AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Adversarial Examples Make Strong Poisons

Liam H Fowl, Micah Goldblum, Ping-yeh Chiang, Jonas Geiping, Wojciech Czaja, Tom Goldstein

NeurIPS 2021 · Adversarial robustness and security › Attacks on text and graphs · approach: Clean-label data poisoning · 34 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
Train on PGD adversarial poisons rather than clean dataResNet-18 · CIFAR-10 validationaccuracy (%): 94.56 -> 6.25Hurt
Use adversarial poisoning instead of Unlearnable ExamplesResNet-18 · CIFAR-10 validationaccuracy (%): 19.85 -> 6.25Hurt
Use class-targeted rather than untargeted poisons (ablation)Adversarial Poisoning · ImageNet validationaccuracy (%): 45.07 -> 3.57Hurt
Increase the perturbation bound from ε=4/255 to ε=8/255 (ablation)Adversarial Poisoning · CIFAR-10 validationaccuracy (%): 52.97 ± 1.23 -> 10.98 ± 0.27Improved
Use class-targeted instead of untargeted poisons at ε=4/255 (ablation)Adversarial Poisoning · CIFAR-10 validationaccuracy (%): 52.97 ± 1.23 -> 28.49 ± 0.65Improved
Increase PGD crafting steps from 50 to 250 (ablation)Adversarial Poisoning · CIFAR-10 validationaccuracy (%): 16.53 ± 0.26 -> 6.25 ± 0.17Improved
Replace PGD with Carlini-Wagner crafting loss (ablation)Adversarial Poisoning · CIFAR-10 validationaccuracy (%): 6.25 ± 0.17 -> 48.40 ± 3.24Hurt
Craft with an adversarially trained model (ablation)Adversarial Poisoning · CIFAR-10 validationaccuracy (%): 6.25 ± 0.17 -> 81.79 ± 0.57Hurt
Relabel poisons using the crafting network's adversarial class (ablation)Adversarially poisoned CIFAR-10 · CIFAR-10 validationaccuracy (%): 6.25 ± 0.17 -> 78.98 ± 0.25Improved
Train with adversarially perturbed SVHN images rather than CIFAR-10 poisons (ablation)Corrected-label training · CIFAR-10 validationaccuracy (%): 78.98 ± 0.25 -> 30.19 ± 0.16Hurt
Apply adversarial training to the poisoned-data victimClass-targeted adversarial poisoning · CIFAR-10 validationaccuracy (%): 6.25 -> 83.01Improved
Use Gaussian smoothing as a defenseClass-targeted adversarial poisoning · CIFAR-10 validationaccuracy (%): 11.94No clear effect

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