Unsupervised Robust Domain Adaptation without Source Data

TL;DR

This paper introduces a method for unsupervised robust domain adaptation that does not require source data, leveraging pseudo-labels and contrastive loss to improve robustness and accuracy against adversarial attacks.

Contribution

It proposes a novel approach that transfers robust source models to target domains without source data, utilizing non-robust pseudo-labels and pair-wise contrastive loss.

Findings

Robust source models can be effectively transferred to target domains.

Using non-robust pseudo-labels enhances performance on clean and adversarial samples.

Achieves over 10% accuracy improvement on benchmark datasets.

Abstract

We study the problem of robust domain adaptation in the context of unavailable target labels and source data. The considered robustness is against adversarial perturbations. This paper aims at answering the question of finding the right strategy to make the target model robust and accurate in the setting of unsupervised domain adaptation without source data. The major findings of this paper are: (i) robust source models can be transferred robustly to the target; (ii) robust domain adaptation can greatly benefit from non-robust pseudo-labels and the pair-wise contrastive loss. The proposed method of using non-robust pseudo-labels performs surprisingly well on both clean and adversarial samples, for the task of image classification. We show a consistent performance improvement of over $10\%$ in accuracy against the tested baselines on four benchmark datasets.