# Accurate, reliable and fast robustness evaluation

**Authors:** Wieland Brendel, Jonas Rauber, Matthias Kümmerer, Ivan Ustyuzhaninov, Matthias Bethge  
**arXiv:** [1907.01003](https://arxiv.org/abs/1907.01003)  
**Date:** 2019-07-01  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/1907.01003v2/

## Key points

- Vulnerability of neural networks to adversarial attacks is a major concern in Deep Learning
- Authors developed new gradient-based adversarial attacks that overcome limitations of existing methods
- Attacks are more reliable and perform better than current state-of-the-art gradient-based attacks
- Minimal hyperparameter tuning required, adaptable to various adversarial criteria, and more query efficient
- Validated across six different models and hold true for various types of perturbations (L0, L1, L2, Linf) in targeted and untargeted scenarios
- Implementations will be available in popular toolboxes such as Foolbox, CleverHans, and ART
- New class of attacks aims to make robustness evaluations easier and more reliable for improved performance against adversarial perturbations
- Expected to lead to the development of more robust machine learning models
- Presents an important advancement in evaluating the robustness of neural network models and offers promising solutions for addressing susceptibility to adversarial attacks in Deep Learning systems.

## Summary

In recent years, the vulnerability of neural networks to adversarial attacks has become a major concern in the field of Deep Learning. To address this issue, the authors of this study have developed a new set of gradient-based adversarial attacks that overcome the limitations of existing methods. These attacks are more reliable in the face of gradient-masking and perform better than current state-of-the-art gradient-based attacks while requiring minimal hyperparameter tuning. Additionally, they can be adapted to various adversarial criteria and are more query efficient. The findings of this study have been carefully validated across six different models and hold true for various types of perturbations (L0, L1, L2, Linf) in both targeted and untargeted scenarios. The implementations of these attacks will soon be available in popular toolboxes such as Foolbox, CleverHans, and ART. The authors hope that their new class of attacks will make robustness evaluations easier and more reliable by providing clearer signals in the search for improved performance against adversarial perturbations. This contribution is expected to lead to the development of more robust machine learning models. Overall, this study presents an important advancement in evaluating the robustness of neural network models and offers promising solutions for addressing the susceptibility to adversarial attacks in Deep Learning systems.