AI Paper Atlas

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

Attacks on Online Learners: a Teacher-Student Analysis

Riccardo Giuseppe Margiotta, Sebastian Goldt, Guido Sanguinetti

NeurIPS 2023 · Adversarial robustness and security › Attacks on text and graphs · approach: Attacks and defenses in sequential settings · 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
raise attack strength (lower action cost C) past a critical valueOnline learner under label attack · Logistic regression, MNISTstudent accuracyHurt
use greedy attack instead of clairvoyant attackGreedy attacker · Teacher-student perceptron/NNrunning costMixed
use constant attack instead of greedy attackConstant attacker · Teacher-student setuprunning costHurt
decrease batch size for greedy attacksGreedy attacker · Linear teacher-studentstudent-target distanceImproved
increase batch size PGreedy attacker · Linear teacher-studentaccuracy transition sharpnessMixed
use RL attacker instead of greedy attackerRL (TD3) attacker · Teacher-student NNcomputational costHurt
use full training versus transfer learning for the attacked learnerGreedy attack · MNIST and CIFAR10 binary tasksaccuracy vs attack strengthNo clear effect
add weight decay in transfer learning (VGG11, ResNet18)Attacked student · CIFAR10 cats vs dogsstudent-teacher distanceMixed
swap SGD for Adam in student trainingAttacked student · Real-data experimentslearning dynamicsNo 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