AI Paper Atlas

Research map / Adversarial robustness and security / Backdoor and poisoning

The Victim and The Beneficiary: Exploiting a Poisoned Model to Train a Clean Model on Poisoned Data

Zixuan Zhu, Rui Wang, Cong Zou, Lihua Jing

ICCV 2023 · Adversarial robustness and security › Backdoor and poisoning · approach: Training-time backdoor defense · 3 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
add AttentionMix augmentation to filtered training (ablation)V&B · CIFAR-10 BlendASR: 76.70% -> 1.42%Improved
add AttentionMix augmentation to filtered training (ablation)V&B · CIFAR-10 WaNetASR: 90.97% -> 25.16%Improved
add AttentionMix augmentation (ablation)V&B · CIFAR-10BA: 93.50% -> 92.27% (BadNets)Hurt
add standard semi-supervised training on top of filter and AttentionMix (ablation)V&B · CIFAR-10 WaNetASR: 25.16% -> 1.23%Improved
add semi-supervised training without suppression (ablation)V&B · CIFAR-10 BadNetsASR: 0.42% -> 1.21%Hurt
add suppression of suspicious samples with Victim predictions (ablation)V&B · CIFAR-10 DynamicASR: 28.79% -> 1.43%Improved
use AttentionMix instead of Mixup, Cutmix or Attentive Cutmix (ablation)V&B · CIFAR-10 WaNet/DynamicASR: 13.38% (Attentive Cutmix) -> 1.13% on DynamicImproved
AttentionMix vs other augmentations on simple triggers (BadNets, Blend) (ablation)V&B · CIFAR-10 BadNetsASR: 0.27% (Mixup) vs 0.62% (AttentionMix)No clear effect
increase mixing threshold t_m from 0.1 toward 0.5V&B · CIFAR-10BAHurt
increase mixing threshold t_m from 0.1 toward 0.5V&B · ImageNet subsetBANo clear effect
apply V&B instead of no defense on ImageNet subsetResNet-18 · ImageNet subsetBA: 94.94% -> 95.42% (BadNets)Improved

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