A very promising direction in the field of black-box adversarial attacks are randomized search schemes for crafting adversarial examples [1, 23, 24]. Combining random search with specific update proposal distributions allows to achieve state-of-the-art black-box efficiency for different threat models such as` and `2 [1], `1 [25], `0, adversarial patches, and adversarial frames [24].

Since our controllers are functions of 2 inputs as described in the Section 3.3 we can illustrate the

Adversarial attacks based on randomized search schemes have obtained state-of-the-art results in black-box robustness evaluation recently.

In this supplementary document, we provide additional materials to supplement our main submission. In Section A, we talk about the societal impacts of our work In Section B, we provide detailed experimental settings as well as further evaluation results on CIFAR-10 and ImageNet. In Section D, we give the proofs w.r .t . In Section E, we give the proofs w.r .t . The proofs of Theorem 3 are given in Section F. In Section C, we provide the analysis and evaluation of decision-based attacks.

We tested front-end enhanced neural models where a frozen classifier was prepended by a differentiable and fully convolutional model with a skip connection. By training them using a small learning rate for about one epoch, we obtained models that retained the accuracy of the backbone classifier while being unusually resistant to gradient attacks including APGD and FAB-T attacks from the AutoAttack package, which we attributed to gradient masking. The gradient masking phenomenon is not new, but the degree of masking was quite remarkable for fully differentiable models that did not have gradient-shattering components such as JPEG compression or components that are expected to cause diminishing gradients. Though black box attacks can be partially effective against gradient masking, they are easily defeated by combining models into randomized ensembles. We estimate that such ensembles achieve near-SOTA AutoAttack accuracy on CIFAR10, CIFAR100, and ImageNet despite having virtually zero accuracy under adaptive attacks. Adversarial training of the backbone classifier can further increase resistance of the front-end enhanced model to gradient attacks. On CIFAR10, the respective randomized ensemble achieved 90.8$\pm 2.5$% (99% CI) accuracy under AutoAttack while having only 18.2$\pm 3.6$% accuracy under the adaptive attack. We do not establish SOTA in adversarial robustness. Instead, we make methodological contributions and further supports the thesis that adaptive attacks designed with the complete knowledge of model architecture are crucial in demonstrating model robustness and that even the so-called white-box gradient attacks can have limited applicability. Although gradient attacks can be complemented with black-box attack such as the SQUARE attack or the zero-order PGD, black-box attacks can be weak against randomized ensembles, e.g., when ensemble models mask gradients.

We show that when taking into account also the image domain $[0,1]^d$, established $l_1$-projected gradient descent (PGD) attacks are suboptimal as they do not consider that the effective threat model is the intersection of the $l_1$-ball and $[0,1]^d$. We study the expected sparsity of the steepest descent step for this effective threat model and show that the exact projection onto this set is computationally feasible and yields better performance. Moreover, we propose an adaptive form of PGD which is highly effective even with a small budget of iterations. Our resulting $l_1$-APGD is a strong white-box attack showing that prior works overestimated their $l_1$-robustness. Using $l_1$-APGD for adversarial training we get a robust classifier with SOTA $l_1$-robustness. Finally, we combine $l_1$-APGD and an adaptation of the Square Attack to $l_1$ into $l_1$-AutoAttack, an ensemble of attacks which reliably assesses adversarial robustness for the threat model of $l_1$-ball intersected with $[0,1]^d$.

Black-box adversarial attacks can fool image classifiers into misclassifying images without requiring access to model structure and weights. Recent studies have reported attack success rates of over 95% with less than 1,000 queries. The question then arises of whether black-box attacks have become a real threat against IoT devices that rely on cloud APIs to achieve image classification. To shed some light on this, note that prior research has primarily focused on increasing the success rate and reducing the number of queries. However, another crucial factor for black-box attacks against cloud APIs is the time required to perform the attack. This paper applies black-box attacks directly to cloud APIs rather than to local models, thereby avoiding mistakes made in prior research that applied the perturbation before image encoding and pre-processing. Further, we exploit load balancing to enable distributed black-box attacks that can reduce the attack time by a factor of about five for both local search and gradient estimation methods.

Adversarial attacks based on randomized search schemes have obtained state-of-the-art results in black-box robustness evaluation recently. However, as we demonstrate in this work, their efficiency in different query budget regimes depends on manual design and heuristic tuning of the underlying proposal distributions. We study how this issue can be addressed by adapting the proposal distribution online based on the information obtained during the attack. We consider Square Attack, which is a state-of-the-art score-based black-box attack, and demonstrate how its performance can be improved by a learned controller that adjusts the parameters of the proposal distribution online during the attack. We train the controller using gradient-based end-to-end training on a CIFAR10 model with white box access. We demonstrate that plugging the learned controller into the attack consistently improves its black-box robustness estimate in different query regimes by up to 20% for a wide range of different models with black-box access. We further show that the learned adaptation principle transfers well to the other data distributions such as CIFAR100 or ImageNet and to the targeted attack setting.