AI Paper Atlas

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

CoPur: Certifiably Robust Collaborative Inference via Feature Purification

Jing Liu, Chulin Xie, Oluwasanmi O Koyejo, Bo Li

NeurIPS 2022 · 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
use CoPur feature purification instead of no defensecollaborative inference · NUS-WIDE targeted PGDrobust accuracy (%): 4.5±0.3 -> 83.7± 0.9Improved
use CoPur instead of Manifold Projection on ExtraSensorycollaborative inference · ExtraSensory targeted PGDrobust accuracy (%): 53.6 ± 35.7 -> 79.0± 5.0Improved
use Manifold Projection alone instead of no defense on NUS-WIDEcollaborative inference · NUS-WIDE targeted PGDrobust accuracy (%): 4.5±0.3 -> 81.3±1.2Improved
use CoPur instead of Manifold Projection on NUS-WIDEcollaborative inference · NUS-WIDE targeted PGDrobust accuracy (%): 81.3±1.2 -> 83.7± 0.9Improved
use adversarial training instead of no defensecollaborative inference · NUS-WIDE targeted PGDrobust accuracy (%): 4.5±0.3 -> 8.0±1.9Mixed
use block-wise adversarial training instead of standard adversarial trainingadversarial training · ExtraSensory targeted PGDrobust accuracy (%): 19.8 ± 34.1 -> 0.0± 0.0Hurt
use CoPur under combined attack instead of Randomized Ablationcollaborative inference · ExtraSensory combined attackrobust accuracy (%): 60.3± 28.1 -> 79.1± 4.5Improved
increase feature-flipping amplification from 1 to 20baseline defenses · ExtraSensory untargetedrobust accuracyHurt
add more malicious agents (agents 7-10 vs 8-10)CoPur · ExtraSensory untargetedrobust accuracyHurt
more than two malicious agentsAdversarial Training · ExtraSensory untargetedrobust accuracyHurt

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