Indicators of Attack Failure: Debugging and Improving Optimization of Adversarial Examples
arXiv:2106.09947
Abstract
Evaluating robustness of machine-learning models to adversarial examples is a challenging problem. Many defenses have been shown to provide a false sense of robustness by causing gradient-based attacks to fail, and they have been broken under more rigorous evaluations. Although guidelines and best practices have been suggested to improve current adversarial robustness evaluations, the lack of automatic testing and debugging tools makes it difficult to apply these recommendations in a systematic manner. In this work, we overcome these limitations by: (i) categorizing attack failures based on how they affect the optimization of gradient-based attacks, while also unveiling two novel failures affecting many popular attack implementations and past evaluations; (ii) proposing six novel indicators of failure, to automatically detect the presence of such failures in the attack optimization process; and (iii) suggesting a systematic protocol to apply the corresponding fixes. Our extensive experimental analysis, involving more than 15 models in 3 distinct application domains, shows that our indicators of failure can be used to debug and improve current adversarial robustness evaluations, thereby providing a first concrete step towards automatizing and systematizing them. Our open-source code is available at: https://github.com/pralab/IndicatorsOfAttackFailure.
Accepted at NeurIPS 2022
References in corpus (9)
- Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks
- On Evaluating Adversarial Robustness
- Security Evaluation of Pattern Classifiers under Attack
- Defensive Distillation is Not Robust to Adversarial Examples
- Improving Adversarial Robustness via Promoting Ensemble Diversity
- RobustBench: a standardized adversarial robustness benchmark
- A New Defense Against Adversarial Images: Turning a Weakness into a Strength
- Fast Minimum-norm Adversarial Attacks through Adaptive Norm Constraints
- A critique of the DeepSec Platform for Security Analysis of Deep Learning Models