DANAA: Towards transferable attacks with double adversarial neuron attribution

TL;DR

DANAA introduces a novel attack method that enhances transferability of feature-level adversarial attacks by accurately estimating neuron importance through a double adversarial attribution approach, improving attack effectiveness across models.

Contribution

The paper proposes DANAA, a new method for more precise neuron importance estimation using adversarial attribution, boosting transferability of feature-level attacks.

Findings

DANAA achieves state-of-the-art transferability performance.

The method effectively identifies important neurons for attack transfer.

Experiments validate the superiority of DANAA on benchmark datasets.

Abstract

While deep neural networks have excellent results in many fields, they are susceptible to interference from attacking samples resulting in erroneous judgments. Feature-level attacks are one of the effective attack types, which targets the learnt features in the hidden layers to improve its transferability across different models. Yet it is observed that the transferability has been largely impacted by the neuron importance estimation results. In this paper, a double adversarial neuron attribution attack method, termed `DANAA', is proposed to obtain more accurate feature importance estimation. In our method, the model outputs are attributed to the middle layer based on an adversarial non-linear path. The goal is to measure the weight of individual neurons and retain the features that are more important towards transferability. We have conducted extensive experiments on the benchmark datasets to demonstrate the state-of-the-art performance of our method. Our code is available at: https://github.com/Davidjinzb/DANAA