AI Paper Atlas

Research map / Adversarial robustness and security / Backdoor and poisoning

Black-Box Detection of Backdoor Attacks With Limited Information and Data

Yinpeng Dong, Xiao Yang, Zhijie Deng, Tianyu Pang, Zihao Xiao, Hang Su, Jun Zhu

ICCV 2021 · Adversarial robustness and security › Backdoor and poisoning · approach: Training-time backdoor defense · 100 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
replace gradient-based NC with gradient-free B3Dbackdoor detection · CIFAR-10detection accuracy: 95.0% -> 97.5%Improved
use Bernoulli discrete mask (B3D) instead of relaxed continuous mask (NC)trigger reverse-engineering · CIFAR-10 normal modelsnormal models wrongly flagged (of 50): 8/50 -> 2/50Improved
use synthetic samples instead of clean data (B3D-SS vs B3D)B3D · ImageNet (10-class subset)detection accuracy: 96.0% -> 95.5%No clear effect
use synthetic samples instead of clean data (B3D-SS vs B3D)B3D · CIFAR-10detection accuracy: 97.5% -> 97.5%No clear effect
increase trigger size from 1x1 to 3x3NC · CIFAR-10Case I detections (of 50): 40/50 -> 47/50 -> 49/50Improved
use B3D reversed-trigger KL score instead of STRIPtriggered-input detection · CIFAR-10AUC: 0.9332 -> 0.9958Improved
use B3D reversed-trigger KL score instead of STRIPtriggered-input detection · GTSRBAUC: 0.4937 -> 0.9946Improved
vary number of gradient samples k within 20 to 100 (ablation)B3D · CIFAR-10detection accuracy: k in [20,100] leads to similar resultsNo clear effect
use BigGAN synthetic images instead of random noise (ablation)B3D-SS · ImageNet (10-class subset)detection qualityImproved
add heuristic rule (mask L1 below one fourth of median) to MAD outlier detection (ablation)NC and B3D · CIFAR-10detection accuracyImproved
add data augmentation to backdoored model trainingbackdoored ResNet-18 · CIFAR-10clean accuracy and ASR: accuracy 94.68%, 94.78%, 95.29%; ASR 99.67%, 99.99%, 100.00%Improved
switch ResNet-18 to VGG-16B3D · CIFAR-10detection accuracy: 97.5% -> 97.0%No 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