AI Paper Atlas

Research map / Adversarial robustness and security / Backdoor and poisoning

Marksman Backdoor: Backdoor Attacks with Arbitrary Target Class

Khoa D Doan, Yingjie Lao, Ping Li

NeurIPS 2022 · Adversarial robustness and security › Backdoor and poisoning · approach: Trigger-based backdoor attacks · 10 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 a class-conditional generative trigger and constrained joint optimization instead of repeated fixed trigger patternsMarksman versus multi-trigger baselines · MNIST, CIFAR10, GTSRB, T-IMNET at 10% poisoningAttack success rate: MNIST 0.784 -> 1.000; CIFAR10 0.730 -> 1.000; GTSRB 0.802 -> 0.997; T-IMNET 0.376 -> 0.999Improved
Use Marksman rather than repeated multi-trigger baselinesMarksman · MNIST, CIFAR10, GTSRB, T-IMNET at 10% poisoningClean accuracy: MNIST: 0.969 -> 0.983; CIFAR10: 0.934 -> 0.943; GTSRB: 0.951 -> 0.979; T-IMNET: 0.577 -> 0.575Mixed
Use Marksman rather than repeated multi-trigger baselinesMarksman · MNIST, CIFAR10, GTSRB, T-IMNET at 50% poisoningAttack success rate: MNIST: 0.909 -> 1.000; CIFAR10: 0.999 -> 1.000; GTSRB: 0.999 -> 0.999; T-IMNET: 0.999 -> 0.999Improved
Transfer the learned trigger generator to a separately trained classifierMarksman trigger · CIFAR10 and GTSRBAttack success rate: CIFAR10: 0.990 -> 0.999; GTSRB: 0.993 -> 0.997Improved
Set alpha to 1.0, removing classifier poisoning (ablation)Marksman · CIFAR10Attack success rate: ASR drops to 0Hurt
Lower alpha from its near-optimal value (ablation)Marksman · CIFAR10Clean accuracyHurt
Vary beta while holding alpha at 0.8 (ablation)Marksman · CIFAR10Clean and attack performanceNo clear effect
Update the trigger generator less frequently than the classifierMarksman optimization · TrainingConvergence speedImproved
Apply Neural Cleanse to the Marksman modelMarksman · MNIST, CIFAR10, GTSRB, T-IMNETBackdoor detectionNo clear effect
Apply STRIP to detect Marksman-triggered inputsMarksman · MNIST, CIFAR10, GTSRB, T-IMNETDetection separationNo clear effect
Apply Spectral Signature to Marksman samplesMarksman · MNIST, CIFAR10, GTSRB, T-IMNETLatent-space separationNo clear effect
Apply Fine-Pruning to the Marksman modelMarksman · MNIST, CIFAR10, GTSRB, T-IMNETAttack success rateNo 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