Robust and Accurate – Compositional Architectures for Randomized Smoothing

by   Miklós Z. Horváth, et al.
ETH Zurich

Randomized Smoothing (RS) is considered the state-of-the-art approach to obtain certifiably robust models for challenging tasks. However, current RS approaches drastically decrease standard accuracy on unperturbed data, severely limiting their real-world utility. To address this limitation, we propose a compositional architecture, ACES, which certifiably decides on a per-sample basis whether to use a smoothed model yielding predictions with guarantees or a more accurate standard model without guarantees. This, in contrast to prior approaches, enables both high standard accuracies and significant provable robustness. On challenging tasks such as ImageNet, we obtain, e.g., 80.0% natural accuracy and 28.2% certifiable accuracy against ℓ_2 perturbations with r=1.0. We release our code and models at


page 1

page 2

page 3

page 4


Randomized Smoothing of All Shapes and Sizes

Randomized smoothing is a recently proposed defense against adversarial ...

Boosting Randomized Smoothing with Variance Reduced Classifiers

Randomized Smoothing (RS) is a promising method for obtaining robustness...

Scalable Certified Segmentation via Randomized Smoothing

We present a new certification method for image and point cloud segmenta...

Randomized Smoothing with Masked Inference for Adversarially Robust Text Classifications

Large-scale pre-trained language models have shown outstanding performan...

Certified Defense via Latent Space Randomized Smoothing with Orthogonal Encoders

Randomized Smoothing (RS), being one of few provable defenses, has been ...

(De-)Randomized Smoothing for Decision Stump Ensembles

Tree-based models are used in many high-stakes application domains such ...

Smooth-Reduce: Leveraging Patches for Improved Certified Robustness

Randomized smoothing (RS) has been shown to be a fast, scalable techniqu...

Please sign up or login with your details

Forgot password? Click here to reset