let us train a Resnet-18 with DP to $47.9\%$ accuracy and privacy parameters $\epsilon = 10, \delta = 10^{-6}$. This is a significant improvement over "naive" DP training of ImageNet models, but a far cry from the $75\%$ accuracy that can be obtained by the same network without privacy. The model we use was pretrained on the Places365 data set as a starting point. We share our code at https://github.com/google-research/dp-imagenet, calling for others to build upon this new baseline to further improve DP at scale.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Differentially Private Contrastive Learning via Bounding Group-level Contribution

    cs.CR 2026-04 unverdicted novelty 7.0

    DP-GCL improves differentially private contrastive learning by bounding group-level contributions through batch partitioning and intra-group augmentation, delivering 5.6% higher image classification accuracy and 20.1%...

  2. High-Dimensional Private Linear Regression with Optimal Rates

    stat.ML 2025-05 accept novelty 7.0

    DP-GD achieves minimax optimal non-asymptotic risk O(γ + γ²/ρ²) for well-conditioned high-dimensional data and power-law scaling for ill-conditioned power-law spectra, with the exponent depending on the privacy parameter ρ.

  3. Privacy Leakage via Output Label Space and Differentially Private Continual Learning

    cs.LG 2024-11 unverdicted novelty 7.0

    Identifies output label space as a privacy side-channel in DP continual learning, formalizes DP for CL, and demonstrates two mitigation methods yielding higher accuracy than prior work.

  4. Fundamental Limitations of Favorable Privacy-Utility Guarantees for DP-SGD

    cs.LG 2026-01 unverdicted novelty 6.0

    Shuffled DP-SGD requires σ ≥ 1/√(2 ln M) or κ ≥ (1/√8)(1 - 1/√(4π ln M)) to limit adversarial advantage, preventing strong privacy and high utility simultaneously.

  5. The pretraining domain outweighs the training objective in setting the privacy-utility trade-off of differentially private medical image analysis

    cs.CV 2026-01 reject novelty 5.0

    In DP-SGD chest X-ray classification, MIMIC-CXR supervised pretraining beats ImageNet and DINOv3 initializations, but the study cannot cleanly separate pretraining domain from objective because key comparison arms are...