AI Paper Atlas

Research map / Adversarial robustness and security / Black-box and transfer attacks

Training A Secure Model against Data-Free Model Extraction

Zhenyi Wang, Li Shen, junfeng guo, Tiehang Duan, Siyu Luan, Tongliang Liu, Mingchen Gao

ECCV 2024 · Adversarial robustness and security › Black-box and transfer attacks · approach: Data-free substitute training · 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
train victim with AAUG (weight perturbation plus attack-aware training) instead of output-perturbation defensesResNet34-8x victim · CIFAR-10 DFME soft-labelclone model accuracy (ResNet18-8x): 87.36 -> 46.82Improved
train victim with AAUGResNet34-8x victim · CIFAR-100 DFME soft-labelclone model accuracy (ResNet18-8x): 58.72 -> 21.31Improved
train victim with AAUGResNet34-8x victim · ImageNet-100 DFME soft-labelclone model accuracy (ResNet18-8x): 35.89 -> 15.58Improved
apply output-perturbation defenses (RandP, P-poison, GRAD) against hard-label DFMS-HLResNet34-8x victim · CIFAR-10 DFMS-HLclone model accuracy (ResNet18-8x): 84.67 -> 84.28 (GRAD)No clear effect
train victim with AAUG against hard-label DFMS-HLResNet34-8x victim · CIFAR-100 DFMS-HLclone model accuracy (ResNet18-8x): 72.57 -> 45.68Improved
train victim with AAUGResNet34-8x victim · CIFAR-10 victim utilitytest accuracy: 94.91 -> 94.21No clear effect
train victim with AAUG instead of output perturbationResNet34-8x victim · CIFAR-100 victim utilitytest accuracy: 76.71 -> 75.73No clear effect
use AAUG instead of P-poison and GRAD at test timedefense deployment · CIFAR-10 and CIFAR-100test-time speed: 25 ~103x speed upImproved
use AAUG instead of standard base model trainingvictim model training · training costtraining cost: 1.2xHurt
fine-tune pre-trained model with AAUG for 20 epochs instead of 200-epoch retraining (ablation)AAUG · CIFAR-10defense performance and cost: training cost reduced by 83%No clear effect
increase attacker query budget (ablation)baseline defenses vs AAUG · DFME attackclone model accuracyMixed

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