AI Paper Atlas

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

TRS: Transferability Reduced Ensemble via Promoting Gradient Diversity and Model Smoothness

Zhuolin Yang, Linyi Li, Xiaojun Xu, Shiliang Zuo, Qian Chen, Pan Zhou, Benjamin I. P. Rubinstein, Ce Zhang, Bo Li

NeurIPS 2021 · Adversarial robustness and security › Black-box and transfer attacks · approach: Transferable adversarial examples · 16 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
Jointly reduce inter-model loss-gradient similarity and enforce neighborhood smoothnessTRS ensemble · MNIST, PGD (50), ε=0.15robust accuracy (%): 47.7 -> 85.1Improved
Jointly reduce inter-model loss-gradient similarity and enforce neighborhood smoothnessTRS ensemble · CIFAR-10, PGD (50), ε=0.01robust accuracy (%): 37.1 -> 50.5Improved
Jointly reduce inter-model loss-gradient similarity and enforce neighborhood smoothnessTRS ensemble · MNIST, FGSM, ε=0.1robust accuracy (%): 95.8 -> 95.6Mixed
Jointly reduce inter-model loss-gradient similarity and enforce neighborhood smoothnessTRS ensemble · CIFAR-10, FGSM, ε=0.02robust accuracy (%): 63.8 -> 44.2Hurt
Add the model-smoothness loss to cosine-similarity regularization (ablation)TRS ensemble · MNIST, PGD (50), ε=0.1robust accuracy (%): 0.0 -> 93.0Improved
Add the model-smoothness loss to cosine-similarity regularization (ablation)TRS ensemble · CIFAR-10, PGD (50), ε=0.01robust accuracy (%): 0.0 -> 50.5Improved
Use min-max neighborhood smoothing instead of pointwise gradient ℓ2 regularization (ablation)TRS ensemble · MNIST, PGD (50), ε=0.1robust accuracy (%): 73.4 -> 93.0Improved
Use min-max neighborhood smoothing instead of pointwise gradient ℓ2 regularization (ablation)TRS ensemble · CIFAR-10, PGD (50), ε=0.01robust accuracy (%): 15.7 -> 50.5Improved
Enforce model smoothness in addition to gradient dissimilarityEnsemble base models · MNIST and CIFAR-10, PGD transferabilityattack success rateImproved

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