Erosion Attack for Adversarial Training to Enhance Semantic Segmentation Robustness

TL;DR

This paper introduces EroSeg-AT, a novel adversarial training framework that generates more effective adversarial examples by disrupting semantic consistency, thereby significantly improving the robustness of segmentation models against attacks.

Contribution

The paper proposes EroSeg-AT, a vulnerability-aware adversarial training method that considers semantic relationships, enhancing robustness beyond existing global-based attack approaches.

Findings

EroSeg-AT outperforms existing attack methods in effectiveness.

Models trained with EroSeg-AT show increased robustness.

Experimental results validate the superiority of the proposed framework.

Abstract

Existing segmentation models exhibit significant vulnerability to adversarial attacks.To improve robustness, adversarial training incorporates adversarial examples into model training. However, existing attack methods consider only global semantic information and ignore contextual semantic relationships within the samples, limiting the effectiveness of adversarial training. To address this issue, we propose EroSeg-AT, a vulnerability-aware adversarial training framework that leverages EroSeg to generate adversarial examples. EroSeg first selects sensitive pixels based on pixel-level confidence and then progressively propagates perturbations to higher-confidence pixels, effectively disrupting the semantic consistency of the samples. Experimental results show that, compared to existing methods, our approach significantly improves attack effectiveness and enhances model robustness under adversarial training.