# Adversarial Robustness Curves

## TL;DR

This paper introduces robustness curves as a general framework to analyze the robustness of models against adversarial examples, independent of specific thresholds and norms, enhancing understanding of model reliability.

## Contribution

It proposes robustness curves to decouple robustness analysis from thresholds and norms, providing a more comprehensive view of model robustness against adversarial attacks.

## Key findings

- Robustness curves can qualitatively depend on the chosen norm under certain conditions.
- The framework allows for a more nuanced understanding of model robustness beyond traditional metrics.
- The approach facilitates separating robustness analysis from specific norm choices.

## Abstract

The existence of adversarial examples has led to considerable uncertainty regarding the trust one can justifiably put in predictions produced by automated systems. This uncertainty has, in turn, lead to considerable research effort in understanding adversarial robustness. In this work, we take first steps towards separating robustness analysis from the choice of robustness threshold and norm. We propose robustness curves as a more general view of the robustness behavior of a model and investigate under which circumstances they can qualitatively depend on the chosen norm.