AI Paper Atlas

Research map / Adversarial robustness and security / Backdoor and poisoning

Excess Capacity and Backdoor Poisoning

Naren Sarayu Manoj, Avrim Blum

NeurIPS 2021 · Adversarial robustness and security › Backdoor and poisoning · approach: Trigger-based backdoor attacks · 1 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 PGD adversarial training instead of vanilla trainingCNN · MNIST target 0, alpha=0.05training robust loss: 1.00 -> 0.13Improved
use PGD adversarial training instead of vanilla trainingCNN · MNIST target 0, alpha=0.05backdoor success rate: 1.00 -> 0.00Improved
use PGD adversarial training instead of vanilla trainingCNN · MNIST target 0, alpha=0.00testing robust loss: 1.00 -> 0.09Improved
use PGD adversarial training instead of vanilla trainingCNN · MNIST target 0, alpha=0.05testing 0-1 loss: 0.01 -> 0.03Hurt
increase poisoned fraction alpha from 0.00 to 0.30 under adversarial trainingCNN with PGD adversarial training · MNIST target 0training robust loss: 0.09 -> 0.41Hurt
increase poisoned fraction alpha from 0.00 to 0.30 under adversarial trainingCNN with PGD adversarial training · MNIST target 0testing robust loss: 0.09 -> 0.19Hurt
increase poisoned fraction alpha from 0.00 to 0.30 under adversarial trainingCNN with PGD adversarial training · MNIST target 0backdoor success rate: 0.00 -> 0.05No clear effect
increase alpha with vanilla trainingCNN with vanilla training · MNIST target 0training 0-1 loss: 0.01 -> 0.01No clear effect
vary target label t from 0 to 9CNN with PGD adversarial training · MNISTrobust loss and backdoor successNo 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