There Is No Free Lunch In Adversarial Robustness (But There Are Unexpected Benefits)

By Dimitris Tsipras, Shibani Santurkar, Logan Engstrom, Alexander Turner, Aleksander Madry. 2018

84 citations recorded by the source corpus.

Subfields: Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Safety, Robustness & Interpretability

Citations in the corpus

Feature Denoising for Improving Adversarial Robustness (2018, background): “, in [22]); we expect this tradeoff to be the subject of future research.”
Towards the first adversarially robust neural network model on MNIST (2018, background): “By “solved” we mean a model that reaches at least 99% accuracy (see accuracy-vs-robustness trade-off (Tsipras et al., 2018; Bubeck et al., 2018)) and whose adversarial examples carry semantic meaning to humans (by which we mean that they start looking like samples that could belong to either class).”
Excessive Invariance Causes Adversarial Vulnerability (2018, background): “The difference to human perception revealed by adversarial examples fundamentally questions which statistics the network use to base its decisions (Jo & Bengio, 2017; Tsipras et al., 2018).”
Adversarial Examples Improve Image Recognition (2019, background): “Though such trained models have several nice properties as described in [51, 47, 43], they cannot generalize well to clean images [23, 45].”
Smooth Adversarial Training (2020, background): “This phenomenon is referred to as no free lunch in adversarial robustness [65, 42, 62].”
On Regularization and Robustness of Deep Neural Networks (2018, background): “This can lead to gains in interpretability, as observed by Tsipras et al. (2018).”
Clean-Label Backdoor Attacks (2018, methodology): “We will construct these perturbations based on adversarially trained models following the methodology in Tsipras et al. (2018).”
Revisiting Adversarial Risk (2018, background): “A number of recent works try to explain the drop in standard accuracy in adversarially trained models [Fawzi et al., 2018, Tsipras et al., 2018].”
Intriguing properties of adversarial training (2019, result): “We note that this tradeoff is also observed in the prior work [20].”
Generalized No Free Lunch Theorem for Adversarial Robustness (2018, methodology): “The `∞ / sup case where q = ∞ (Tsipras et al., 2018) is particularly important: the corresponding threat model allows the adversary to separately increase or decrease each feature by an amount at most .”