AI Paper Atlas

Research map / Label-efficient and robust learning / Semi-supervised learning

Backdoor Defense via Decoupling the Training Process

Kunzhe Huang, Yiming Li, Baoyuan Wu, Zhan Qin, Kui Ren

ICLR 2022 · Label-efficient and robust learning › Semi-supervised learning · approach: Self-training theory · 41 citations (OpenAlex)

Read the paper

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
Decouple training with self-supervised backbone learning, followed by classifier training and semi-supervised fine-tuningResNet-18 · CIFAR-10, BadNetsASR: 100 -> 0.96Improved
Decouple training with self-supervised backbone learning, followed by classifier training and semi-supervised fine-tuningResNet-18 · ImageNet, BadNetsASR: 90.49 -> 0.26Improved
Replace the self-supervised backbone with a supervised-trained backbone (ablation)DBD · CIFAR-10, BadNetsASR: 0.96 -> 100Hurt
Use SCE rather than CE for second-stage classifier training (ablation)Self-supervised backbone pipeline · CIFAR-10, BlendedASR: 12.19 -> 6.24Improved
Use SCE rather than CE for second-stage classifier training (ablation)Self-supervised backbone pipeline · CIFAR-10, BadNetsASR: 5.20 -> 5.12No clear effect
Filter out low-credible samples before tuning FC layers (ablation)SS with SCE · CIFAR-10, BadNetsASR: 5.12 -> 4.02Improved
Filter out low-credible samples before tuning FC layers (ablation)SS with SCE · CIFAR-10, BadNetsBA: 82.34 -> 78.94Hurt
Fine-tune the whole model semi-supervised rather than only tuning FC layers on high-credible samples (ablation)SS with SCE + Tuning · CIFAR-10, BadNetsASR: 4.02 -> 0.96Improved
Fine-tune the whole model semi-supervised rather than only tuning FC layers on high-credible samples (ablation)SS with SCE + Tuning · CIFAR-10, BadNetsBA: 78.94 -> 92.41Improved
Increase WaNet kernel size from 4 to 224 (ablation)WaNet on ImageNet · ImageNet, WaNetASR: 6.87 -> 96.22Improved
Increase WaNet strength from 0.4 to 1 (ablation)WaNet on ImageNet · ImageNet, WaNetASR: 80.06 -> 96.22Improved
Remove WaNet noise mode (ablation)WaNet on ImageNet · ImageNet, WaNetASR: 31.72 -> 96.22Improved

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