Energy Attack: On Transferring Adversarial Examples

TL;DR

Energy Attack introduces a novel transfer-based black-box adversarial attack method that models the energy distribution of perturbations using PCA, achieving state-of-the-art results without gradient access.

Contribution

The paper presents a parameter-free, gradient-free black-box attack leveraging PCA to model energy distribution, enabling effective transferability across models and datasets.

Findings

Achieves state-of-the-art black-box attack performance.

Perturbation energy distribution transfers across models and datasets.

Does not require gradient approximation or model access.

Abstract

In this work we propose Energy Attack, a transfer-based black-box $L_\infty$-adversarial attack. The attack is parameter-free and does not require gradient approximation. In particular, we first obtain white-box adversarial perturbations of a surrogate model and divide these perturbations into small patches. Then we extract the unit component vectors and eigenvalues of these patches with principal component analysis (PCA). Base on the eigenvalues, we can model the energy distribution of adversarial perturbations. We then perform black-box attacks by sampling from the perturbation patches according to their energy distribution, and tiling the sampled patches to form a full-size adversarial perturbation. This can be done without the available access to victim models. Extensive experiments well demonstrate that the proposed Energy Attack achieves state-of-the-art performance in black-box attacks on various models and several datasets. Moreover, the extracted distribution is able to transfer among different model architectures and different datasets, and is therefore intrinsic to vision architectures.