AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

Dataset Inference for Self-Supervised Models

Adam Dziedzic, Haonan Duan, Muhammad Ahmad Kaleem, Nikita Dhawan, Jonas Guan, Yannis Cattan, Franziska Boenisch, Nicolas Papernot

NeurIPS 2022 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Data-efficient and robust self-supervision · 7 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
increase stealing queries from 5K to 250K (ImageNet queries) (ablation)dataset inference · ImageNet victimp-value: 9.88e-1 -> 3.14e-5Improved
increase stealing queries from 500 to 50K (CIFAR10 queries) (ablation)stolen encoder, SVHN victim · SVHN victimp-value: 6.89e-1 -> 1.19e-17Improved
steal with the victim's own data type (SVHN queries on SVHN victim) vs other datadataset inference · SVHN victimp-value: 1.19e-17 -> 1.05e-75Improved
apply shuffle obfuscation to stolen encoder (ablation)dataset inference · CIFAR10 victimp-value: 8.73e-7 -> 1.72e-6No clear effect
apply pad obfuscation to stolen encoder (ablation)dataset inference · CIFAR10 victimp-value: 8.73e-7 -> 3.44e-6No clear effect
apply linear transform obfuscation (ablation)dataset inference · CIFAR10 victimp-value: 8.73e-7 -> 6.81e-7No clear effect
shuffle the representation elements (ablation)cosine similarity score · CIFAR10 victimcosine similarity C: 0.504 -> 0.002Hurt
pad the representation with zeros (ablation)cosine similarity score · CIFAR10 victimcosine similarity C: 0.504 -> 0.007Hurt
shuffle the representation elements (ablation)mutual information score · CIFAR10 victimmutual information score S: 0.73 -> 0.74No clear effect
use independent encoder instead of stolen (cosine similarity)encoder similarity · CIFAR10 victimcosine similarity C: 0.95 -> 0.0007Hurt
use independent encoder instead of stolen (mutual information)encoder similarity · SVHN victimmutual information score S: 0.96 -> 0.90No clear effect
increase queries from 500 to 50K (CIFAR10 queries) (ablation)stolen encoder · SVHN downstreamlinear eval accuracy: 19.3 -> 75.1Improved

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