AI Paper Atlas

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

Revealing and Protecting Labels in Distributed Training

Trung Dang, Om Thakkar, Swaroop Ramaswamy, Rajiv Mathews, Sang Chin, Françoise Beaufays

NeurIPS 2021 · Federated learning and privacy › Privacy attacks and secure learning · approach: Gradient inversion attacks · 8 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 RLG instead of Yin et al.'s methodEfficientNetB0 label-set inference · ImageNet validation, N=10/50/100F1: .020 -> 1; .093 -> .999; .174 -> .999Improved
Use RLG instead of Yin et al.'s methodResNet50 label-set inference · ImageNet validation, N=10/50F1: .998 -> .998; .987 -> .987No clear effect
Use RLG instead of Yin et al.'s methodResNet50 label-set inference · ImageNet validation, N=100F1: .979 -> .972Hurt
Use RLG on an untrained model's projection-layer updateASR bag-of-words inference · LibriSpeech testEM: .018 -> .988Improved
Train the ASR model for 4k rather than use the untrained model (ablation)RLG bag-of-words inference · LibriSpeech testEM: .988 -> .955Hurt
Train the ASR model for 10k rather than use the untrained model (ablation)RLG bag-of-words inference · LibriSpeech testEM: .988 -> .933Hurt
Use a single-step mini-batch update instead of aggregating multiple steps (ablation)RLG bag-of-words inference · LibriSpeech test, N=4 or 8, K=1EM: .535 -> 1; .160 -> 1Improved
Aggregate single-sample updates over 4 or 8 steps (ablation)RLG bag-of-words inference · LibriSpeech testEM: 1 -> .535; 1 -> .160Hurt
Aggregate 2-sample updates over 2 steps (ablation)RLG bag-of-words inference · LibriSpeech test, N=2, K=2EM: 1 -> .330Hurt
Apply Sign-SGD to the projection-layer update (ablation)RLG · EfficientNetB0 and ASREM: 1 -> 0; .988 -> .104Hurt
Apply GradDrop at 50% (ablation)RLG · EfficientNetB0 and ASREM: 1 -> .17; .988 -> .881Mixed
Apply GradDrop at 90% (ablation)RLG · EfficientNetB0 and ASREM: 1 -> 0; .988 -> .866Mixed

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