Paper ID: 2204.00487

Robust and Accurate -- Compositional Architectures for Randomized Smoothing

Miklós Z. Horváth, Mark Niklas Müller, Marc Fischer, Martin Vechev

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 $\ell_2$ perturbations with $r=1.0$. We release our code and models at https://github.com/eth-sri/aces.

Submitted: Apr 1, 2022