… construct a causal graph to model the generation process of adversarial examples and define
the adversarial distribution to formalize the intuition of adversarial attacks. From the causal …

… In Appendix ?? we also perform adversarial robustness tests on a natural image binary
classification task derived from CIFAR-10. Again deep CAMA out-performs a discriminative CNN …

… Besides, this paper also provides a fundamental viewpoint of the relationship between
adversarial robustness and causal intervention, which may guide the design of future defenders …

… we present a new alternative robustness auditing method that uses causal inference to measure
… largely falls into two categories: adversarial and natural/non-adversarial robustness. We …

… causal perspective. This section concludes with the introduction of an adversarial model for
attacks on causal … , to argue that adversarial robustness is generically unachievable. But as …

… We refer to this notion of robustness as the adversarial robustness of recourse. We study the
adversarial robustness of recourse from the lens of causality (Pearl, 2009). Causal recourse …

… Given the lack of rigorous definitions of non-adversarial robustness in the literature prior
to this review (discussed next), this causal framework acts as a useful frame of reference in …

… are able to fairly measure their adversarial robustness on three standard image classification
… decorrelate causal and confounder signals and their adversarial robustness. Additionally, …

… and adversarial robustness can be simultaneously defined through the lens of causality to
… When dealing with a causal structure underlying data, traditional metrics like the Euclidean …

… Using our proposed fair metric we establish a causal adversarial perturbation (CAP) set
to … fairness and adversarial robustness. Finally, we define a novel causal individual fairness …